ontology maturing for searching, managing, and retrieving resources
DESCRIPTION
presentation of the paper "Using the Ontology Maturing Proces Model for Searching, Managing, and Retrieving Resources with Semantic Technologies" at the ODBASE 2008 conference, Monterrey, Mexico, Nov 13 2008TRANSCRIPT
Using the Ontology Maturing Process Model for Searching, Managing, and RetrievingResources with Semantic Technologies
FZI Research Center for Information TechnologiesKarlsruhe, GERMANY
{braun|aschmidt|awalter|zach}@fzi.dehttp://www.fzi.de/ipe
Simone Braun, Andreas Schmidt,Andreas Walter, Valentin Zacharias
2
Problem & Research Question
How to improve searching in, managing of, and retrieving ofresources through the use of
(semantic) annotations
3© FZI Research Center for Information Technologies Karlsruhe, Germany | Information Process Engineering | www.fzi.de/ipe
Motivation and Current Approaches
Motivation for our approach comes fromdeficiencies in current systems:
• Tagging, its advantages and its problems
• Semantic annotation, its advantages and itsproblems
4
Tagging
The use of arbitrary keywords for managing, searching, and finding resources
Advantages:• Lightweight, easy, adaptable,
no setup, proven - used by millions
Disadvantages:• Lack of precision due to problems like homonyms,
synonyms, multilinguality, typos, different waysto write words, tags at different levels
noodle (pasta) vs noodle (swear word)spaghettoni vs vermicellini
noodle vs Nudelspagetti vs spaghetti
SpaghettiCarbonara vs Spaghetti_Carbonarapasta vs spaghetti
5
Semantic Annotation
The use of (semantically) described entities for managing, searching, and finding resources
Advantages:• Through the use of concepts (instead of words) avoids tagging
problems such as homonyms, synonyms etc. • Potentially better management, searching, and browsing
Disadvantages:• Despite years of research so far not widely used• Needs ontology that is used for annotation
o this is often created by different users (KE experts) and updated only seldomly
o hence it’s often out-of date, incomplete, inaccurrate and incomprehensible
© FZI Research Center for Information Technologies Karlsruhe, Germany | Information Process Engineering | www.fzi.de/ipe
6
Hypotheses
Tagging and semantic annotationapproaches can be combined in a waythat avoids their respective drawbacks
while retaining the advantages
The core concept is the lightweightand simple collaborative evolution of
the ontology used for annotation
More on Motivation: Simone Braun, Valentin ZachariasSocial Semantic Bookmarking, PAKM 2008
7
Structure of Work & Presentation
ProcessModel
Implement. Evaluation
Iterative Co-DependentDevelopment
8
Structure of Work & Presentation
ProcessModel
Implement. Evaluation
Iterative Co-DependentDevelopment
Implementations:• Image annotation
with ImageNotion• Web resource
annotation withSOBOLEO
Process Model:• Ontology Maturing
model to explain collaborative ontologydevelopment processes and guide tool development
Evaluation:• Multiple
evaluations tovalidate processmodel & toolsand to guide tooldevelopment
9
Structure of Work & Presentation
ProcessModel
Implement. Evaluation
Iterative Co-DependentDevelopment
10
Quality of a Collaboratively Created Ontology
Good ontologies for semantic applications are a balance ofAppropriateness• Representation of the domain• wrt. the purpose of the ontology for the semantic application• Tight coupling between usage and updating of ontology elements
Social Agreement• Ontology represents a shared understanding of the community
elaborated in social & collaborative processes• Learning process of the users
o deepen their understanding of the real worldo the vocabulary (ontology elements) to describe the world
Formality• Ontology development is a process of continuous evolution• Different levels of formality might coexist
11
Process of Ontology Maturing
Based on the assumption that ontologies cannot be formalized in a single activityRather the result of continuous negotiation & collaborative learning processes taking place when applying the ontologies
12
Process of Ontology Maturing
Users annotate resources with arbitrary tagsNew concept ideas emergee.g. recent/specific tags like ‘whole grain spaghetti’
13
Process of Ontology Maturing
A common terminology evolves through the collaborative (re-)usage of the tags
Tags are defined and refined, useless or incorrect ones are rejectede.g. adding German ‘Vollkornspaghetti’ and a description
14
Process of Ontology Maturing
Community members begin to organize the concepts with hierarchical & ad hoc relations
