d2.2 tracking technologies and social attitudes€¦ · of the deliverable, an update is due in...

Project co

© Copyright

the Sense4Us project

D2.2 Tracking

attitudes

Grant agreement no.:

Document Reference:

Dissemination Level:

Project co-funded by the European Union under the Seventh Framework Programme

© Copyright 2014 University of Southampton IT Innovation Centre and other partners of

the Sense4Us project

ing technologies and social

Project acronym:

Project full title: Data Insights for Policy Makers and Citizens

Grant agreement no.:

Responsible:

Contributors: All Partners in the SENSE4US Consortium

Document Reference:

Dissemination Level:

Version:

Date:

funded by the European Union under the Seventh Framework Programme

Innovation Centre and other partners of

technologies and social

SENSE4US

Insights for Policy Makers and Citizens

611242

Juri Papay

SENSE4US Consortium

D2.2

PU

FINAL

06/10/14

D2.2 Report that tracks technologies and social attitudes

History

Version Date

0.1 28/08

0.2 08/09

0.3 29/09

0.4 29/09

0.5 01/10

1.0 06/10

Report that tracks technologies and social attitudes

2

Date Modification reason Modified by

28/08/14 Initial draft Juri Papay

/09/14 Contributions from project

partners

Juri Papay

/09/14 Revision Juri Papay

/09/14 Draft reviewed by project

partners

Hansard Society and

GESI

01/10/14 Final edits Juri Papay

/10/14 Submitted version Juri Papay

Modified by

Juri Papay

Juri Papay

Juri Papay

Hansard Society and

GESIS

Juri Papay

Juri Papay


Table of contents

History ................................

Table of contents ................................

List of tables ................................

List of abbreviations ................................

Executive summary ................................

1 Introduction ................................

2 Document summarisation

2.1 Definition ................................

2.2 Automatic labelling

2.3 Topic extraction ................................

2.4 Semantic mapping

2.5 Semantic search................................

2.6 Document summarisation toolkits

3 Finding related information

3.1 Definition ................................

3.2 Data integration ................................

3.3 Data linking ................................

3.4 Information discovery

3.5 Data visualisation ................................

4 Policy impact simulation

4.1 Definition ................................

4.2 Modelling techniques

4.3 Modelling Tools ................................

5 Finding related public opinion

5.1 Information from social media

5.2 Sentiment analysis

6 Social attitudes ................................

6.1 Definition ................................

6.2 Engaging the public in policy making

6.3 Public perception of policy making process

7 Summary ................................

8 References ................................


3

contents

.............................................................................................................................

................................................................................................

................................................................................................

................................................................................................

................................................................................................

................................................................................................

Document summarisation .........................................................................................

................................................................................................

Automatic labelling ................................................................................................

................................................................................................

Semantic mapping ................................................................................................

................................................................................................

Document summarisation toolkits ................................................................

Finding related information ................................................................

................................................................................................

................................................................................................

................................................................................................

Information discovery ...............................................................................................

................................................................................................

Policy impact simulation .........................................................................................

................................................................................................

niques ................................................................................................

................................................................................................

Finding related public opinion ................................................................

Information from social media ................................................................

Sentiment analysis ................................................................................................

................................................................................................

................................................................................................

Engaging the public in policy making ................................................................

Public perception of policy making process ..............................................................

................................................................................................

................................................................................................

............................. 2

............................................. 3

.................................................... 4

......................................... 5

.......................................... 6

............................................. 7

......................... 9

..................................................... 9

...................................... 9

......................................... 10

..................................... 11

......................................... 12

............................................ 12

.................................................... 14

................................................... 14

......................................... 14

................................................ 15

............................... 16

...................................... 16

......................... 18

................................................... 18

................................ 18

......................................... 20

................................................. 21

.................................................. 21

..................................... 21

....................................... 23

................................................... 23

........................................ 23

.............................. 24

................................................ 25

............................................. 27


List of tables

No table of figures entries found.


4

figures entries found.


List of abbreviations

Abbreviation Description

ALOE Assisted Linked Data Consumption Engine

API Application Programming Interface

DoW Description of Work

LDA Latent Dirichlet Allocation

LOD Linked Open Data

Lemon Lexicon Model for Ontologies

RDF Resource Definition Framework

LOD Linked Open Data

OR Operational research

OWL Web Ontology Language

PM Project Month

PSM Problem Structuring Methods

RDF Resource

REST Representational state transfer

SOFIE Self-Organizing Frame

SPARQL Protocol and RDF Query Language

SVR Support Vector Regression

WP Work package


5

List of abbreviations

Description

Assisted Linked Data Consumption Engine

Application Programming Interface

Description of Work

Latent Dirichlet Allocation

Linked Open Data

Lexicon Model for Ontologies

Resource Definition Framework

Linked Open Data

Operational research

Web Ontology Language

Project Month – PM1 is the first month of the project, etc.

roblem Structuring Methods

Resource Definition Framework

Representational state transfer

Organizing Framework for Information Extraction

Protocol and RDF Query Language

Support Vector Regression

package

PM1 is the first month of the project, etc.


Executive summary

Sense4us is a three year project that was launched in October 2013 and co

Seventh Framework Programme (FP7

giving them tools to access a wide array of current data, take into account the views of

citizens on policy issues in real time and help them to better understand the implications of

proposed policies. Policy-makers at the EU, national (UK) and local (Germany) levels will be

engaged in the development of Sense4us to ensure the toolkit has the widest possib

application.

The main objective of the Sense4us

networking sites for the benefit

to information will also increase

to the transparency of the policy making process

project is the integration of

may refer to the same real world entity, the

subjective thoughts and opinions. The integration of information originating from these data

sources enables the composition of two complementary points of view of the same

and thus provides a better,

support in the decision making process.

The Description of Work (DoW

Task 2.2 – Tracking other

In order to keep the project relevant and to grasp opportunities where they arise,

research will be carried out and reports produced that take account of new

technologies, social attitudes towards technologies and

in the SENSE4US space. The outcome of this research will be fed into the decision

making process of the project.

In this deliverable we have

technologies and social attitudes relevant for the

report will be used for the assessment of various options that arise

architecture design. In this

areas, these are:

• Document summarisation

• Finding related information

• Policy impact simulation

• Finding related public opinion

• Social attitudes

These subject areas cover the main components of

of the alternatives which will be taken

of the deliverable, an update is due in month 24

the project.


6

Executive summary

project that was launched in October 2013 and co-

Seventh Framework Programme (FP7-ICT-2013-10). Sense4us aims to assist policy makers by


olicy issues in real time and help them to better understand the implications of

makers at the EU, national (UK) and local (Germany) levels will be

engaged in the development of Sense4us to ensure the toolkit has the widest possib

Sense4us project is to exploit the potential of open data

the benefit of policy makers and citizens. We expect that a w

increase the participation of citizens in public matters and contribute

the policy making process. One of the key elements

the integration of information from open data and social media. A

to the same real world entity, the first provides objective facts while the

nd opinions. The integration of information originating from these data

the composition of two complementary points of view of the same

a better, more in depth understanding of the subject area and

support in the decision making process.

DoW) specifies the requirements for Task 2.2 as follows:

Tracking other technologies and social attitudes (M03-33)


carried out and reports produced that take account of new

technologies, social attitudes towards technologies and the impact of other research


making process of the project.

have surveyed the main directions of research,

social attitudes relevant for the Sense4us project. The outcome of this

report will be used for the assessment of various options that arise during the

first instalment of the deliverable we describe five main su

summarisation

Finding related information

impact simulation

Finding related public opinion

These subject areas cover the main components of Sense4us project and provide

which will be taken into account during the design. This is the first version

update is due in month 24 that will summarise the lessons learnt during

-funded under the

10). Sense4us aims to assist policy makers by


olicy issues in real time and help them to better understand the implications of


engaged in the development of Sense4us to ensure the toolkit has the widest possible

the potential of open data and social

We expect that a wider access

in public matters and contribute

One of the key elements of the Sense4us

. Although the data

provides objective facts while the other

nd opinions. The integration of information originating from these data

the composition of two complementary points of view of the same problem

of the subject area and can also

as follows:


carried out and reports produced that take account of new

the impact of other research


research, key concepts,

project. The outcome of this

the functional and

describe five main subject

provide an overview

into account during the design. This is the first version

lessons learnt during


1 Introduction

Sense4us is a three year project that was launched in October 2013 and co

Seventh Framework Programme (FP7


citizens on policy issues in real time and help them to better understand the implications of

proposed policies. Policy-makers at the EU, national (UK) and local (Germany) levels will be

engaged in the development of Sense4us to ensure the toolkit has the widest poss

application.

One of the objectives of the

social networking sites for the benefit

access to information will contribute

supporting the policies. One of the key themes

information from open data and

world entity, the first provides

opinions. The integration of information originating from these data sources

composition of two complementary points of view of the same probl

better, more in depth understanding

making process. Other objective

modelling and simulation, data analytics and social networ

economic and social benefits to policy

These objectives in terms of software development can be translated into a set of tools

enable:

• the extraction of information from

• the automatic annotation

• the lexical analysis of sources

• the creation of policy models combining quantitative open data sources with

qualitative data from social media

• the simulation of likely impacts of a policy

• the tracking of discussion

In this deliverable we’ll review the state of art in the context of the above objectives

according to the requirements for Task 2.2 specified in the Description of Work (DoW):

Task 2.2 – Tracking other


research will carried out and reports produced that take account of new technologies,

social attitudes towards technologies and th

SENSE4US space. The outcome of this research will be fed into the decision making

process of the project.


7

r project that was launched in October 2013 and co-

Seventh Framework Programme (FP7-ICT-2013-10). Sense4us aims to assist policy makers by


policy issues in real time and help them to better understand the implications of


engaged in the development of Sense4us to ensure the toolkit has the widest poss

the Sense4us project is to exploit the potential of open data

for the benefit of policy makers and citizens. We expect that a wider

access to information will contribute to the policy making process by providing evidence for

One of the key themes of the Sense4us project is the integration of

data and social media. Although the data may refer to the same real

provides objective facts while the other subjective thoughts a

opinions. The integration of information originating from these data sources

composition of two complementary points of view of the same problem and thus provides a

understanding of the subject area and can also support in the decision

objectives of the Sense4us project are: to advance the fields of policy

modelling and simulation, data analytics and social network discussion dynamics, provide

onomic and social benefits to policy-makers on all governmental levels across Europe.

in terms of software development can be translated into a set of tools

the extraction of information from big data and open data sources;

the automatic annotation and linkage of homogeneous data;

exical analysis of sources;

the creation of policy models combining quantitative open data sources with

litative data from social media;

likely impacts of a policy; and

he tracking of discussion dynamics in social media.



Tracking other technologies and social attitudes (M03-33)



social attitudes towards technologies and the impact of other research in the


process of the project.

-funded under the

10). Sense4us aims to assist policy makers by


policy issues in real time and help them to better understand the implications of


engaged in the development of Sense4us to ensure the toolkit has the widest possible

the potential of open data and

of policy makers and citizens. We expect that a wider

by providing evidence for

the integration of

may refer to the same real

subjective thoughts and

opinions. The integration of information originating from these data sources enables the

em and thus provides a

of the subject area and can also support in the decision

to advance the fields of policy

k discussion dynamics, provide

all governmental levels across Europe.

in terms of software development can be translated into a set of tools that

the creation of policy models combining quantitative open data sources with





e impact of other research in the




technologies and social attitudes relevant for the

report will be used for the assessment of various options that arise

architecture design. In the first

areas, these are:

• Document summarisation

• Finding related information

• Policy impact simulation

• Finding related public opinion

• Social attitudes.

These subject areas cover the main components of the project and give an overview of the

alternatives which will be taken into account during the design. This is the first version of the

deliverable, the update is due in month 24

project.

