What is beyond collocations?: insights from machine learning experiments
- Autores: Leo Wanner, Bernd Bohnet, Mark Giereth
- Localización: Atti del XII Congresso Internazionale di Lessicografia: Torino, 6-9 settembre 2006 / Elisa Corino (ed. lit.), Carla Marello (ed. lit.), Cristina Onesti (ed. lit.), Vol. 2, 2006, ISBN 88-7694-918-6, págs. 1071-1087
- Idioma: inglés
- Enlaces
- Dialnet Métricas: 3 Citas
-
Resumen
Traditionally, collocations are treated in lexicography as idiosyncratic word combinations that must be learnt by heart by second language learners and which must thus be listed explicitly in collocation dictionaries. However, the learners' capacity to understand and to produce collocations they have never heard before indicates that collocations are not as opaque as often assumed. In our work on the extraction of collocations from corpora and their classification with respect to a fine-grained semantically oriented typology, we experiment with several alternative machine learning techniques that exploit different characteristic features of collocations. These techniques can be viewed to model different strategies used by learners for the recognition of collocations. Their results can be thus expected to give us some evidence on how collocation dictionaries should be structured in order to provide best access to this important part of lexis.
