Negations and document length in logical retrieval

Abstract. Terms which are not explicitly mentioned in the text of a document receive often a minor role in current retrieval systems. In this work we connect the management of such terms with the ability of the retrieval model to handle partial representations. A simple logical indexing process capable of expressing negated terms and omitting some other terms in the representation of a document was designed. Partial representations of documents can be built taking into account document length and global term distribution. A propositional model of information retrieval is used to exemplify the advantages from such expressive modeling. A number of experiments applying these partial representations are reported. The benefits of the expressive framework became apparent in the evaluation.

• Type: Journal
• Year: 2006
• Where: IS
• Authors: David E. Losada, Alvaro Barreiro
• Keywords: Artificial Intelligence, Current Retrieval Systems, Document, IS 2006, Partial Representations