The document is organised according to the identified subject areas.

document summarisation. T

found in Section 3. Section

techniques for discovering public opinion.

politics, in general, and to the p

findings of the report and makes recommendations for the


8

have surveyed the main directions of research,

technologies and social attitudes relevant for the Sense4us project. The outcome of this

report will be used for the assessment of various options that arise during the functional and

architecture design. In the first instalment of the deliverable we describe five main subject

Document summarisation;

Finding related information;

Policy impact simulation;

Finding related public opinion; and


taken into account during the design. This is the first version of the

deliverable, the update is due in month 24 that will summarise the lessons learnt during the

The document is organised according to the identified subject areas. Section

document summarisation. The issues related to Open Data and information

. Section 4 details policy impact simulation and Section

public opinion. Section 6 discusses social attitudes in relation to

and to the policy making process specifically. Section 7

findings of the report and makes recommendations for the Sense4us project.

research, key concepts,

project. The outcome of this

the functional and

ibe five main subject


taken into account during the design. This is the first version of the

ssons learnt during the

Section 2 describes

issues related to Open Data and information search can be

impact simulation and Section 5 outlines

iscusses social attitudes in relation to

7 summarises the

project.


2 Document summarisation

2.1 Definition

The purpose of document summarization

that contains the key points of

concise and comprehensive. The key topics contained in the document (or in multiple

documents) should be clearly identified and discusse

advantages of automatic document summarisa

human intervention and therefore it is unbiased. However this

accuracy and consistency before it is published

In general there are two techniques used for producing a summary, these are:

abstraction (Mani 2014). The extractive

or sentences in the original text

semantic representation of the text

creating a summary. In this case the

explicitly present in the original

research, but the majority of

There are several issues related to document summarisation

several publications (Mani 2014)

a) dealing with the thematic diversity of a large number of documents

b) combining the main themes with completeness

c) readability;

d) conciseness;

e) representing a wide

f) reflecting the logical structure of the main content

g) dividing the output into meaningful paragraphs

h) transition from general topic

i) consistent referencing of information sources

The detailed analysis of the above listed issues is beyond the

focus only on the themes that have been identified by the research partners

to be relevant in the context of the

a) automatic labelling;

b) topic extraction;

c) semantic mapping; and

d) semantic search.

In the following sections provide a short survey of

state of art.

2.2 Automatic labelling

Labelling is the first step for determining the topics

generic approach to automatic labelling


9

Document summarisation

document summarization is to reduce a large volume of text to a summary

that contains the key points of the original document. The produced summary must be both


ly identified and discussed from different perspectives.

automatic document summarisation is that it can produce an o

on and therefore it is unbiased. However this output must be assessed for

accuracy and consistency before it is published (Mani 2014).

there are two techniques used for producing a summary, these are:

. The extractive approach selects a subset of existing words, phrases,

or sentences in the original text for generating a summary. The abstractive approach builds a

of the text and then uses natural language generation techniques

In this case the produced summary might contain words

explicitly present in the original text. Abstractive methods represent a promising direction

the majority of summarisation tools still use extractive methods.

issues related to document summarisation, these have been

(Mani 2014), (Lloret 2010), (Nenkova 2011):

the thematic diversity of a large number of documents;

combining the main themes with completeness;

diversity of views on the topic;

logical structure of the main content;

dividing the output into meaningful paragraphs;

transition from general topics to more specific themes; and

consistent referencing of information sources.

The detailed analysis of the above listed issues is beyond the scope of this report and we’ll

focus only on the themes that have been identified by the research partners

context of the Sense4us project. These themes are:

; and

provide a short survey of these themes and describe the current

labelling

Labelling is the first step for determining the topics that a document addresses.

utomatic labelling uses primitive labels as described in

ge volume of text to a summary

The produced summary must be both


perspectives. One of the

tion is that it can produce an output without

output must be assessed for

there are two techniques used for producing a summary, these are: extraction and

approach selects a subset of existing words, phrases,

summary. The abstractive approach builds a

natural language generation techniques for

produced summary might contain words that are not

represent a promising direction of

extractive methods.

, these have been derived from

scope of this report and we’ll

focus only on the themes that have been identified by the research partners and considered

these themes and describe the current

document addresses. The most

as described in (Griffiths 2004),


(Blei, D. M, et al 2003). In this technique the

associated with a given topic.

Current research however,

interpreting the coherent meaning of a

approaches have explored the use of external

supporting automatic labelling of topics by deriving

(Lau, J.H., et al 2011), or graph

al proposed an unsupervised

topic model (Mei, Q., et al 2007)

problem involving the minimisation of

between the given topic and the candidate labels while maximising

between these two words.

selecting the “top n” terms

mutual information and conditional probabilities.

Another frequently used technique

as dictionaries, thesauruses

sources for automatic labelling

The authors derived candidate topic

algorithm (LDA 2014) and the hierarchy

initial candidate labels were

(Lau, J.H., et al 2011), described a case study where the

extracted from Wikipedia articles.

One of the promising approaches

Regression (SVR) model that enables t

recently, (Hulpus 2013) proposed

and graph centrality measures

2.3 Topic extraction

Topic extraction is a starting point for various tasks

sentiment analysis. One of the main challenges is understanding

ensuring “semantic coherence”

developed to enable topic extraction. These

for Standard Development Kits

survey the following tools:

a) Semantria’s (Semantria 2014)

b) Texalytics (Texalytics 2014)

c) AlchemyAPI (AlchemyAPI 2014)

d) Yahoo Content Analysis API

Semantria (Semantria 2014)

collection of documents. Semantria

matrix by using a deep learning algorithm.

formulation of queries, tag


10

In this technique the top words are ranked using the

associated with a given topic.

, indicates that selecting the top terms is not

meaning of a given topic (Mei, Q., et al 2007)

approaches have explored the use of external sources for example DBpedia

automatic labelling of topics by deriving so called candidate labels

or graph-based (Hulpus 2013) algorithms. The paper authored by Mei

proposed an unsupervised probabilistic algorithm for automatic assign

(Mei, Q., et al 2007). The proposed approach was defined as an optimisation

the minimisation of the Kullback–Leibler divergence (Kullback

given topic and the candidate labels while maximising the mutua

. It was proposed by (Lau, J.H., et al 2011) to

terms by using different ranking mechanisms including point

conditional probabilities.

used technique for automatic labelling utilises external data

es or specialised topic ontologies. Methods relying on external

labelling have been described by Magatti et al (Magatti,D., et

derived candidate topic labels using the Latent Dirichlet Allocation

the hierarchy obtained from the Google Directory service

initial candidate labels were extended with the Open Office English Thesaurus.

described a case study where the label candidates for

Wikipedia articles.

One of the promising approaches in this area is the construction of a Support Vector

that enables the ranking of label candidates (Support 2014)

proposed the use of structured data sources (for example

and graph centrality measures for generating semantic concept labels.

starting point for various tasks, for example text summarisation and

sentiment analysis. One of the main challenges is understanding the meaning of topics by

“semantic coherence” (Aletras 2013). Recently numerous t

topic extraction. These tools usually provide REST API and also

tandard Development Kits supporting popular programming languages.

(Semantria 2014);

(Texalytics 2014);

(AlchemyAPI 2014); and

Yahoo Content Analysis API (YahooAPI 2014).

(Semantria 2014) allows topics to be extracted that are relevant for

Semantria relies on Wikipedia’s ontology for constructing a concept

sing a deep learning algorithm. The tool offers a simple API that allows

tagging, grouping and structuring the collection of derived

the probabilities

not sufficient for

(Mei, Q., et al 2007). More recent

and Wikipedia for

candidate labels by using lexical

paper authored by Mei et

ing of labels in a

was defined as an optimisation

(Kullback-Leiber 2014)

the mutual information

to label topics by

including point-wise

external data sources such

Methods relying on external

(Magatti,D., et al 2009).

Latent Dirichlet Allocation (LDA)

obtained from the Google Directory service. Then the

ish Thesaurus. In their paper,

label candidates for topics were

a Support Vector

(Support 2014). More

se of structured data sources (for example DBpedia)

for example text summarisation and

the meaning of topics by

Recently numerous tools have been

and also endpoints

programming languages. In this section we

that are relevant for single or a

Wikipedia’s ontology for constructing a concept

simple API that allows the

collection of derived topics.


Semantria also provides sentiment analysis for the data extracted from the social media.

API can be accessed directly

customisation of output by editing the

Topic extraction in Textalytics

natural language processing techniques that produce morphological, syntactic and semantic

analysis. The tool produces different types of elements which are classified as

• Named entities: people, organizations, places, etc

• Concepts: significant keywords in the text

• Time expressions;

• Money expressions;

• URIs;

• Phone number expressions

• Other expressions: alphanumeric patterns

• Quotations; and

• Relationships.

The AlchemyAPI (AlchemyAPI 2014)

sentiment analysis of web pages, documents, tweets and photos.

network algorithms for the

tool is frequently is used in the business environment

customers, competitors, company operations, marketing, sales and new product

developments. AlchemyAPI provides language support for Python, PHP,

C/C++, C#(.NET) and Android OS.

The Yahoo Content Analysis API

of entities/concepts, categories, and r

entities and concepts by their overall relevance. These items are then linked

pages and annotated with tags

2.4 Semantic mapping

Semantic mapping can be described as a

another (Suchanek F. M. 2009)

by topics which are mapped into semantic relationships

by semantic nets or ontologies. The

process: extracting the topics

the topics to Open Data resources

relevant for Sense4us due to the fact that their functionality

aims of the project. These tools are:

• LODifier (LODifier 2014)

• SOFIE (Self-Organizing Framework for Information Extraction)

and


11

sentiment analysis for the data extracted from the social media.

API can be accessed directly from Excel via a plug-in. One of the big benefits

by editing the categories and tags in the configuration file

Textalytics (Texalytics 2014) is based on combining a number of complex


different types of elements which are classified as

Named entities: people, organizations, places, etc;

Concepts: significant keywords in the text;

Phone number expressions;

Other expressions: alphanumeric patterns;

(AlchemyAPI 2014) tool provides topic extraction, image taggin

sentiment analysis of web pages, documents, tweets and photos. AlchemyAPI uses

the linguistic and statistical processing of large volumes of text. This

sed in the business environment for extracting knowledge about the


AlchemyAPI provides language support for Python, PHP,

C/C++, C#(.NET) and Android OS.

Yahoo Content Analysis API (YahooAPI 2014) is a Web Service that allows

entities/concepts, categories, and relationships in the text. The service ranks the

epts by their overall relevance. These items are then linked

tags and relevant meta-data.

Semantic mapping can be described as a transformation of data from one

(Suchanek F. M. 2009). In the context of the Sense4us project the data is represented

by topics which are mapped into semantic relationships. These relationships can be

by semantic nets or ontologies. The task in this case can be described as a three stage

topics from the text, creating a semantic network of topics

Open Data resources. In this section we survey three tools that might be

due to the fact that their functionality a close match wi

aims of the project. These tools are:

(LODifier 2014);

Organizing Framework for Information Extraction) (Suchanek F. M. 2009)

sentiment analysis for the data extracted from the social media. The

benefits of this tool is the

in the configuration file.

is based on combining a number of complex


different types of elements which are classified as:

topic extraction, image tagging and

AlchemyAPI uses neural

volumes of text. This

for extracting knowledge about the


Ruby, Java, Perl,

Web Service that allows the detection

text. The service ranks the detected

epts by their overall relevance. These items are then linked to Wikipedia

one namespace into

of the Sense4us project the data is represented

. These relationships can be expressed

can be described as a three stage

semantic network of topics and linking

In this section we survey three tools that might be

a close match with the original

(Suchanek F. M. 2009);


• Lemon (Lexicon Model for Ontologies)

The main purpose of LODifier (as the name indicates) is converting unstructured natural

language into Linked Data (LODifier 2014)

disambiguation, Semantic Web vocabularies for extracting entities and

them. The output is translated into RDF representation where the individual nodes are linked

to DBpedia and WordNet. The key features of LODifier are:

surface variations, using a semantic graph

linking up the concepts and relations to the LOD cloud (for additional

I et al 2012).

