towards a distributional semantic web stack
DESCRIPTION
The ability of distributional semantic models (DSMs) to dis- cover similarities over large scale heterogeneous and poorly structured data brings them as a promising universal and low-effort framework to support semantic approximation and knowledge discovery. This position paper explores the role of distributional semantics in the Semantic Web vision, based on the state-of-the-art distributional-relational models, categorizing and generalizing existing approaches into a Distributional Semantic Web stack.TRANSCRIPT
Towards a Distributional Semantic Web Stack
André Freitas, Edward Curry, Siegfried HandschuhInsight Centre for Data Analytics
University of Passau
URSW 2014
Riva del Garda
Position paper
Model targeting semantic approximations
(from a praxis perspective)
Interested in collecting references / creating bridges with this community
2
Outline
Motivation
Distributional Semantic Models (DSMs)
Distributional-Relational Models (DRMs)
Applications
Take-away message
3
Motivation Semantic intelligent behaviour is highly dependent
on (commonsense, semantic) knowledge scale
Semantics =
Formal meaning representation model (lots of data)
+ inference model
4
Motivation Scalability problems
1st Hard problem: Acquisition
Semantics =
Formal meaning representation model (lots of data)
+ inference model
5
Motivation Scalability problems
2nd Hard problem: Consistency
Semantics =
Formal meaning representation model (lots of data)
+ inference model
6
“Most semantic models have dealt with particular types of constructions, and have been carried out under very simplifying assumptions, in true lab conditions.
If these idealizations are removed it is not clear at all that modern semantics can give a full account of all but the simplest models/statements.”
Baroni et al. 2013
Semantics for a Complex World
7
Distributional Semantic Models Semantic Model with low acquisition effort
(automatically built from text)
Simplification of the representation(vector-based)
Enables the construction of comprehensive commonsense/semantic KBs
Trades formal structure for volume of commonsense knowledge
What is the cost?Some level of noise
(semantic best-effort)
8
Distributional Hypothesis
“Words occurring in similar (linguistic) contexts tend to be semantically similar”
He filled the wampimuk with the substance, passed it around and we all drunk some
9
Distributional Semantic Models (DSMs)“The dog barked in the park. The owner of the dog put
him on theleash since he barked.” contexts = nouns and verbs in the
same sentence
10
Distributional Semantic Models (DSMs)“The dog barked in the park. The owner of the dog put
him on theleash since he barked.”
bark
dog
park
leash
contexts = nouns and verbs in the same sentence
bark : 2park : 1leash : 1owner : 1
11
Distributional Semantic Models (DSMs)
car
dog
bark
run
leash
12
Semantic Similarity & Relatedness
car
dog
bark
run
leash
13
Query: cat
Semantic Similarity & Relatedness
θ
car
dog
cat
bark
run
leash
14
Query: cat
Distributional Semantic Models (DSMs)
DSMs as Commonsense Reasoning
Commonsense data is here
θ
car
dog
cat
bark
run
leash
16
Semantic Approximation is here
Distributional-Relational Models (DRMs) Hybrid distributional + structured data
Semantic approximation as a first-class citizen
Structured data + user query provides a contextual support for the semantic approximation
17
Your Algorithm goes here
DRM
Text Collection
Structured Data
Distributional Semantic
Model
Distributional-Relational Models (DRMs)
Heuristics to minimize the
approximation errors
18
Your Algorithm goes here
DRM
Structured Data (the same or another one)
Structured Data
Distributional Semantic
Model
Distributional-Relational Models (DRMs)
Heuristics to minimize the
approximation errors
19
DRM
20
Application: Flexible Querying / Semantic Search
Freitas et al., ICSC 2011 Freitas & Curry, IUI 2014
Application: Selective Reasoning (1)
Speer et al. AAAI 2009 Freitas et al, NLDB 2014
22
Application: Distributional Semantics and Logic Programming
Pereira da Silva & Freitas, FOIKS 2014
23
Application: Knowledge Discovery
Entity similarity/Entity consolidation
Relationship discovery
Novacek et al. ISWC 2011 Cohen et al. T. AMIA Annu Symp 2009 Speer et al. AAAI 2009
24
Distributional Semantics / Semantic Web Stack?
25
26
Take-away message
Effective semantic approximation that works+
Automatic construction of comprehensive semantic models from unstructured data
+Simple to use
Powerful semantic pattern in practice.
27
Do-it-yourself
http://easy-esa.org
28