resulting in a lightweight ontologye.g. ‘spaghetti’ <is broader> ‘whole grain spaghetti’
15
Process of Ontology Maturing
Adding axioms allows for exploiting relationships forreasoning
Users add more precise relations between entites; such aspartonomic relations, disjunction etc. e.g. ‘water‘,‘semolina‘ <is part of> ‘spaghetti‘
16
The Artifact, Knowledge & Social Dimensions
Concentrating only on the development of the ontology is not sufficient to create & analyze community-driven semantic applications
Need to consider that users have different levels of understanding of parts of the domain and that this understanding also evolves within usage processes
Viewing ontology development as collaborative learning processes requires to consider interaction, communication, and coordination processes
The development of social processes & competencies© FZI Research Center for Information Technologies Karlsruhe, Germany | Information Process Engineering | www.fzi.de/ipe
17© FZI Research Center for Information Technologies Karlsruhe, Germany | Information Process Engineering | www.fzi.de/ipe
18
The Artifact Dimension
Artifacts – a product of human conceptionThat mature from simple tags to formalized or evenaxiomatized ontology elements
The artifact dimension identifies available ontologyelements and their relations
© FZI Research Center for Information Technologies Karlsruhe, Germany | Information Process Engineering | www.fzi.de/ipe
19
Knowledge Dimension
The knowledge dimension is concerned with theknowledge of the users that ultimatively determineswhat they can model
On the individual level: • Alignment processes bringing forth a sufficient level of shared
understanding of the domain• Learning processes on artifacts creation methods
On the collective level:• Development of an understanding as such
© FZI Research Center for Information Technologies Karlsruhe, Germany | Information Process Engineering | www.fzi.de/ipe
20
Social Dimension
The social dimension is concerned with socialstructures & processesUsers need to learn to collaborateOn the individual level: • General willingness & competencies to interact, communicate,
negotiate, compromise, and accept rules
On the collective level:• The development of rules, best practices, identification of
leaders etc.
© FZI Research Center for Information Technologies Karlsruhe, Germany | Information Process Engineering | www.fzi.de/ipe
21
Structure of Work & Presentation
ProcessModel
Implement. Evaluation
Iterative Co-DependentDevelopment
22
ImageNotion: Semantic Image Annotation & Search
Semantic Image Annotation & Search
Benefits for image annotation• Semantics allow for improved navigation through image
archives (e.g. images with the same persons, events)• Multilinguality, reusability of ontology elements:
saves time for image annotation compared to textual annotation
Requirements for semantic image annotation• Work integrated, collaborative creation process of ontologies• Easy understandability of ontology elements• Usability and simplicity: tools and work steps must be
informal, lightweight, easy-to-use and easy to understand
© FZI Research Center for Information Technologies Karlsruhe, Germany | Information Process Engineering | www.fzi.de/ipe
23
ImageNotion: Semantic Image Annotation & Search
An imagenotion represents a semantic notion graphically through an imageGuides the process of visually creating an ontology that contains imagenotions and relationsAllows for collaborative creation and maturing of ontologies Allows for semantic annotation of images & maturing of their quality
© FZI Research Center for Information Technologies Karlsruhe, Germany | Information Process Engineering | www.fzi.de/ipe
Usage of imagenotions for semantic image annotation
• Imagenotion
Imagenotion
2. Consolidation in communities
• Descriptive• Textual
- Label text• - Synonyms• Date information• Links
VisualAssociate an
image
3. Formalization: Rules and relations
1. Createimagenotions
Emergence of newideas
24
ImageNotion: Semantic Image Annotation & Search
© FZI Research Center for Information Technologies Karlsruhe, Germany | Information Process Engineering | www.fzi.de/ipe
25
SOBOLEO: Social Semantic Bookmarking of Webpages
Use CaseSupporting knowledge workers working together in one domain in developing a shared ontology and a shared index of relevant web resources organized with this ontology
Course• Users encounter a web page• Annotating with concepts from the ontology or arbitrary tags• Gathering arbitrary tags as “prototypical concepts” for later