The SOFIE system enables

bases (Suchanek F. M. 2009)

the discovered facts into ontologies

identifying the most probable meaning

patterns in a knowledge database. The key idea of SOPHIE is that the knowledge stored

existing ontology can be used for gathering

hypotheses. This feature enables not only a growth but also validation and re

ontology.

Lemon (Lexicon Model for Ontologies) is a higher level model

lexical resources to ontologies and thereby providing

(NLP) (McCrae 2011). The

(WordNet 2014) can be mapped into Lemon’s semantic network and thus enabling linking

with other ontologies.

2.5 Semantic search

The purpose of semantic search

intention of the user for improving the accuracy of output.

account various aspects, for example

natural language queries, concept

2012). The popular search engines such as Bing and Google also use elements of semantic

search for finding the most relevant information.

five techniques used in semantic search

• RDF Path Traversal;

• Keyword to Concept Mapping

• Graph Patterns;

• Logics - by using inference

• Fuzzy concepts, relations and

2.6 Document summarisation t

There are several text summarisation toolkits used in practice that represent interest for the

Sense4us project. These are the


12

Lemon (Lexicon Model for Ontologies) (McCrae 2011).


(LODifier 2014). LODifier uses deep semantic analysis, word

disambiguation, Semantic Web vocabularies for extracting entities and relatioships


The key features of LODifier are: abstracting away from linguistic

semantic graph for representing explicit structural information and

linking up the concepts and relations to the LOD cloud (for additional details see

enables the integration of the discovered data into existing knowledge

(Suchanek F. M. 2009). SOFIE can parse documents, extract ontological facts and li

the discovered facts into ontologies. Algorithms based on logical reasoning are used for

identifying the most probable meaning of words. This process relies on searching

patterns in a knowledge database. The key idea of SOPHIE is that the knowledge stored

existing ontology can be used for gathering new data and also for

enables not only a growth but also validation and re

Lemon (Lexicon Model for Ontologies) is a higher level model that enables

lexical resources to ontologies and thereby providing input for Natural Language

The paper also described that the existing RDF model of WordNet

can be mapped into Lemon’s semantic network and thus enabling linking

emantic search is to use the conceptual meaning of terms and also the

intention of the user for improving the accuracy of output. Semantic search

for example, context of search, location, intent, synonyms

concept matching etc. for providing relevant search results

The popular search engines such as Bing and Google also use elements of semantic

search for finding the most relevant information. In the literature we can find references to

five techniques used in semantic search (Makela 2005), these are:

Keyword to Concept Mapping;

inference based on OWL; and

concepts, relations and logics.

Document summarisation toolkits


Sense4us project. These are the Ultimate Research Assistant (Assistant 201


LODifier uses deep semantic analysis, word-sense

relatioships between


abstracting away from linguistic

cit structural information and

details see (Augenstein,

discovered data into existing knowledge

ontological facts and link

. Algorithms based on logical reasoning are used for

of words. This process relies on searching for text

patterns in a knowledge database. The key idea of SOPHIE is that the knowledge stored in an

also for reasoning about

enables not only a growth but also validation and restructuring of an

that enables the linking of

anguage Processing

that the existing RDF model of WordNet

can be mapped into Lemon’s semantic network and thus enabling linking

meaning of terms and also the

Semantic search takes into

synonyms of words,

relevant search results (John

The popular search engines such as Bing and Google also use elements of semantic

In the literature we can find references to


(Assistant 2014) and iResearch


Reporter (iResearch 2014). There are also tools specialised for processing news items, these

are: Newsblaster (Newsblaster 2014)

NewsInEssence (Radev 2005)

concept extraction, hierarchical concept clustering, text summarisation, taxonomy g

and various visualisation tools (


13

There are also tools specialised for processing news items, these

(Newsblaster 2014), NewsFeed Researcher (Researcher 2014)

(Radev 2005). Typical features of these tools are: text mining of documents,

concept extraction, hierarchical concept clustering, text summarisation, taxonomy g

sation tools (for example mind map generation).

There are also tools specialised for processing news items, these

(Researcher 2014),

. Typical features of these tools are: text mining of documents,

concept extraction, hierarchical concept clustering, text summarisation, taxonomy generation


3 Finding related information

3.1 Definition

There are several data resources considered

social media, academic research

objective is to extract informa

the Sense4us project Linked Open Data (LOD)

concept of Linked Open Data

from the Semantic Web (Berners

Web and the resources published o

suggested that this semantic augmentation could be

hyperlinks between Web resources.

the benefits of Linked Data as follo

“Linked Data provides a unifying data model

hyperlink-based data discovery and

In the context of the Sense4us project based on the communication with the

we can identify four main themes

a) integration of heterogeneous data sources

b) data linking;

c) information discovery

d) data visualisation.

These themes will be described in details in the

3.2 Data integration

The term “data integration”

and providing the user with a unified view of th

this respect are the different structure

With the growing volumes of data available on the Web, it has been recognised that the

traditional data integration approaches, which assume

sources, are inadequate for the challenges

“dataspaces” which represent an

dataspace is based on the following

a) providing suppor

b) offering various

of data sources; and

c) integration of data sources

A good survey of dataspace solutions can be found in

(Cafarella, M., et al 2011), described two

Crawler, which investigated the problem of structuring information published on the web.

The WebTables project compile

trying to combine the tables that are embedded in HTML pages.

Crawler attempted to extract data from the so called


14

Finding related information

data resources considered by the Sense4us project, these are: Open D

cademic research, policy-makers’ own data and policy documents

extract information that is relevant for a specific subject area. In the context of

Linked Open Data (LOD) is the main source of information

concept of Linked Open Data was introduced by (Berners-Lee 2006). This concept was

(Berners-Lee, T. et al 2001), which proposed an enhance

resources published on the Web with more expressive semantics

suggested that this semantic augmentation could be achieved by establishing and describing

perlinks between Web resources. Heath and Bizer (Heath 2011) in their paper summarised

of Linked Data as follows:

Linked Data provides a unifying data model a standardized data access mechanism

based data discovery and self-descriptive data”.

Sense4us project based on the communication with the

we can identify four main themes relevant for the usage of data resources, these are:

integration of heterogeneous data sources;

discovery; and

These themes will be described in details in the following sections.

” refers to the problem of combining data from

h a unified view of this data (Lenzerini 2002). The main issues

different structures, different formats and different dynamics


traditional data integration approaches, which assumed a limited number of relational data

sources, are inadequate for the challenges of the Web. This led to the emergence of

which represent an abstraction for data integration. The generic vision of a

dataspace is based on the following principles (Franklin, M., et al 2005):

support for a wide variety of data formats;

levels of service depending on the type of query and parameters

; and

data sources.

dataspace solutions can be found in (Hedeler, C., et al 2010)

described two Google projects: WebTables and Google Deep Web

which investigated the problem of structuring information published on the web.

compiled a large collection of databases by crawling the Web

trying to combine the tables that are embedded in HTML pages. The Google Deep Web

to extract data from the so called Deep Web that is made up

by the Sense4us project, these are: Open Data,

documents. The overall

tion that is relevant for a specific subject area. In the context of

source of information. The original

. This concept was derived

enhancement of the

expressive semantics. The authors

establishing and describing

in their paper summarised

zed data access mechanism,

Sense4us project based on the communication with the project partners

the usage of data resources, these are:

different sources

The main issues in

different dynamics of data.


a limited number of relational data

This led to the emergence of so called

abstraction for data integration. The generic vision of a

levels of service depending on the type of query and parameters

(Hedeler, C., et al 2010). In their paper,

WebTables and Google Deep Web

which investigated the problem of structuring information published on the web.

by crawling the Web and

Google Deep Web

made up by a large


number of Web forms. These two projects

Web that cannot be found by the

There are numerous approaches to

following four systems might be relevant:

a) Sindice + Sig.ma (Tummarello, G., et al 2010)

based search over semantic data collected from the Web

b) Information Workbench

Data application development. T

data sources, which are

c) LDIF (Schulte, A., et al 2011)

retrieved from the Web. Important

co-reference resolution

d) ALOE (Assisted Linked Data Consumption Engine

consumption engine.

relation to the increasingly complex Web of Data. This complexity

rapidly growing size of data, the diversity of vocabularies that describe the data and

the lack of a common

solution to address this problem by

schema and data-level mappings.

3.3 Data linking

The objective of data linking is to establish hyperlinks between related data

stored in different data sources.

Web browsers in a similar way

that in the case of the semantic search the users follow RDF links for navigating between data

sources. The RDF links can also be followed automatically by

search engines (Heath 2011)

Publishing data on the web is as

recommendations described in Tim Berners

(Berners-Lee 2006):

a) use URIs as names for things

b) use HTTP URIs so that people can look up those names

c) provide useful information, using the standards (RDF,

d) include links to other URIs

In relation to data linking there are several

are Microformats (Microformats 2014)

Microformats (Microformats 2014)

define a set of simple data formats that can be embedded into HTML via class attributes. The

data items that are included in HTML pages via Microformats do not have their own

identifier. This makes linking the data across documents and Web sites a major issue.

Linked Data initiative promotes a set of best practices for publishing and connecting

structured data on the Web.

RDF.


15

number of Web forms. These two projects tried to harvest the vast amount of data on the

eb that cannot be found by the traditional web crawlers.

are numerous approaches to data integration, in respect to the Sense4us project the

following four systems might be relevant:

(Tummarello, G., et al 2010) aim to provide light-

semantic data collected from the Web;

Information Workbench (Haase 2011) is intended as a platform to support Linke

Data application development. The system primarily targets the virtual integration of

data sources, which are accessed using query federation;

(Schulte, A., et al 2011) provides a suite of tools for the integration

retrieved from the Web. Important features of the system are data translation and

reference resolution; and

Assisted Linked Data Consumption Engine) (ALOE 2014) provides a linked data

consumption engine. Consumption of Linked Data is one of the

relation to the increasingly complex Web of Data. This complexity


common schema for knowledge bases. ALOE provides

solution to address this problem by using light-weight techniques for generating

level mappings.

The objective of data linking is to establish hyperlinks between related data

stored in different data sources. Once the data is linked it can be accessed by using Semantic

Web browsers in a similar way to the traditional web browsers. The difference is however

case of the semantic search the users follow RDF links for navigating between data

he RDF links can also be followed automatically by specialised robots and semantic

(Heath 2011).

Publishing data on the web is as easy as publishing HTML pages. However t

described in Tim Berners-Lee’s paper that allow the data

se URIs as names for things;

se HTTP URIs so that people can look up those names;

provide useful information, using the standards (RDF, SPARQL); and

include links to other URIs.

In relation to data linking there are several initiatives that Sense4us can benefit from, these

(Microformats 2014) and Linked Data (LinkedData 2014)

(Microformats 2014) is to extend the web with structured data. The idea is to



linking the data across documents and Web sites a major issue.


structured data on the Web. The key technologies used by Linked Data are URI, HTTP and

to harvest the vast amount of data on the

, in respect to the Sense4us project the

-weight keyword-

is intended as a platform to support Linked

virtual integration of

integration of data

data translation and

provides a linked data

Consumption of Linked Data is one of the relevant tasks in

relation to the increasingly complex Web of Data. This complexity is due to the


a semi-automatic

weight techniques for generating

The objective of data linking is to establish hyperlinks between related data items that are

Once the data is linked it can be accessed by using Semantic

. The difference is however,

case of the semantic search the users follow RDF links for navigating between data

robots and semantic

easy as publishing HTML pages. However there are general

the data to be found

can benefit from, these

2014). The aim of

is to extend the web with structured data. The idea is to



linking the data across documents and Web sites a major issue. The


ey technologies used by Linked Data are URI, HTTP and


3.4 Information discovery

Linked Data is typically published either as static RDF files on a web s

HTML files or via database access (

requirement for the publication

by HTTP clients. Linked Data

integrated by specialised web

LDSpider (Isele 2011)). Data can be consumed also by

currently in use by the given

typically Linked Data Browsers.

