Syntactic Indexes for Text Retrieval

Abstract

In this paper, we present three techniques for incorporating syntactic metadata in a textual retrieval system. The first technique involves just a syntactic analysis of the query and it generates a different weight for each term of the query, depending on its grammar category in the query phrase. These weights will be used for each term in the retrieval process. The second technique involves a storage optimization of the system's inverted index that is the inverse index will store only terms that are subjects or predicates in the document they appear in. Finally, the third technique builds a full syntactic index, meaning that for each term in the term collection, the inverse index stores besides the term-frequency and the inverse-document-frequency, also the grammar category of the term for each of its occurrences in a document.