Workshop on Computational Approaches to Synonymy
at the Symposium on Re-Thinking Synonymy, Helsinki, 28-30 October 2010Many problems and applications in computational linguistics and natural language processing implicitly invoke, in various forms, the concept of synonymy or identity of meaning. In one way or another, they involve either determining identity (or non-identity) of meaning in different surface forms or creating different surface forms for a single meaning.
For example, paraphrase recognition is an important component of the more-general problem of recognizing textual entailment. Textual tailoring and personalization seeks to find the most effective linguistic realization of a message for a particular user; automatically simplifying texts and creating stylistic variations are special cases of this. Lexical choice in text generation tries to find the best word for a given meaning and to discriminate it from other words that are close in meaning but not synonymous in the context. Cross-lingual document retrieval and other cross-lingual applications such as, in particular, machine translation conflate the ideas of synonymy and translation equivalence.
But while there has been a large amount of research on computational methods for determining degree of similarity in lexical meaning and for recognizing paraphrase, little attention has been given to theoretical considerations of synonymy. Mostly, it is treated as a boolean property (two words are or aren't in the same synset; two sentences are or aren't mutual entailments) with little thought of any theoretical underpinning.
On the other hand, the real-world linguistic problems that natural language processing addresses provide useful test cases for linguistic theories of synonymy, and the computational methods developed are de facto theories of synonymy even if not intended as such.
This workshop will explore computational approaches to synonymy, with an emphasis on explicating their implicit theoretical notions and their implications for linguistic theory. Papers are solicited on the following topics:
- Computational theories of lexical, phrasal, and sentential synonymy.
- Principled methods of paraphrase recognition and generation.
- Principled methods of text tailoring and stylistic variation.
- Cross-lingual synonymy and machine translation.
- Beyond the synset: Principles of synonymy in computational lexical resources
- Yet another paraphrase-recognition or textual entailment system.
- Yet another lexical similarity metric.
- Yet another personalization or text-simplification system.
Submissions: Abstracts of up to 500 words should be submitted (as pdf attachments) to synonymy@cs.toronto.edu by 30 June 2010. Decisions on acceptance will be notified by 15 July 2010.
Organizers: Graeme Hirst (University of Toronto); Kentaro Inui (Tohoku University); Manfred Stede (University of Potsdam).
| Last modified: Monday, 12-Apr-2010 10:38:32 EEST |