Hartig et al., (Hartig, O., et al 2009)

Some applications enable to

SPARQL endpoints which are identified in advance, see

are mainly similarity-based but they

consider to be relevant in the context of the Sense4us project are the following:

a) Silk (Volz, J. et al 2009)

b) LIMES (Ngomo, A. et al 2011)

c) SERIMI (Araújo 2011)

d) Amalgame (van Ossenbruggen 2011)

e) RiMOM (Li, J. et al 2009)

Another approach for improving

catalogues and meta-level descriptions

repositories. Examples of this concept are

catalogues.

3.5 Data visualisation

A suitable presentation of

(Halevy 2012). The visualisation is

allows not only easy navigation but also

amounts of data (Wills 2009)

recommend taking a closer look at the following initi

2010) and TWC portal (Ding 2011)

The LESS framework enables

based on parameterised data and queries

which allows flexible integration

The visualisation used in the

source ontology of linked data and

presentation of information.

implemented in the TWC porta

(Ding 2011). The presentation layer of

(TWC) and Linking Open Government Data


16

on discovery

Linked Data is typically published either as static RDF files on a web server, embedded into

HTML files or via database access (relational databases or RDF triple stores).

for the publication of Linked Data is that the URIs of resources

Linked Data can be consumed in different ways. It can be

specialised web-spiders for providing a consistent view o

Data can be consumed also by de-referencing only

given application. Linked Data applications using this

typically Linked Data Browsers. Techniques for on-the-fly dereferencing are presented in

(Hartig, O., et al 2009).

enable to federate the query process by sending complex SPARQL queries

which are identified in advance, see (Langegger 2008). Link Discovery Tools

based but they also apply machine-learning techniques

in the context of the Sense4us project are the following:

(Volz, J. et al 2009);

(Ngomo, A. et al 2011);

(Araújo 2011);

(van Ossenbruggen 2011); and

(Li, J. et al 2009).

Another approach for improving information discovery is based the creation

level descriptions that summarise the information stored in the

. Examples of this concept are (CKAN 2014), (VoId 2014) and

of linked data is one of the critical requirements as described in

. The visualisation is usually based on the hierarchical graph

allows not only easy navigation but also better understanding and summaris

(Wills 2009), (Ell 2011). In the context of the Sense4us project

a closer look at the following initiatives: LESS (Auer 2010)

(Ding 2011).

enables end-to-end integration of linked data from multiple sources

ised data and queries (Auer 2010). LESS also uses a template

flexible integration of different formats, such as text, plots, graphs, maps, etc.

the Fusion project (Araujo 2010) allows mappings between the

linked data and the application ontology for the processing and visual

presentation of information. Various demos of linked data visualization have been

implemented in the TWC portal which demonstrates examples of linked governmental data

. The presentation layer of the TWC LOGD portal (Tetherless World Constellation

(TWC) and Linking Open Government Data (LOGD)) uses mash-ups based on

erver, embedded into

relational databases or RDF triple stores). An essential

Is of resources can be looked up

can be crawled and

a consistent view of the data (see

only the URIs that are

application. Linked Data applications using this method are

fly dereferencing are presented in

complex SPARQL queries

Link Discovery Tools

learning techniques. Tools that we

in the context of the Sense4us project are the following:

creation of source

the information stored in the

and (DCAT 2014)

is one of the critical requirements as described in

the hierarchical graph structure that

and summarising of large

In the context of the Sense4us project we

(Auer 2010), Fusion (Araujo

end integration of linked data from multiple sources

template mechanism

ots, graphs, maps, etc.

mappings between the

application ontology for the processing and visual

linked data visualization have been

linked governmental data

Tetherless World Constellation

ups based on the Google


visualization API. In the Fusion Tables

integration and visualization


17

Fusion Tables project Google also provides a framework for data

integration and visualization (Gonzalez 2010).

Google also provides a framework for data


4 Policy impact simulation

4.1 Definition

The academic literature describes d

alternatives based on the

objective of this process is to reduce

can be selected. This is an iterative process

assumptions, parameters and

enable the policy maker to predict the consequences of

With these tools the user can

analysed from the perspective

and groups of individuals (Jiwani 2010)

One of the key components of

several activities. In relation to modelling the following activities can be identified

(Druckenmiller et al 2009), (Mintzberg 1994)

a) describing the scenario

b) identifying the key variables,

c) describing the stakeholders and

d) specifying the assumptions

e) specifying the mathematical model or a

f) interpreting of the output

g) providing feedback to the modelling process

4.2 Modelling techniques

In the literature we can find various approaches to modelling

techniques can be found in

techniques have been considered

a) Operational research

b) Multi-Criteria Decision Analysis

c) Multi-attribute utility theory

d) Portfolio decision analysis

e) Game theory.

Operational Research (OR) is one of the

and decision support. There are numerous tools developed in OR that can be used for the

evaluation of alternative scenarios, see details in

2014). There are also advanced

taking into account the qualitative

referred to as Problem Structuring Methods (PSM)

works with well-structured problems, the PSM

unstructured and contain many uncertainties. Typical features

multiple actors, multiple perspectives, conflicting interests, important intangi

number of uncertainties (Mingers 2004)


18

Policy impact simulation

The academic literature describes decision making as a process of identifying and choosing

es based on the considerations and preferences of the decision maker. The

objective of this process is to reduce the number of alternatives so that a favourable

can be selected. This is an iterative process which requires the continuous

and outcomes (Harris 2012). Decision support tools and techniques

enable the policy maker to predict the consequences of a policy before they are enacted

can carry out “what if” studies that allow a given policy

from the perspective of different stakeholders and for the likely effects on

(Jiwani 2010), (Mitchell 2009).

One of the key components of a decision support system is modelling whi

In relation to modelling the following activities can be identified

(Mintzberg 1994):

the scenarios;

key variables, ranges of variables and uncertainties;

stakeholders and preferences;

the assumptions of the model;

mathematical model or a simulation tool;

output; and

to the modelling process.

techniques

In the literature we can find various approaches to modelling, an overview

(Mingers 2004). In the context of Sense4us project the following

have been considered:

Operational research and Problem Structuring Methods (PSM);

Criteria Decision Analysis;

attribute utility theory;

tfolio decision analysis; and

Operational Research (OR) is one of the well-established methodologies used for modelling


ation of alternative scenarios, see details in (Weimer and Vining 2005),

advanced techniques that aim to extend the traditional OR

qualitative data in the modelling process. These techniques are

Problem Structuring Methods (PSM) (Mingers 2004). The traditional OR mainly

structured problems, the PSM on the other hand tackles problems which are

unstructured and contain many uncertainties. Typical features of unstructured problems are

multiple actors, multiple perspectives, conflicting interests, important intangi

(Mingers 2004). A detailed account of PSM is beyond the scope of

making as a process of identifying and choosing

and preferences of the decision maker. The

number of alternatives so that a favourable outcome

continuous checking of initial

pport tools and techniques

policy before they are enacted.

a given policy to be

likely effects on society

which incorporates

In relation to modelling the following activities can be identified

n overview of various

the context of Sense4us project the following

used for modelling


, (Gass 2011), (OR

techniques that aim to extend the traditional OR methods by

data in the modelling process. These techniques are

raditional OR mainly

problems which are

unstructured problems are

multiple actors, multiple perspectives, conflicting interests, important intangibles and large

A detailed account of PSM is beyond the scope of


this report, therefore we just mention the

1989):

• Strategic options development and analysis

• Soft systems methodology (SSM)

• Strategic choice approach (SCA)

• Drama theory (Rosenhead 1996)

• Viable systems model (VSM)

• Decision conferencing

The information available for modelling can be either

quantitative information can be integrated

qualitative information is essential

(Sterman 2000). Needless to say it is harder to deal with qualitative

generally no accepted rules for obtaining, analysis and interpreting this type of information

The problems involved with qualitative information can be summarised by

quotation from (Luna-Reyes et al 2003)

“Although there is general agreement about the importance of qualitative data during the

development of a system dynamics model, there is not a clear description about how or w

to use it. The lack of an integrated set of procedures to obtain and analyze qualitative

information creates, among several possible problems, a gap between the problem modeled

and the model of the problem.

Another approach to modelling is based on

method is widely used in

leading to the achievement of multiple

of this technique is that the alternatives are defined at the beginning of the modelling process

and they do not change. However

and they can also change as the modelling work progresses and

available (Franco L. A. 2011)

A recently developed method based Multi

approach for integrating both

(Danielson, M., et al 2006). The

selection of a portfolio of alternatives

resource constraints, for example

Game theory in general terms can be described as the

conflict and cooperation between intelligent rational decision

theory is used in numerous fields

etc. This theory assumes that games are well defined

The objective is to design game


19

this report, therefore we just mention the typical examples of PSM techniques

Strategic options development and analysis (SODA) (Eden 2000);

Soft systems methodology (SSM) (Connell 2001);

trategic choice approach (SCA) (Friend 2001);

(Rosenhead 1996);

Viable systems model (VSM) (Harnden 1990); and

Decision conferencing (Phillips 1989).

available for modelling can be either quantitative or

can be integrated into mathematical expressions

is essential for the conceptualisation and formulation of the model

to say it is harder to deal with qualitative data

accepted rules for obtaining, analysis and interpreting this type of information

The problems involved with qualitative information can be summarised by

Reyes et al 2003):

Although there is general agreement about the importance of qualitative data during the

development of a system dynamics model, there is not a clear description about how or w



and the model of the problem.”

Another approach to modelling is based on Multi-Criteria Decision Analysis (MCDA). This

in complex decision problems for evaluating various alternatives

the achievement of multiple objectives (Keeney R.L 1993). One of the assumptions

is that the alternatives are defined at the beginning of the modelling process

and they do not change. However, in practice these alternatives are often difficult to specify

also change as the modelling work progresses and more information becomes

(Franco L. A. 2011).

A recently developed method based Multi-attribute Utility Theory proposes an alternative

approach for integrating both quantitative and qualitative data into the same model

The main idea of Portfolio Decision Analysis (PDA)

alternatives (instead of a single one only) which satisfies

resource constraints, for example costs, budget etc. (Fasth 2012).

in general terms can be described as the "the study of mathematical models

conflict and cooperation between intelligent rational decision-makers" (Myerson

theory is used in numerous fields such as economics, politics, management, defence, biology

theory assumes that games are well defined in terms of players, rules and outcomes.

game strategies that can produce the desired outcome.

techniques (J. Rosenhead

or qualitative. The

mathematical expressions while the

conceptualisation and formulation of the model

data since there are

accepted rules for obtaining, analysis and interpreting this type of information.

The problems involved with qualitative information can be summarised by the following

Although there is general agreement about the importance of qualitative data during the

development of a system dynamics model, there is not a clear description about how or when



iteria Decision Analysis (MCDA). This

various alternatives

. One of the assumptions

is that the alternatives are defined at the beginning of the modelling process

in practice these alternatives are often difficult to specify

nformation becomes

proposes an alternative

into the same model

nalysis (PDA) is based on the

(instead of a single one only) which satisfies certain

mathematical models of

(Myerson 1991). This

mics, politics, management, defence, biology

players, rules and outcomes.

produce the desired outcome.


