Grant: Discovery Grants Program - Individual
University of Alberta
$42,000.00
Developing Theoretically-Grounded Explainable Approaches in Lexical Semantics
EF 2024-25
Résumé
We aim to develop a unified theory of synonymy and translational equivalence that would provide
theoreti- cal foundations for wordnets and multi-wordnets. A wordnet is a rich semantic and
lexicographic resource which groups words into synonym sets, or synsets, and which records relations
between synsets, such as hypernymy and hyponymy. Multi-wordnets are a multilingual generalization,
in which words from dif- ferent languages which represent the same lexicalized concept may share
mutli-lingual synsets. Wordnets and multi-wordnets have proven to be extremely useful resources,
especially in unsupervised settings, for a wide variety of important tasks in natural language
processing (NLP) and machine learning, including sentiment analysis, relation extraction, and word
sense disambiguation (WSD). However, wordnets are language-specific, precluding the application of
many state-of-the-art methods to languages with fewer available resources.
We anticipate that the aforementioned theoretical developments will lead, via testable hypotheses,
to innovative methods for advancing the state of the art in semantics. Such predictive theories are
vital to guiding scientific inquiry and interpreting experimental results. Recent research in
computational lexical semantics has focused almost exclusively on the engineering aspects of the
field, leaving important theo- retical issues unaddressed, and leading to a severe lack of
explainable, interpretable models for semantic tasks. Of particular interest to us is the
relationship between synonymy, sense distinctions, and cross- lingual lexicalization distinctions.
Various, and often conflicting, assumptions about these phenomena form the basis for much of the
relevant literature; however, these assumptions are rarely stated explicitly, and are even more
rarely tested. Publications pointing out this fundamental weakness in the foundations of the field
have largely been ignored by the broader community. We intend to address these deficien- cies, and
aim, in doing so, to not only make vital theoretical contributions, but also to develop novel,
theoretically-sound methods and resources for important semantic tasks.
In the long term, the problem we seek to address is the automatic construction of wordnets and
multi- wordnets from multilingual corpora and thesauri, which will allow us to both validate our
theory, and demonstrate its practical application. Construction of a wordnet typically involves a
large amount of manual work by language experts, which often results in limited coverage and
accuracy, especially for lower-resourced languages. Maintaining or expanding a wordnet requires a
further investment of time and resources. A language-independent method for wordnet construction
would therefore greatly increase the number of languages covered by modern NLP systems. Since the
method would be based on our own explicitly formulated theory, this method would be more
interpretable than previously proposed methods.
Programme
Discovery Grants Program - Individual
Bénéficiaire
Emplacement
Alberta, Canada
Type de bénéficiaire
Academia
Type
Grant
Source
NSERC's Awards Data
Tel que publié
Application id
RGPIN-2024-04959
Name nom
Kondrak, Grzegorz
Department département
Computing Science
Organization id
9
Institution établissement
University of Alberta
Province en
Alberta
Province fr
Alberta
Country en
CANADA
Country fr
CANADA
Fiscal year exercice financier
2024
Competition year année de concours
2024
Award amount
42000
Program id
RGPIN
Program name en
Discovery Grants Program - Individual
Program name fr
Programme de subventions à la découverte - individuelles
Group en
DISCOVERY RESEARCH
Group fr
RECHERCHE AXÉE SUR LA DÉCOUVERTE
Committee code
1507
Committee name en
Computer Science
Committee name fr
Sciences informatiques
Area of application code
1207
Area of application group en
Advancement of knowledge
Area of application group fr
Avancement des connaissances
Area of application en
Information, computer and communication technologies
Area of application fr
Technologie de l'information, des ordinateurs et des réseaux de communications
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
Field of research list names en
Not available
Field of research list names fr
Non disponible
Socioeconomic objective list names en
Not available
Socioeconomic objective list names fr
Non disponible
Application title
Developing Theoretically-Grounded Explainable Approaches in Lexical Semantics
Keywords
natural language processing; lexical semantics; word sense disambiguation; translation; synonymy; wordnets; semantic resources; less-studied languages; explainability; interpretability
Application summary
We aim to develop a unified theory of synonymy and translational equivalence that would provide
theoreti- cal foundations for wordnets and multi-wordnets. A wordnet is a rich semantic and
lexicographic resource which groups words into synonym sets, or synsets, and which records relations
between synsets, such as hypernymy and hyponymy. Multi-wordnets are a multilingual generalization,
in which words from dif- ferent languages which represent the same lexicalized concept may share
mutli-lingual synsets. Wordnets and multi-wordnets have proven to be extremely useful resources,
especially in unsupervised settings, for a wide variety of important tasks in natural language
processing (NLP) and machine learning, including sentiment analysis, relation extraction, and word
sense disambiguation (WSD). However, wordnets are language-specific, precluding the application of
many state-of-the-art methods to languages with fewer available resources.
We anticipate that the aforementioned theoretical developments will lead, via testable hypotheses,
to innovative methods for advancing the state of the art in semantics. Such predictive theories are
vital to guiding scientific inquiry and interpreting experimental results. Recent research in
computational lexical semantics has focused almost exclusively on the engineering aspects of the
field, leaving important theo- retical issues unaddressed, and leading to a severe lack of
explainable, interpretable models for semantic tasks. Of particular interest to us is the
relationship between synonymy, sense distinctions, and cross- lingual lexicalization distinctions.
Various, and often conflicting, assumptions about these phenomena form the basis for much of the
relevant literature; however, these assumptions are rarely stated explicitly, and are even more
rarely tested. Publications pointing out this fundamental weakness in the foundations of the field
have largely been ignored by the broader community. We intend to address these deficien- cies, and
aim, in doing so, to not only make vital theoretical contributions, but also to develop novel,
theoretically-sound methods and resources for important semantic tasks.
In the long term, the problem we seek to address is the automatic construction of wordnets and
multi- wordnets from multilingual corpora and thesauri, which will allow us to both validate our
theory, and demonstrate its practical application. Construction of a wordnet typically involves a
large amount of manual work by language experts, which often results in limited coverage and
accuracy, especially for lower-resourced languages. Maintaining or expanding a wordnet requires a
further investment of time and resources. A language-independent method for wordnet construction
would therefore greatly increase the number of languages covered by modern NLP systems. Since the
method would be based on our own explicitly formulated theory, this method would be more
interpretable than previously proposed methods.