Matching index expressions for information retrieval
Bernd Wondergem, P. van Bommel, Th.P. van der Weide · 1998
Index expressions are expressive descriptors that capture some of the linguistic structure of natural language. Index expressions have proven of value for Information Retrieval (IR) in many applications. However, quantitative matching schemes, or similarity measures, have not been provided for index expressions yet. In this article, several quantitative matching strategies for index expressions are devised. This vastly increases the potential of applying index expressions for IR. Both content and structure play a role in these matching strategies. Dierent views on the semantics of index expressions are analysed: order of subexpressions, embedding, and headedness. These views are formalised in criteria for which corresponding similarity measures are devised. These measures are proven optimal for the corresponding criteria. In designing the similarity measures, both the inductive and the structural representations of index expressions are exploited. 1 Introduction Index expressions (...