consolidation and placement• Or immediate switch to the ontology editor
o e.g. for adding synonyms or structuring with broader/narrower/ related relations
© FZI Research Center for Information Technologies Karlsruhe, Germany | Information Process Engineering | www.fzi.de/ipe
© FZI Research Center for Information Technologies Karlsruhe, Germany | Information Process Engineering | www.fzi.de/ipe
27
SOBOLEO: Social Semantic Bookmarking of Webpages
Use CaseSupporting knowledge workers working together in one domain in developing a shared ontology and a shared index of relevant web resources organized with this ontology
Course• Users encounter a web page• Annotating with concepts from the ontology or arbitrary tags• Gathering arbitrary tags as “prototypical concepts” for later
consolidation and placement• Or immediate switch to the ontology editor
o e.g. for adding synonyms or structuring with broader/narrower/ related relations
© FZI Research Center for Information Technologies Karlsruhe, Germany | Information Process Engineering | www.fzi.de/ipe
© FZI Research Center for Information Technologies Karlsruhe, Germany | Information Process Engineering | www.fzi.de/ipe
29
Structure of Work & Presentation
ProcessModel
Implement. Evaluation
Iterative Co-DependentDevelopment
30
Evaluations - Overview
5 Evaluations:(S1) CKC Workshop @ WWW 2007• 33 participants• 202 concepts, 393 relations, 155 resources, ∅ 3 concepts per
resource
(S2) Workshop of the IM WISSENSNETZ project• 4 participants with no modeling background• Guided user tests with observation, thinking aloud, interviews,
questionnaires & screenrecording
(S3) Workshop @ EATEL SummerSchool 2008• 24 participants with mixed background (CS, pedagogy etc.)• 182 concepts, 323 relations, 76 resources, ∅ 3 concepts per
resource© FZI Research Center for Information Technologies Karlsruhe, Germany | Information Process Engineering | www.fzi.de/ipe
31
Evaluations - Overview
(I1) Online survey• 137 participants• Task: create imagenotion „Manuel Barroso“
(I2) Workshop of the IMAGINATION project• 3x6 participants (Wikipedia users, French image agency
employees, Italian history students)• Task: annotate historical images according to the users area of
expertise
© FZI Research Center for Information Technologies Karlsruhe, Germany | Information Process Engineering | www.fzi.de/ipe
32
User Acceptance & Usefulness
Majority of users appreciated both tools and we could show that people from a variety of backgrounds are able to understand & interact with semantic annotation
Users liked in particular• the ease of use of ontology editing• the simple way for annotating with concepts & tags• possibility to integrate not yet well defined concepts• having ”starter concepts” & “to get the ontology building
almost for free“
© FZI Research Center for Information Technologies Karlsruhe, Germany | Information Process Engineering | www.fzi.de/ipe
33
Evaluation Results II
Evaluations also uncovered interesting effects showing importance of social and knowledge dimension, e.g.:
Mutual Support• Specialists for tool use or domain areas quickly emerged and
were asked by others for helpExtend tools to support users in identification and contacting these specialists
Interest in Background Knowledge• Users showed great interest in learning more about the subject
matter of the current resources they were annotating (e.g. by looking things up in Wikipedia)Encourage and extend tools to support this, e.g. by automatically adding texts from wikipedia as tag descriptions
34
Conclusions
Sustainable community-driven semantic applications need thefacilities such that (almost) all parts of the semantic model can beevolved by the community
The Ontology Maturing process model describes the maturingprocess of the semantic model at the artifact, social, andknowledge dimension
SOBOLEO and ImageNotion implement these vision of community-driven semantic applications and have been favourably received byusers in multiple evaluations
For the future we plan more long-term evaluations to furthervalidate the model and improve the tools
© FZI Research Center for Information Technologies Karlsruhe, Germany | Information Process Engineering | www.fzi.de/ipe
35
Contact
© FZI Research Center for Information Technologies Karlsruhe, Germany | Information Process Engineering | www.fzi.de/ipe
Simone Braun
FZI Research Center for Information Technologies
Karlsruhe, GERMANY
http://fzi.de/ipe
http://mature-ip.euhttp://imagination-project.org
http://www.imagenotion.com
http://tool.soboleo.com