Context-sensitive synonymy identification for the study of language variation.

The choice of synonyms and translations to a word depends on the context in which that word is used. For example, Dutch 'bos' has different meanings in the sentences 'ik wandelde door het bos' and 'ik gaf haar een bos bloemen'. This influence of context is ignored by most corpus-based models of word meaning. I now aim to make these models context-sensitive and demonstrate their application in the study of language variation. The project will proceed in three stages. The first stage consists of the actual development of the context-sensitive models. We will explore what contextual factors influence the meaning of a word, and how we can use these factors to find context-sensitive lists of synonyms and translations automatically. In the second and third stage, we will apply the results of stage one to the study of language variation. We will first extend our findings to the automatic discovery of words with a different meaning in two language varieties, like Belgian Dutch and Netherlandic Dutch. These results will then support a large-scale investigation of the differences in word use between language varieties in several languages.

Keywords:Polysemy, Distributional semantics, Lexical semantics, Corpus linguistics, Computational linguistics, Language variation

Disciplines:Linguistics