4.3 Modelling Tools

There are numerous design

for assessing the consequences

tools, the following categories

• Information Control

information and knowledge

• Paradigm Models -

model;

• Simulation Models -

• Ways of Choosing - reducing the number of alternatives

• Representation Aids

In the context of the case study

we can also highlight an

UK Department for Energy and Climate Change

outcomes of different energy policy scenarios, and the impact of their choices on the

climate (calculator 2014)

and demonstrates the impact of real scientific data. The

whether the data included in

emissions reduction. A f

called the Global Calculator

energy, land and food systems on the climate


20

tools that allow conceptual and simulation model

consequences of various policy decisions. According to this

egories can be identified (tools 2014):

Information Control - gathering, storage, retrieval, and organisation of data,

information and knowledge;

paradigms, frameworks or perspectives for conceptualising

models that provide answers to "what if" questions

reducing the number of alternatives; and

Representation Aids - visualisation of data or problem space.

context of the case study (“green energy”) investigated by the

an award winning tool “2050 carbon calculator”

Department for Energy and Climate Change. The tool allows users to simulate the


(calculator 2014). This climate calculator produces emission reduction pathway

and demonstrates the impact of real scientific data. The tool also allows

the data included in policy documents supports the long term objectives on

A further example of a similar tool, extended for the entire

Global Calculator which allows a user to estimate the impact

energy, land and food systems on the climate (Global 2014).

models to be created

this survey of these

gathering, storage, retrieval, and organisation of data,

conceptualising the

" questions;

Sense4us project

developed by the

users to simulate the


reduction pathways

tool also allows a user to check

the long term objectives on

urther example of a similar tool, extended for the entire world is

the impact of the world’s


5 Finding related public opinion

5.1 Information from social media

The objective of social media

and interpret the discussions and the exchange of information related to specific topics.

information found in social media

evidence required for policy

concerns about using this source of information in the context of policy making or

assessing the public reaction

information about the participants of political discussions on the social media

2012). The recent data collected by

percentage of users generate most of the discussions, in real terms about

than 36% of all collected tweets. It is also important to

generated by professional “opinion forming” organisation

tanks rather than individual citizens. This finding, although it is only a preliminary study

many questions regarding the use o

public opinion.

5.2 Sentiment analysis

The objective of sentiment analysis is to extract evidence from social media that can support

the policy making process and to get feedback on certain topics

of public debate, analysing the dynamics of discussions and determining

regard to specific policies.

mechanisms and tools for harnessing this rich content

The main directions of this

(Beevolve 2012):

a) collecting data on the evolution of political debates

b) monitoring the change of

c) modelling the evolution of a political

d) providing information from open data sources to

e) extracting evidence

Recent studies have shown a strong correlation between the online opinion

their offline attitude towards political issues in the

public opinion on various social media platforms

makers. By collecting information we can get a picture

policies as well as positive and negative argu

Business of Government “Next Four Y

turning to social media to discuss their political views

some social networking sites have managed to attr

of all Internet traffic (Traffic 2014)

content and an easy and quick way to reach out to millions of people. While social networks


21

Finding related public opinion

Information from social media

media analysis in the context of Sense4us is to collect, analyse, track

and interpret the discussions and the exchange of information related to specific topics.

social media has certainly got the potential to improve the quality of

policy document drafting (Leavy 2013). There are however

concerns about using this source of information in the context of policy making or

assessing the public reaction to various policies. One of the concerns

information about the participants of political discussions on the social media

data collected by (Fernandez, M. et al 2012) show that

generate most of the discussions, in real terms about 6%

than 36% of all collected tweets. It is also important to mention that most of these tweets are

generated by professional “opinion forming” organisations such as news agencies and think

tanks rather than individual citizens. This finding, although it is only a preliminary study

the use of information coming from the social media

analysis is to extract evidence from social media that can support

and to get feedback on certain topics. This involves

, analysing the dynamics of discussions and determining

policies. Data mining of social media, along with the

mechanisms and tools for harnessing this rich content are the subject of intensive research

this research have been outlined by (Agarwal, A., et al 2011)

the evolution of political debates;

monitoring the change of sentiment;

modelling the evolution of a political debate;

information from open data sources to enrich the political

from debates and summarising the key arguments

have shown a strong correlation between the online opinion

towards political issues in the real world (Hussain 2011)

public opinion on various social media platforms can provide important insights for policy

. By collecting information we can get a picture of the citizens’ attitude

policies as well as positive and negative arguments. According to The IBM Center for The

Next Four Years: Citizen Participation” more and more people are

turning to social media to discuss their political views (Citizen Participation 2012)

some social networking sites have managed to attract millions of active users, and

(Traffic 2014). Social media have therefore become a


in the context of Sense4us is to collect, analyse, track

and interpret the discussions and the exchange of information related to specific topics. The

to improve the quality of

There are however, several

concerns about using this source of information in the context of policy making or for

One of the concerns is the lack of

information about the participants of political discussions on the social media (Wheeler

show that only a small

6% generate more

that most of these tweets are

such as news agencies and think

tanks rather than individual citizens. This finding, although it is only a preliminary study, raises

social media for assessing

analysis is to extract evidence from social media that can support

involves the monitoring

, analysing the dynamics of discussions and determining public opinion in

along with the development of

are the subject of intensive research.

(Agarwal, A., et al 2011) and

debate; and

from debates and summarising the key arguments.

have shown a strong correlation between the online opinion of individuals and

(Hussain 2011). Hence tracking

important insights for policy

the citizens’ attitudes towards

According to The IBM Center for The

” more and more people are

(Citizen Participation 2012). Today,

active users, and about 20%

media have therefore become a rich source of



are known to be highly accepted among younger generations, they

with older generations (Madden 2014)

Most existing approaches to sentiment analysis focus on classifying the individual

subjective (positive or negative) or objective (neutral). They can be categorised

approaches and lexicon-based

classifiers from various combinations of features such as wor

Speech (POS) tags (Agarwal, A., et al 2011)

tweets, punctuations, etc.

accuracy, however, training data is usually

evolving subject domains for example

Lexicon-based methods use a

excellent etc. for calculating the overall sentiment

on sentiment dictionaries are

(Wilson 2005) and SentiStrength

Another approach to sentiment analysis is based on

processing techniques capturing

The results produced by conceptual semantic approaches

produced by syntactical approaches, however

knowledge bases (Cambria 2013)

Sentiment analysis was one of the key components of two previous projects

2014) and ROBUST (ROBUST 2014)

discussion analysis methods

type of content and social features

the ROBUST project (ROBUST 2014)

models to larger communities

on capturing and understanding the state of a


22

ghly accepted among younger generations, they are also gaining popularity

(Madden 2014).

Most existing approaches to sentiment analysis focus on classifying the individual

subjective (positive or negative) or objective (neutral). They can be categorised

based techniques. Supervised methods are based on training

classifiers from various combinations of features such as word “n-grams” (Pak 2010)

(Agarwal, A., et al 2011) and tweets syntax features such as

s, etc. (Kouloumpis 2011). These methods can achieve

owever, training data is usually difficult to obtain especially for continuously

for example in Twitter (Liu 2010).

based methods use a dictionary of opinionated words for example

excellent etc. for calculating the overall sentiment (Bollen 2011). Examples of methods based

sentiment dictionaries are SentiWordNet (sentiwordnet 2014), MPQA subjectivity lexicon

SentiStrength (sentistrength 2014).

Another approach to sentiment analysis is based on ontologies and natural language

capturing the conceptual representations of words (Saif, H. et al 2012)

conceptual semantic approaches are better than the output

syntactical approaches, however they are often limited by the size an

(Cambria 2013).

was one of the key components of two previous projects

(ROBUST 2014). In the WeGov project (WeGov 2014)

discussion analysis methods were implemented. These methods allowed the

content and social features and also identified the behavioural roles

(ROBUST 2014) the behavioural models were extended

to larger communities. In both WeGov and ROBUST projects, the analysis was focused

understanding the state of a discussion at any given point in time.

are also gaining popularity

Most existing approaches to sentiment analysis focus on classifying the individual words into

subjective (positive or negative) or objective (neutral). They can be categorised by supervised

are based on training

(Pak 2010), Part-Of-

d tweets syntax features such as hash-tags, re-

hods can achieve around 80%

to obtain especially for continuously

ionated words for example; great, sad,

Examples of methods based

MPQA subjectivity lexicon

natural language

(Saif, H. et al 2012).

are better than the output

size and scope of

was one of the key components of two previous projects WeGov (WeGov

(WeGov 2014) several generic

the analysis of the

roles of participants. In

extended by applying these

, the analysis was focused

discussion at any given point in time.


6 Social attitudes

6.1 Definition

In this section we give a short overview of

the policy making process in general. We also discuss the ways

contribute to policy making by participating in debates, providing feedback and suggestions

to decision makers. The term, “

captured by the following quotation

“Social attitude is a person or

Social attitude measures the

(Answers 2014)

The heterogeneity of individuals and groups within

reactions to policies. Some policies

groups or individuals.

6.2 Engaging the public

The participation of the public in politics is regularly measure

statistics (Politics 2014) the overall interest of public

the UK the following quotation can illustrate this point

“There is a common concern that the

politics.” (Politics 2014)

In terms of policy making it is important to consider the behavioural characteristics

groups and individuals. Regarding behaviour

2014):

• people display ‘bounded rationality’

have not got the time for complex calculations and reasoning

• people’s preferences

preferences are not in line with long term goals

• people exhibit reciprocity and value fairness

For engaging the public in policy related debates

called “4E” requirements should be addressed

• Encouraging - giving the right signals,

target audience responds

• Enabling - putting in place the capacity, infrastructure, services, skills, guidance,

information and support needed

• Engaging - getting people involved


23

In this section we give a short overview of the social attitudes of the public to policies and to

the policy making process in general. We also discuss the ways that the public can engage


The term, “social attitude”, is a wide concept and perhaps can be

captured by the following quotation:

Social attitude is a person or groups reaction to other people, races, cultures, ideas, or traits.

Social attitude measures the person or groups like or dislike toward a certain subject.

The heterogeneity of individuals and groups within society will naturally result in

. Some policies may change the behaviour and preferences

Engaging the public in policy making

public in politics is regularly measured and evaluated.

the overall interest of public the in politics is low.

quotation can illustrate this point.

“There is a common concern that the British public is increasingly becoming disengaged with

In terms of policy making it is important to consider the behavioural characteristics

groups and individuals. Regarding behaviour, the following observations can be made

bounded rationality’ since they do not have access to the data and

the time for complex calculations and reasoning;

preferences change over time – it also often happens that short term

preferences are not in line with long term goals;

eople exhibit reciprocity and value fairness.

For engaging the public in policy related debates according to DEFRA’s study the following so

” requirements should be addressed (Collier 2010):

giving the right signals, incentives and disincentives ensuring that

target audience responds;

in place the capacity, infrastructure, services, skills, guidance,

information and support needed;

people involved; and

public to policies and to

the public can engage and


is a wide concept and perhaps can be

races, cultures, ideas, or traits.

person or groups like or dislike toward a certain subject.”

will naturally result in different

and preferences of social

d and evaluated. According to

in politics is low. In the context of

British public is increasingly becoming disengaged with

In terms of policy making it is important to consider the behavioural characteristics of social

the following observations can be made (Politics

since they do not have access to the data and

it also often happens that short term

FRA’s study the following so

entives and disincentives ensuring that the

in place the capacity, infrastructure, services, skills, guidance,


• Exemplifying - leading

6.3 Public perceptions of

The latest figures from the 2014 Hansard Society report

the following (Audit 2014):

• 50% of the British public say they are

• 50% of the British public say they feel they know

about politics;

• 49% of the British public say they are certain to vote in the event of an immediate

general election; and

• 33% of the British public think that the system of

well.

Despite the relatively high level of interest in politics the

large extent has been disconnected from t

from the public into the policy making

from the public regarding the

benefits of the rapid development of social

greater say in the policy making process.

makers from becoming engaged with

summarised as follows (Collier 2010)

a) citizens feel that they are not

made;

b) they also think that they do not have much

c) most of the information which the public receive and react to is delivered via

traditional media such as television and news

d) the news items are interpreted by

editorial bias; and

e) the supporting evidence and

makes the validation of policies diff


24

ing by example and sharing responsibility.

of the policy making process

latest figures from the 2014 Hansard Society report into political engagement

50% of the British public say they are 'very' or 'fairly' interested in politics

50% of the British public say they feel they know 'a great deal' or

49% of the British public say they are certain to vote in the event of an immediate

; and

33% of the British public think that the system of governing in this country works

Despite the relatively high level of interest in politics the policy making process

disconnected from the public. In other words there is little

policy making process. It is also difficult to measure the feedback

the policies that have already been implemented

benefits of the rapid development of social media, it is expected that the public will have a

er say in the policy making process. However, there are still barriers that prevent policy

makers from becoming engaged with the citizens in political debates. These aspects can be

(Collier 2010):

feel that they are not well enough informed about the way decisions are

that they do not have much influence over decision making

ost of the information which the public receive and react to is delivered via

a such as television and news;

interpreted by professional journalists and often published with

supporting evidence and data are not presented or are difficult to access. This

makes the validation of policies difficult if not impossible.

into political engagement highlight

interested in politics;

' or 'a fair amount'

49% of the British public say they are certain to vote in the event of an immediate

governing in this country works

icy making process itself to a

public. In other words there is little if any input

process. It is also difficult to measure the feedback

implemented. As one of the

it is expected that the public will have a

However, there are still barriers that prevent policy

These aspects can be

well enough informed about the way decisions are

influence over decision making;

ost of the information which the public receive and react to is delivered via

n published with

difficult to access. This


7 Summary


tools that we consider to be relevant for

range of topics that involve open data sources, harvesting information

visualisation of semantic data,

making. Producing a comprehensive

scope of the presented report

key components of the Sense4us system. The themes surveyed in this report included:

document summarisation,

related public opinion and social attitudes

numerous references to the literature that capture the main directions of research, the key

concepts and tools.

