Identification and Treatment of Multiword Expressions Applied to Information Retrieval

Abstract

The extensive use of Multiword Expressions (MWE) in natural language texts prompts more detailed studies that aim for a more adequate treatment of these expressions. A MWE typically expresses concepts and ideas that usually cannot be expressed by a single word. Intuitively, with the appropriate treatment of MWEs, the results of an Information Retrieval (IR) system could be improved. The aim of this paper is to apply techniques for the automatic extraction of MWEs from corpora to index them as a single unit. Experimental results show improvements on the retrieval of relevant documents when identifying MWEs and treating them as a single indexing unit. 1