The original view of the lexicon is that it was "a repository of exceptions". The modern view is that the lexicon has a rich structure. Word senses and Multiword Expressions (MWEs) are two areas at the cutting edge of work on this structure.
There are several things in common between Word senses and MWEs:
1. The target for learning is only partially understood.
2. Recognizing them helps to reduce ambiguity.
3. They are important for natural language processing that occurs further upstream.
4. Both areas became more active and a subject for evaluation about twenty years ago.
The Research Topic aims to make a greater connection between theory and empirical work in the areas of word senses and multiword expressions. The key is exploring and explicating the links between (i) theoretical work in these areas, (ii) empirical evaluation of that work, and (iii) the relationship to linguistic and computational practice. In particular, consider work on Pustejovsky’s Generative Lexicon, and Hanks’ work on norms and exploitations, as well as potential connections between these areas and work on neural networks.
The Article Collection will solicit papers on the following topics and questions:
- Using MWEs and word senses in NLP tasks (e.g., parsing, translation, question answering)
- Representing semantic properties of MWEs and word senses
- Methods for representing variation in form
- Combining information from different lexicons
- How often do we have a norm vs. exploitation for word senses and for MWEs?
- How much variance is there between different senses in the accuracy of disambiguation?
- How can we deal with machine learning for MWEs and word senses when the target for classification is only partially understood?
- What are some of the commonalities and some of the differences in research on these topics?
- How should we evaluate more complex approaches such as the Generative Lexicon?
The Research Topic will focus on word senses and MWEs, but we also welcome papers about the evaluation of other aspects of the lexicon, e.g., subcategorization, sentiment lexicons, and semantic role labeling. The expected audience would be people who contribute to Semeval, LREC, and the workshops on MWEs.
The original view of the lexicon is that it was "a repository of exceptions". The modern view is that the lexicon has a rich structure. Word senses and Multiword Expressions (MWEs) are two areas at the cutting edge of work on this structure.
There are several things in common between Word senses and MWEs:
1. The target for learning is only partially understood.
2. Recognizing them helps to reduce ambiguity.
3. They are important for natural language processing that occurs further upstream.
4. Both areas became more active and a subject for evaluation about twenty years ago.
The Research Topic aims to make a greater connection between theory and empirical work in the areas of word senses and multiword expressions. The key is exploring and explicating the links between (i) theoretical work in these areas, (ii) empirical evaluation of that work, and (iii) the relationship to linguistic and computational practice. In particular, consider work on Pustejovsky’s Generative Lexicon, and Hanks’ work on norms and exploitations, as well as potential connections between these areas and work on neural networks.
The Article Collection will solicit papers on the following topics and questions:
- Using MWEs and word senses in NLP tasks (e.g., parsing, translation, question answering)
- Representing semantic properties of MWEs and word senses
- Methods for representing variation in form
- Combining information from different lexicons
- How often do we have a norm vs. exploitation for word senses and for MWEs?
- How much variance is there between different senses in the accuracy of disambiguation?
- How can we deal with machine learning for MWEs and word senses when the target for classification is only partially understood?
- What are some of the commonalities and some of the differences in research on these topics?
- How should we evaluate more complex approaches such as the Generative Lexicon?
The Research Topic will focus on word senses and MWEs, but we also welcome papers about the evaluation of other aspects of the lexicon, e.g., subcategorization, sentiment lexicons, and semantic role labeling. The expected audience would be people who contribute to Semeval, LREC, and the workshops on MWEs.