The intention of this report

technologies and provide pointers to alternatives that can be used

specification and architecture design.

updated version will be published

during the project.

The findings of the report for

a) Document summarisation

publishing, is becoming

research promises

comprehensive and also

been a significant progress in this direction, however there are

provide a critical assessment of the quality

humans.

b) Finding related information

drafting policies. If we look

mainly text based. Althou

in databases and difficult to

The Linked Open Data initia

sources and making them public. However, it must be said that until data sea

becomes an integral

hidden.

c) Policy impact simulation

the investigation of

of the model. The m

terms of a numerical simulation

can be achieved automatically. Although some rudimentary

extracted directly from the

of this model most certainly

d) Finding related public information

mood of the society in relation to vario


25

In this deliverable we have surveyed the main directions of research, the

tools that we consider to be relevant for the Sense4us project. The project itself covers a wide

range of topics that involve open data sources, harvesting information from

visualisation of semantic data, and the analysis of social and demographics aspects in policy

making. Producing a comprehensive survey of all these areas is far beyond the capacity and

scope of the presented report. Therefore we have focused only on the themes that


, finding related information, policy impact simulation,

and social attitudes. In each of these subject areas we have provided


The intention of this report was to highlight the opportunities, take account

and provide pointers to alternatives that can be used for

specification and architecture design. This is the first instalment of the deliverable, the

published in month 24, which will summarise the

for individual subject areas can be summarised as follows:

ocument summarisation, in light of the rapidly increasing volume

becoming one of the key research areas of computer science

research promises the automatic production of digests that are

and also reflect the key points of input documents. Recently there has

progress in this direction, however there are only a

provide a critical assessment of the quality of output against the digest

Finding related information that can aid reasoning and decision making is essential for

If we looked at the information that is published on the web

mainly text based. Although there is an increasing amount of data, it is mainly hidden

in databases and difficult to find since the data is not indexed by the

The Linked Open Data initiative is trying to improve this situation by integrating data

sources and making them public. However, it must be said that until data sea

an integral part of standard web browsers the wealth of data will remain

Policy impact simulation aims to develop analytical and simulation models that allow

various "what if" studies by changing the numerical

he main issue in this respect is interpreting the policy document

terms of a numerical simulation model. This is a complex task and it is unlike

automatically. Although some rudimentary model can probably be

from the text, further conceptualisation, refinement

most certainly will require human intervention.

Finding related public information in social media has got the potential to gauge the

the society in relation to various events and policies. However, it must also

key concepts and

The project itself covers a wide

from social media, the

analysis of social and demographics aspects in policy

survey of all these areas is far beyond the capacity and

herefore we have focused only on the themes that cover the


ation, policy impact simulation, finding

. In each of these subject areas we have provided


take account of new

for the functional

of the deliverable, the

summarise the lessons learnt

as follows:

volume of digital

of computer science. This

that are concise,

s. Recently there has

a few studies that

against the digests produced by

aid reasoning and decision making is essential for

is published on the web, it is

of data, it is mainly hidden

the search engines.

improve this situation by integrating data

sources and making them public. However, it must be said that until data search

wealth of data will remain

simulation models that allow

" studies by changing the numerical input values

the policy document(s) in

task and it is unlikely that it

model can probably be

inement and validation

in social media has got the potential to gauge the

s and policies. However, it must also


be said that this information can be distorted since according to recent studies most

of the traffic on certain topics is generated by a small fraction of participants. One of

the main problems with harvesting social m

profile of participants which makes any sound statistical studies about the real public

opinion difficult to conduct.

e) Social attitudes to politics

organisations. The reason

find out whether there is

interested in politics

making itself. There are o

provide input to the policy making process. The consequence is that there is a lack of

understanding why certain policies have been accepted and what was the

justification for their introduction


26



the main problems with harvesting social media is the lack information about the


opinion difficult to conduct.

to politics are regularly monitored and measured by numerous

The reason is to measure the level of public engagement

find out whether there is acceptance of a policy. Although some

interested in politics they have little if any involvement in the process of policy

re are only a few opportunities (if any) to express opinion

the policy making process. The consequence is that there is a lack of

why certain policies have been accepted and what was the

justification for their introduction.



edia is the lack information about the


are regularly monitored and measured by numerous

is to measure the level of public engagement and also to

some citizens are

involvement in the process of policy

to express opinions or to

the policy making process. The consequence is that there is a lack of

why certain policies have been accepted and what was the


8 References

Aday, S. Blogs and bullets: New media in contentious politics.

http://www.usip.org/files/resources/pw65.pdf.

Agarwal, A., et al. “Sentiment analysis of twitter data.”

workshop on languages in

AlchemyAPI. 2014. http://www.alchemyapi.com/) .

Aletras, N. and Stevenson, M. “ Evaluating topic coherence using distributional semantics on

Computational Semantics (IWCS 2013).”

Conference . Potsdam: Association for Computational Linguistics, 2013. 13

ALOE. 2014. http://aksw.org/projects/aloe.

Answers. 2014. http://www.answers.com/Q/What_is_the_definition_of_social_attitude.

Araujo, S. et al. “Fusion – visually exploring and eliciting

Semantic Web – ISWC 2010.

Araújo, S., et al. “SERIMI

interlinking.” Proceedings of the 7th International Workshop on

2011.

Assistant, Ultimate Research. 2014. http://www.ultimate

assistant.com/GenerateResearchReport.asp.

Audit. 2014. http://www.auditofpoliticalengagement.org/.

Auer, S. et al. “Less template

Semantic Web: Research and Applications.

224.

Augenstein, I et al. “LODifier: generating linked data from unstructured text.”

Proceedings of the 9th international conference on The Semanti

applications. Berlin: Springer

Beevolve. An Exhaustive Study of Twitter Users Across the World.

http://www.beevolve.com/twitter

Berners-Lee, T. 2006. http://www.w3.org/DesignIssues/LinkedData.ht

Berners-Lee, T. et al. “The Semantic Web.”

Blei, D. M, et al . “Latent dirichlet allocation.”

993–1022.

Bollen, J. et al. “Twitter mood predicts the stock market.”

no. 1 (2011): 1-8.

Cafarella, M., et al. “Structured Data on the Web.”

(2011): 72-79.


27

Blogs and bullets: New media in contentious politics.

http://www.usip.org/files/resources/pw65.pdf.

“Sentiment analysis of twitter data.” In proceedings of the acl 2011

workshop on languages in social media . 2011. 30–38.

2014. http://www.alchemyapi.com/) .


Computational Semantics (IWCS 2013).” In Proceedings of the 10th International

Potsdam: Association for Computational Linguistics, 2013. 13

ALOE. 2014. http://aksw.org/projects/aloe.


visually exploring and eliciting relationships in linked data.”

ISWC 2010. Springer Berlin / Heidelberg, 2010. 1–15.

Araújo, S., et al. “SERIMI - resource description similarity, RDF instance matching and

Proceedings of the 7th International Workshop on Ontology Matching.

Assistant, Ultimate Research. 2014. http://www.ultimate

assistant.com/GenerateResearchReport.asp.

Audit. 2014. http://www.auditofpoliticalengagement.org/.

Auer, S. et al. “Less template-based syndication and presentation of linked data.”

Semantic Web: Research and Applications. Springer Berlin / Heidelberg, 2010. 211

Augenstein, I et al. “LODifier: generating linked data from unstructured text.”

Proceedings of the 9th international conference on The Semantic Web: research and

Berlin: Springer-Verlag, 2012. 210-224 .

An Exhaustive Study of Twitter Users Across the World.

http://www.beevolve.com/twitter-statistics.

Lee, T. 2006. http://www.w3.org/DesignIssues/LinkedData.html.

Lee, T. et al. “The Semantic Web.” Scientific American 284 (2001): 34

Blei, D. M, et al . “Latent dirichlet allocation.” Journal of Machine Learning Research

Bollen, J. et al. “Twitter mood predicts the stock market.” Journal of Computational Science

Cafarella, M., et al. “Structured Data on the Web.” Communications of the ACM

Blogs and bullets: New media in contentious politics. 2010.

In proceedings of the acl 2011


In Proceedings of the 10th International

Potsdam: Association for Computational Linguistics, 2013. 13–22.


relationships in linked data.” The

15.

resource description similarity, RDF instance matching and

Ontology Matching.

Assistant, Ultimate Research. 2014. http://www.ultimate-research-

f linked data.” The

Springer Berlin / Heidelberg, 2010. 211–

Augenstein, I et al. “LODifier: generating linked data from unstructured text.” ESWC'12

c Web: research and

An Exhaustive Study of Twitter Users Across the World. 2012.

284 (2001): 34-43.

Journal of Machine Learning Research, 3 2003:

Journal of Computational Science 2,

Communications of the ACM 54, no. 2


calculator, Climate. 2014. https://www.gov.uk/2050

Cambria, E. “An introduction to concept

Computing and Its Applications.

“Citizen Participation.” 2012. http://www.businessofgovernment.org/blog/business

government/next-four

CKAN. 2014. http://thedatahub.org/.

Collier, A., et al. “Understanding and influencing behaviours: a review of social research,

economics and policy making in Defra.” February 2010.

http://archive.defra.gov.uk/evidence/series/documents/unde

behaviour-discuss.pdf.

Connell, N. “Evaluating soft OR: Some reflections on an apparently unsuccessful

implementation using a soft systems methodology (SSM) based approach.”

the Operational Research Society

Danielson, M., et al. “Cross

Proceedings of the 28th International Conference Information Tech. Interfaces.

2006.

DCAT. 2014. http://www.w3.org/egov/wiki/Data_Catalog_Vocabulary.

de Vries, B. et al. “Understanding Design Through Design Support Tools.”

the Netherlands 2005.

214.

Ding, L., et al. “Twc logd: A portal for linked open government data ecosystems.”

Semantics: Science, Services and Agents on the World Wide Web.

333.

Druckenmiller et al. “Agent

Situation Formulation.”

(2009).

Eden, C. “On evaluating the performance of GSS:Furthering the debate.”

Operational Research

Ell, B., et al. “Labels in the web of data.”

Berlin/Heidelberg, 2011. 162

Fasth, T., Larsson, A. “Portfolio Decision Analysis in Vague Domains.”

IEEM. IEEE, 2012. 61

Fernandez, M. et al. “Using Social Media To Inform Policy Making: To whom are we listening?

.” In European Conference on Social Media: ECSM

Franco L. A., Montibeller G. “Problem structuring for multicriteria decision analysis

interventions.” In Wiley encyclopedia of operations research and management

science, by Cochran et al. (ed). USA: Wiley, 2011.


28

calculator, Climate. 2014. https://www.gov.uk/2050-pathways-analysis.

Cambria, E. “An introduction to concept-level sentiment analysis.” Advances in Soft

Computing and Its Applications. Springer, 2013. 478–483.

“Citizen Participation.” 2012. http://www.businessofgovernment.org/blog/business

four-years-citizen-participation.

