The Treatment of Word Sense Inventories in the ‘LACELL WSD Project’

Authors

  • Moisés Almela
Keywords: WSD, lexicology, lexicography, corpus linguistics, computational linguistics

Abstract

The WSD community has long debated whether the criteria for representing polysemy in general purpose dictionaries meet the specific demands of sense disambiguation tasks. Concern is growing that pre-defined sense inventories might not adjust well to the needs of WSD, because word occurrences can rarely be paired with rigid sense classes in a one-toone fashion. A second cause for concern is the level of sense granularity adopted in conventional dictionary entries.Fine-grained distinctions can be useful for a dictionary user but complicate the design and evaluation of WSD systems in a way that is often unnecessary. As a result of these objections, many experts have voiced the opinion that dictionaries are not adequate sources of sense inventories for WSD. However, the problem of word sense overlaps can also be resolved by modifying the way in which dictionary entries are processed by WSD programs. This is the solution applied in the LACELL WSD system. The algorithm selects simultaneously two or more dictionary senses if the context does not allow sufficient discrimination between/among them. This article explains the underpinnings of such proposal, as well as discussing some advantages and disadvantages.

Downloads

Download data is not yet available.
How to Cite
Almela, M. (2009). The Treatment of Word Sense Inventories in the ‘LACELL WSD Project’. International Journal of English Studies, 9(3), 21–38. Retrieved from https://revistas.um.es/ijes/article/view/99491