Canada Spends Logo

Explore our Budget 2026 analysis

Grant: Discovery Grants Program - Individual

University of Ottawa

$14,000.00

Making sense of semantic relations

FY 2012-13

Summary
Natural Language Processing appears to suffer from overspecialization. Highly focussed new research initiatives come with their own workshops. People work with more and more languages. Statistical methods, which dominate the field, often make success out of an increase of a few impenetrable scores. There ensues a troubling fragmentation: we can no longer evaluate progress of the whole discipline. To reclaim the big picture is unrealistic, but it may be feasible to reintegrate one central area: semantic analysis. Its mainstay is a system of semantic relations, but there is little agreement on what relations to look for. This project will seek a unified view of relations at sentence, phrase and word level, useful in a variety of applications. Semantic relation analysis relies for inspiration on the study of semantics in linguistics, and for practice on text collections annotated with instances of relations. WordNet is the best-known such resource for lexical semantics; several others cover specifically verbs or nouns. There exist mappings between pairs of those resources, but it should be desirable to align them more systematically, by proposing a ""master set"" of semantic relations. Concrete applications, for example semantic role labelling or recognition of relations between nominals, also rely on language resources annotated with semantic information. It should help if annotations were the same rather than merely compatible via complicated bridging. Semantic relations cannot be too few and too general, or too many and too specific to hinder recognition and classification tasks. A hierarchy of relations may be an answer; experiments with encoding knowledge based on various relation sets should help determine an adequate granularity. I will work, then, toward a hierarchically organized system of semantic relations at the word and sentence level, for wide use in language technology. There will be practical studies, the building of annotated data sets, proof-of-concept applications and software to share with the community, but at heart, this is a longer-term plan of work on some of the basic tenets of Natural Language Processing, seen both from the theoretical and the utilitarian perspective.
Program
Discovery Grants Program - Individual
Location
Ontario, Canada
Recipient Type
Academia
Type
Grant
Source
NSERC's Awards Data

As published

Application id
8193-2012
Name nom
Szpakowicz, Stan
Department département
Electrical Engineering and Computer Science , School of
Organization id
28
Institution établissement
University of Ottawa
Province en
Ontario
Province fr
Ontario
Country en
CANADA
Country fr
CANADA
Fiscal year exercice financier
2012
Competition year année de concours
2012
Award amount
14000
Program id
RGPIN
Program name en
Discovery Grants Program - Individual
Program name fr
Programme de subventions à la découverte - individuelles
Group en
Discovery grants programs
Group fr
Programmes de subventions à la découverte
Committee code
1507
Committee name en
Computer Science
Committee name fr
Sciences informatiques
Area of application code
1200
Area of application group en
Advancement of knowledge
Area of application group fr
Avancement des connaissances
Area of application en
Advancement of knowledge
Area of application fr
Avancement des connaissances
Research subject code
2801
Research subject group en
Artificial intelligence
Research subject group fr
Intelligence artificielle
Research subject en
Natural language and speech understanding
Research subject fr
Langage naturel et reconnaissance de la parole
Partie
1
Nb partie
5
Application title
Making sense of semantic relations
Keyword
NSERC
Application summary
Natural Language Processing appears to suffer from overspecialization. Highly focussed new research initiatives come with their own workshops. People work with more and more languages. Statistical methods, which dominate the field, often make success out of an increase of a few impenetrable scores. There ensues a troubling fragmentation: we can no longer evaluate progress of the whole discipline. To reclaim the big picture is unrealistic, but it may be feasible to reintegrate one central area: semantic analysis. Its mainstay is a system of semantic relations, but there is little agreement on what relations to look for. This project will seek a unified view of relations at sentence, phrase and word level, useful in a variety of applications. Semantic relation analysis relies for inspiration on the study of semantics in linguistics, and for practice on text collections annotated with instances of relations. WordNet is the best-known such resource for lexical semantics; several others cover specifically verbs or nouns. There exist mappings between pairs of those resources, but it should be desirable to align them more systematically, by proposing a ""master set"" of semantic relations. Concrete applications, for example semantic role labelling or recognition of relations between nominals, also rely on language resources annotated with semantic information. It should help if annotations were the same rather than merely compatible via complicated bridging. Semantic relations cannot be too few and too general, or too many and too specific to hinder recognition and classification tasks. A hierarchy of relations may be an answer; experiments with encoding knowledge based on various relation sets should help determine an adequate granularity. I will work, then, toward a hierarchically organized system of semantic relations at the word and sentence level, for wide use in language technology. There will be practical studies, the building of annotated data sets, proof-of-concept applications and software to share with the community, but at heart, this is a longer-term plan of work on some of the basic tenets of Natural Language Processing, seen both from the theoretical and the utilitarian perspective.