CKAN. 2014. http://thedatahub.org/.



http://archive.defra.gov.uk/evidence/series/documents/understand

discuss.pdf.


implementation using a soft systems methodology (SSM) based approach.”

the Operational Research Society, 2001: 150–160.

lson, M., et al. “Cross-disciplinary research in analytic decision support systems.”

Proceedings of the 28th International Conference Information Tech. Interfaces.

DCAT. 2014. http://www.w3.org/egov/wiki/Data_Catalog_Vocabulary.

“Understanding Design Through Design Support Tools.” Design Research in

the Netherlands 2005. Eindhoven: Eindhoven University of Technology, 2005. 205

Ding, L., et al. “Twc logd: A portal for linked open government data ecosystems.”

: Science, Services and Agents on the World Wide Web. Elsevier, 2011. 325

Druckenmiller et al. “Agent-Based Collaborative Approach to Graphing Causal Maps for

Situation Formulation.” Journal of the Association for Information Systems

Eden, C. “On evaluating the performance of GSS:Furthering the debate.” European Journal of

Operational Research, 2000: 218–222.

Ell, B., et al. “Labels in the web of data.” The Semantic Web – ISWC 2011.

Berlin/Heidelberg, 2011. 162–176.

, Larsson, A. “Portfolio Decision Analysis in Vague Domains.” Proceedings of IEEE

IEEE, 2012. 61-65.


European Conference on Social Media: ECSM, 174-182. ePub , 2012.


Wiley encyclopedia of operations research and management

, by Cochran et al. (ed). USA: Wiley, 2011.

Advances in Soft

“Citizen Participation.” 2012. http://www.businessofgovernment.org/blog/business-



rstand-influence-


implementation using a soft systems methodology (SSM) based approach.” Journal of

disciplinary research in analytic decision support systems.”

Proceedings of the 28th International Conference Information Tech. Interfaces. IEEE,

Design Research in

Eindhoven: Eindhoven University of Technology, 2005. 205-

Ding, L., et al. “Twc logd: A portal for linked open government data ecosystems.” Web

Elsevier, 2011. 325–

Based Collaborative Approach to Graphing Causal Maps for

Journal of the Association for Information Systems 10, no. 3

European Journal of

ISWC 2011. Springer

Proceedings of IEEE


ePub , 2012.


Wiley encyclopedia of operations research and management


Franklin, M., et al. “From da

management.” SIGMOD Record

Friend, J. “The strategic choice approach.” In

Revisited: Problem Structuring Methods for

Mingers, J. (Eds.) Rosenhead, 115

Fung, A., Shkabatur, J. iral Engagement: Fast, Cheap, and Broad, but Good for Democracy?

2012. http://www.archonfung.net/docs/articles/2012/ViralEng

Gass, S. I. “"George B. Dantzig". Profiles in Operations Research.”

Operations Research & Management Science

Global. Global. 2014. http://www.globalcalculator.org/.

Gonzalez, H. et al. “Google fusion ta

the cloud.” Proceedings of the 1st ACM symposium on Cloud computing.

175–180.

Griffiths, T.L, Steyvers, M. “Finding scientific topics.”

Haase, P., et al. “The Information Workbench as a Self

Applications.” Workshop on Consuming Linked Data (COLD 2011).

2011.

Halevy, A. Y. “Towards an ecosystem of structured data on the web.”

International Conference on Extending Database Technology.

Hansard. “Audit of Political Engagement.” 2014. http://www.hansardsociety.org.uk/wp

content/uploads/2014/04/Audit

Harnden, R. “The languaging of models: The understanding and communication of models

with particular reference to Stafford Beer's cybernetic model of organization

structure.” Systems Practice

Harris, R. Introduction to Decision Ma

Hartig, O., et al . “Executing SPARQL Queries over the Web of Linked Data.”

Semantic Web Conference.

Heath, T. and Bizer, C. Linked Data: Evolving the Web into a Global Data Space.

Claypool, 2011.

Hedeler, C., et al. “Dataspaces.” In

Springer-Verlag, 2010.

Hulpus, I., et al. “Unsupervised graph

the sixth ACM international conference on

New York: ACM, 2013. 465

Hussain, M., and Howard, P. “The role of digital media.”

(2011): 35–48.

iResearch. 2014. http://iresearch


29

Franklin, M., et al. “From databases to dataspaces: a new abstraction for information

SIGMOD Record 34, no. 4 (2005): 27-33.

Friend, J. “The strategic choice approach.” In Rational Analysis for a Problematic World

Revisited: Problem Structuring Methods for Complexity,Uncertainty and Conflict

Mingers, J. (Eds.) Rosenhead, 115–150. Chichester: Wiley, 2001.

iral Engagement: Fast, Cheap, and Broad, but Good for Democracy?

2012. http://www.archonfung.net/docs/articles/2012/ViralEngagement5.pdf.

Gass, S. I. “"George B. Dantzig". Profiles in Operations Research.” International Series in

Operations Research & Management Science, 2011: 217–240.

2014. http://www.globalcalculator.org/.

Gonzalez, H. et al. “Google fusion tables: data management, integration and collaboration in

Proceedings of the 1st ACM symposium on Cloud computing.

Griffiths, T.L, Steyvers, M. “Finding scientific topics.” PNAS, 101(suppl. 1), 2004: 5228

l. “The Information Workbench as a Self-Service Platform for Linked Data

Workshop on Consuming Linked Data (COLD 2011). Bonn: ISWC 2011,

Halevy, A. Y. “Towards an ecosystem of structured data on the web.” Proceedings of the 15th

ational Conference on Extending Database Technology. EDBT ’12, 2012. 1

Hansard. “Audit of Political Engagement.” 2014. http://www.hansardsociety.org.uk/wp

content/uploads/2014/04/Audit-of-Political-Engagement-11-2014.pdf.



Systems Practice 3, no. 3 (1990): 289–302.

Introduction to Decision Making. VirtualSalt, 2012.

Hartig, O., et al . “Executing SPARQL Queries over the Web of Linked Data.”

Semantic Web Conference. 2009. 293-309.

Linked Data: Evolving the Web into a Global Data Space.

Hedeler, C., et al. “Dataspaces.” In Search Computing, by S., Brambilla, M. Ceri, 114

Verlag, 2010.

Hulpus, I., et al. “Unsupervised graph-based topic labelling using dbpedia.”

the sixth ACM international conference on Web search and data mining, WSDM ’13.

New York: ACM, 2013. 465–474.

Hussain, M., and Howard, P. “The role of digital media.” Journal of democracy

2014. http://iresearch-reporter.com/.

tabases to dataspaces: a new abstraction for information

Rational Analysis for a Problematic World

Complexity,Uncertainty and Conflict, by J.,

iral Engagement: Fast, Cheap, and Broad, but Good for Democracy?

agement5.pdf.

International Series in

bles: data management, integration and collaboration in

Proceedings of the 1st ACM symposium on Cloud computing. ACM, 2010.

, 2004: 5228–5235.

Service Platform for Linked Data

Bonn: ISWC 2011,

Proceedings of the 15th

EDBT ’12, 2012. 1-2.

Hansard. “Audit of Political Engagement.” 2014. http://www.hansardsociety.org.uk/wp-

2014.pdf.



Hartig, O., et al . “Executing SPARQL Queries over the Web of Linked Data.” International

Linked Data: Evolving the Web into a Global Data Space. Morgan &

, by S., Brambilla, M. Ceri, 114-134.

based topic labelling using dbpedia.” In Proceedings of

Web search and data mining, WSDM ’13.

Journal of democracy 22, no. 3


Isele, R., et al. “LDSpider: An open

Proceedings of 9th International Semantic Web Conference (ISWC 2010).

Jiwani, G.N. Uncovering the Unknown of Government Policy Decision

Levels: Multiple Case Study.

John, Tony. What is Semantic Search?

What-Semantic-Search.aspx.

Keeney R.L, Raiffa H. Decisions with multiple objectives: preferences and value tradeoffs.

Cambridge University Press

Kouloumpis, E. et al. “Twitter sentiment analysis: The good the bad and the omg!”

Proceedings of the ICWSM.

Kullback-Leiber. 2014. http://en.wikipedia.org/wiki/Kullback%E2%80%93Leibler_divergence.

Langegger, A. et al. “A semantic web middleware for virtual data integration on the web.”

Proceedings of the 5th European semantic web conference on The semantic web:

research and applications.

Lau, J.H., et al . “Automatic labelling of topic mode

Meeting of the Association for Computational Linguistics: Human Language

Technologies - Volume 1.

Linguistics, 2011. 1536

LDA. 2014. http://en.wikipedia.org/wi

Leavy, J. Social Media and Public Policy, What is the evidence? .

report, 2013.

Lenzerini, M. “Data integration: A theoretical perspective.”

symposium on Principles

Li, J. et al. “RiMOM: A Dynamic Multistrategy Ontology Alignment Framework.”

Transactions on Knowledge and Data Engineering.

1218-1232.

Liesiö, J., et al. “Preference Programming for Robust Portfolio Modeling.”


LinkedData. 2014. http://linkeddata.org/.

Liu, B. “Sentiment analysis and subjectivity.” In

N. Indurkhya and F. J. Damerau (eds). 2010.

Lloret, E., Palomar, M. “Challenging Issues of Automatic Summarization: Relevance Detection

and Quality-based Evaluation.”

LODifier. 2014. http://www.aifb.k

Lourenço, J., et al. “PROBE –

Evaluation.” Decision Support Systems


30

“LDSpider: An open-source crawling framework for the Web of Linked Data.”

Proceedings of 9th International Semantic Web Conference (ISWC 2010).

Uncovering the Unknown of Government Policy Decision-making Process at Senior

Levels: Multiple Case Study. University of Washington, 2010.

What is Semantic Search? 2012. http://www.techulator.com/resources/5933

Search.aspx.

Decisions with multiple objectives: preferences and value tradeoffs.

Cambridge University Press: Cambridge, 1993.


Proceedings of the ICWSM. Barcelona, 2011.

Leiber. 2014. http://en.wikipedia.org/wiki/Kullback%E2%80%93Leibler_divergence.

semantic web middleware for virtual data integration on the web.”


research and applications. Springer-Verlag, 2008. 493-507.

Lau, J.H., et al . “Automatic labelling of topic models.” In Proceedings of the 49th Annual


Volume 1. Stroudsburg, PA, USA: Association for Computational

Linguistics, 2011. 1536–1545.

LDA. 2014. http://en.wikipedia.org/wiki/Latent_Dirichlet_allocation.

Social Media and Public Policy, What is the evidence? . Alliance for Useful Evidence

Lenzerini, M. “Data integration: A theoretical perspective.” 21st ACM SIGMOD

symposium on Principles of database systems. Madison: ACM, 2002. 233

Li, J. et al. “RiMOM: A Dynamic Multistrategy Ontology Alignment Framework.”

Transactions on Knowledge and Data Engineering. IEEE Computer Society, 2009.

Liesiö, J., et al. “Preference Programming for Robust Portfolio Modeling.” European Journal of

Operational Research 181, no. 3 (2000): 1488-1505.

LinkedData. 2014. http://linkeddata.org/.

Liu, B. “Sentiment analysis and subjectivity.” In Handbook of Natural Language Processing

N. Indurkhya and F. J. Damerau (eds). 2010.


based Evaluation.” Informatica 34 (2010): 29–35.

LODifier. 2014. http://www.aifb.kit.edu/web/LODifier/en.

– A Multicriteria Decision Support System for Portfolio Robustness

Decision Support Systems 54, no. 1 (2012): 534-550.

source crawling framework for the Web of Linked Data.”

Proceedings of 9th International Semantic Web Conference (ISWC 2010). 2011.

making Process at Senior

2012. http://www.techulator.com/resources/5933-

Decisions with multiple objectives: preferences and value tradeoffs.


Leiber. 2014. http://en.wikipedia.org/wiki/Kullback%E2%80%93Leibler_divergence.

semantic web middleware for virtual data integration on the web.”


