Enhancing Electronic Dictionaries with an Index Based on Associations

Proceedings of the 21st International Conference on Computational Linguistics and 44th Annual Meeting of the ACL, pages 281–288, Sydney, July 2006. c⃝2006 Association for Computational Linguistics Enhancing electronic dictionaries with an index based on associations Olivier Ferret CEA –LIST/LIC2M 18 Route du Panorama F-92265 Fontenay-aux-Roses ferreto@zoe.cea.fr Michael Zock1 LIF-CNRS 163 Avenue de Luminy F-13288 Marseille Cedex 9 michael.zock@lif.univ-mrs.fr Abstract A good dictionary contains not only many entries and a lot of information concerning each one of them, but also adequate means to reveal the stored in- formation. Information access depends crucially on the quality of the index. We will present here some ideas of how a dictionary could be enhanced to support a speaker/writer to find the word s/he is looking for. To this end we suggest to add to an existing electronic resource an index based on the notion of association. We will also present preliminary work of how a subset of such associations, for ex- ample, topical associations, can be ac- quired by filtering a network of lexical co-occurrences extracted from a corpus. 1 Introduction A dictionary user typically pursues one of two goals (Humble, 2001): as a decoder (reading, listening), he may look for the definition or the translation of a specific target word, while as an encoder (speaker, writer) he may want to find a word that expresses well not only a given con- cept, but is also appropriate in a given context. Obviously, readers and writers come to the dictionary with different mindsets, information and expectations concerning input and output. While the decoder can provide the word he wants additional information for, the encoder (language producer) provides the meaning of a word for which he lacks the corresponding form. In sum, users with different goals need access to different indexes, one that is based on form (decoding), 1 In alphabetical order the other being based on meaning or meaning relations (encoding). Our concern here is more with the encoder, i.e. lexical access in language production, a feature largely neglected in lexicographical work. Yet, a good dictionary contains not only many entries and a lot of information concerning each one of them, but also efficient means to reveal the stored information. Because, what is a huge dic- tionary good for, if one cannot access the infor- mation it contains? 2 Lexical access on the basis of what: concepts (i.e. meanings) or words? Broadly speaking, there are two views concern- ing lexicalization: the process is conceptually- driven (meaning, or parts of it are the starting point) or lexically-driven2: the target word is accessed via a source word. This is typically the case when we are looking for a synonym, anto- nym, hypernym (paradigmatic associations), or any of its syntagmatic associates (red-rose, cof- fee-black), the kind of association we will be concerned with here. Yet, besides conceptual knowledge, people seem also to know a lot of things concerning the lexical form (Brown and Mc Neill, 1966): num- ber of syllables, beginning/ending of the target word, part of speech …

Content truncated to 500 words