In Proceedings of the 49th Annual


Stroudsburg, PA, USA: Association for Computational

Alliance for Useful Evidence

21st ACM SIGMOD-SIGACT-SIGART

Madison: ACM, 2002. 233-246.

Li, J. et al. “RiMOM: A Dynamic Multistrategy Ontology Alignment Framework.” IEEE

IEEE Computer Society, 2009.

European Journal of

l Language Processing, by


A Multicriteria Decision Support System for Portfolio Robustness


Luna-Reyes et al. “Collecting and analyzing qualitative data for system

and models.” System Dynamics Review

Madden, M. “Older Adults and Social Media, Pew Internet.” 2014.

http://pewinternet.org/Reports/2010/Older

Magatti,D., et al. “Automatic labeling of topics.”

International Conference on Intelligent Systems Design and Applications, ISDA ’09.

Washington: IEEE Computer Society, 2009. 1227

Makela, E. Survey of Semantic Search

http://www.seco.tkk.fi/publications/2005/makela

Mani, I. Summarization Evaluation: An Overview.

http://research.nii.ac.jp/ntcir/workshop/OnlineProceedings2/sum

McCrae, J., Spohr,D., Cimiano, P. “Lin

web with lemon.” In

Heidelberg: Springer, 2011.

Mei, Q., et al. “Automatic labeling of multinomial topic models.”

ACM SIGKDD international conference on Knowledge discovery and data mining.

York: ACM, 2007. 490

Microformats. 2014. http://microformats.org.

Mingers, J., Rosenhead,J. “Problem structuring methods in action.”


Mintzberg, H. The Rise and Fall of Strategic Planning.

Mitchell, B. Policy-making process.

Making-Process.html.

Myerson, R.B. Game Theory: Analysis of Conflic

Nenkova, A., Kathleen, K. “Automatic Summarization.”

103–233.

Newsblaster. 2014. http://www1.cs.columbia.edu/~sable/research/hlt

Ngomo, A. et al. “LIMES A Time

of Data.” IJCAI. 2011. 2312

OR. What is Operations Research?

Operations-Research.

Pak, A., Paroubek, P. “Twitter as a corpus for sentiment analy

Proceedings of LREC 2010.

Phillips, L. “People-centered group decision support.” In

Support Systems, by Doukidis et al, 208

“Politics.” 2014. http://bsa-30.natcen.ac.uk/read


31

Reyes et al. “Collecting and analyzing qualitative data for system dynamics: methods

System Dynamics Review 19, no. 4 (2003): 271-296.


http://pewinternet.org/Reports/2010/Older-Adults-and-Social-Media.aspx.

Magatti,D., et al. “Automatic labeling of topics.” In Proceedings of the 2009 Ninth


Washington: IEEE Computer Society, 2009. 1227–1232.

Survey of Semantic Search Research.

http://www.seco.tkk.fi/publications/2005/makela-semantic-search-2005.pdf.

Summarization Evaluation: An Overview.

http://research.nii.ac.jp/ntcir/workshop/OnlineProceedings2/sum-mani.pdf.

McCrae, J., Spohr,D., Cimiano, P. “Linking lexical resources and ontologies on the semantic

web with lemon.” In The Semantic Web: Research and Applications

Heidelberg: Springer, 2011.

Mei, Q., et al. “Automatic labeling of multinomial topic models.” In Proceedings of the 13th

KDD international conference on Knowledge discovery and data mining.

York: ACM, 2007. 490–499.

Microformats. 2014. http://microformats.org.

Mingers, J., Rosenhead,J. “Problem structuring methods in action.” European Journal of

Operational Research, 2004: 530–554.

The Rise and Fall of Strategic Planning. New York: Prentice-Hall, 1994.

making process. 2009. http://www.waterencyclopedia.com/Oc

Process.html.

Game Theory: Analysis of Conflict. Harvard University Press, 1991.

Nenkova, A., Kathleen, K. “Automatic Summarization.” Information Retrieval

Newsblaster. 2014. http://www1.cs.columbia.edu/~sable/research/hlt-blaster.pdf.

Ngomo, A. et al. “LIMES A Time-Efficient Approach for Large-Scale Link Discovery on the Web

2011. 2312-2317.

What is Operations Research? 2014. https://www.informs.org/About-INFORMS/What

Research.

Pak, A., Paroubek, P. “Twitter as a corpus for sentiment analysis and opinion mining.”

Proceedings of LREC 2010. Valletta, 2010.

centered group decision support.” In Knowledge-Based Management

, by Doukidis et al, 208–224. Chichester: Ellis-Horwood, 1989.

30.natcen.ac.uk/read-the-report/politics/introduction.aspx.

dynamics: methods


Media.aspx.

In Proceedings of the 2009 Ninth


Research. 2005.

2005.pdf.

Summarization Evaluation: An Overview. 2014.

mani.pdf.

king lexical resources and ontologies on the semantic

The Semantic Web: Research and Applications, 245–259.

In Proceedings of the 13th

KDD international conference on Knowledge discovery and data mining. New

European Journal of

Hall, 1994.

2009. http://www.waterencyclopedia.com/Oc-Po/Policy-

Harvard University Press, 1991.

Information Retrieval 5, no. 2-3 (2011):

blaster.pdf.

Scale Link Discovery on the Web

INFORMS/What-is-

sis and opinion mining.”

Based Management

Horwood, 1989.

report/politics/introduction.aspx.


Puron-Cid, G., et al. “IT-Enabled Policy Analysis: New Technologies, Sophisticated Analysis and

Open Data for Better Government Decisions.”

International Conferen

Radev, D., et al. “Newsinessence: Summarizing Online News Topics.”

ACM - The digital society

Researcher, NewsFeed. 2014. http://newsfeedresearcher.com/.

ROBUST. 2014. http://www.robust

Rosenhead. “What is the problem. An introduction to problem structuring methods.”

Interfaces 26, no. 6 (1996): 117

Rosenhead, J. Rational Analysis for a Problematic.

Saif, H. et al. “Semantic sentime

ISWC, 2012.

Salo, A., et al. “Improving Resource Allocation with Portfolio Decision Analysis.”

Nov/Dec 2011: 23-26.

Schulte, A., et al. “LDIF - Linked Data Integration Framework.”

Data (COLD 2011). Bonn: ISWC 2011, 2011.

Semantria. 2014. http://www.sementria.com.

sentistrength. 2014. http://sentistrength.wlv.ac.uk/.

sentiwordnet. 2014. http://sentiwordnet.isti.cnr.it/.

Sterman, J. Systems Thinking and

2000.

Suchanek F. M., Sozio, M., Weikum, G. “Sofie: a self

extraction.” Proceedings of the 18th international conference on World wide web

2009: 631–640.

Support. 2014. http://en.wikipedia.org/wiki/Support_vector_machine.

Texalytics. 2014. http://textalytics.com/core/topics

tools. 2014. http://www.ifm.eng.cam.ac.uk/research/dstools/#2b.

“Traffic.” 2014. http://www.comscore.com/.

Tumasjan, A., et al. “Predict

political sentiment.”

weblogs and social media.

Tummarello, G., et al. “Sig. ma: Live views on the Web of Data.”

Services and Agents on the World Wide Web


32

Enabled Policy Analysis: New Technologies, Sophisticated Analysis and

Open Data for Better Government Decisions.” Proceedings of the 13th Annual

International Conference on Digital Government Research. 2012. 97-106.

Radev, D., et al. “Newsinessence: Summarizing Online News Topics.” Communications of the

The digital society, 2005: 95-98.

Researcher, NewsFeed. 2014. http://newsfeedresearcher.com/.

://www.robust-project.eu/.


26, no. 6 (1996): 117–131.

Rational Analysis for a Problematic. Chichester: Wiley, 1989.

Saif, H. et al. “Semantic sentiment analysis of twitter.” 11th Int. Semantic Web Conf.

Salo, A., et al. “Improving Resource Allocation with Portfolio Decision Analysis.”

26.

Linked Data Integration Framework.” Workshop on Consuming Linked

Bonn: ISWC 2011, 2011.

2014. http://www.sementria.com.

sentistrength. 2014. http://sentistrength.wlv.ac.uk/.

sentiwordnet. 2014. http://sentiwordnet.isti.cnr.it/.

Systems Thinking and Modeling for a Complex World. Boston: Irwin/McGraw

Suchanek F. M., Sozio, M., Weikum, G. “Sofie: a self-organizing framework for information

Proceedings of the 18th international conference on World wide web

2014. http://en.wikipedia.org/wiki/Support_vector_machine.

2014. http://textalytics.com/core/topics-info.

tools. 2014. http://www.ifm.eng.cam.ac.uk/research/dstools/#2b.

“Traffic.” 2014. http://www.comscore.com/.

Tumasjan, A., et al. “Predicting elections with twitter: What 140 characters reveal about

political sentiment.” In Proceedings of the fourth international aaai conference on

weblogs and social media. 2010. 178-185.

Tummarello, G., et al. “Sig. ma: Live views on the Web of Data.” Web Semantics: Science,

Services and Agents on the World Wide Web 8, no. 4 (2010): 355-364.

Enabled Policy Analysis: New Technologies, Sophisticated Analysis and

Proceedings of the 13th Annual

106.

Communications of the


11th Int. Semantic Web Conf. Boston:

Salo, A., et al. “Improving Resource Allocation with Portfolio Decision Analysis.” Analytics,

Workshop on Consuming Linked

Boston: Irwin/McGraw-Hill,

organizing framework for information

Proceedings of the 18th international conference on World wide web,

ing elections with twitter: What 140 characters reveal about

In Proceedings of the fourth international aaai conference on

emantics: Science,

364.


van Ossenbruggen, J. et al.

Theory and Practice of Digital Libraries.

VoId. 2014. http://vocab.deri.ie/void/guide.

Volz, J. et al. “Discovering and Maintaining Links on the Web of Data.”

Web Conference. 2009. 650

WeGov. 2014. http://www.wegov

Weimer, D., and A. Vining.

Pearson Prentice Hall, 2005.

Wheeler, B. Why not let social media run the country?

politics-19555756.

Wills, G. J. “Visualizing hierarchical data.”

2009. 3425–3432.

Wilson, T. et al. “Recognizing contextual polarity in phrase

Proceedings of the conference on Human Language Technology and Empirical

Methods in Natural Language Processing.

WordNet. 2014. http://wordnet.princeton.edu/.

YahooAPI. 2014. http://developer.yahoo.com/search/content/V1/termExtraction.html.

Yardi, S., Boyd, D. “Dynamic debates: an analysis of group polarization over time on Twitter.”

Bulletin of Science, Technology & Society


33

van Ossenbruggen, J. et al. “Interactive Vocabulary Alignment.” International Conference on

Theory and Practice of Digital Libraries. 2011. 296-307.

ocab.deri.ie/void/guide.

Volz, J. et al. “Discovering and Maintaining Links on the Web of Data.” International Semantic

2009. 650-665.

WeGov. 2014. http://www.wegov-project.eu/.

Weimer, D., and A. Vining. Policy Analysis: Concepts and Practice . Upper Saddle River, NJ:

Pearson Prentice Hall, 2005.

Why not let social media run the country? 2012. http://www.bbc.co.uk/news/uk

Wills, G. J. “Visualizing hierarchical data.” Encyclopedia of Database Systems.

Wilson, T. et al. “Recognizing contextual polarity in phrase-level sentiment analysis.”


Methods in Natural Language Processing. Vancouver, 2005.

4. http://wordnet.princeton.edu/.

2014. http://developer.yahoo.com/search/content/V1/termExtraction.html.


Bulletin of Science, Technology & Society 30, no. 5 (2010): 316-327.

International Conference on

International Semantic

Upper Saddle River, NJ:

2012. http://www.bbc.co.uk/news/uk-

Encyclopedia of Database Systems. Springer US,

level sentiment analysis.”


2014. http://developer.yahoo.com/search/content/V1/termExtraction.html.


d2.2 tracking technologies and social attitudes€¦ · of the deliverable, an update is due in...

Documents