Term Weighting Using Term Dependence

Performance of an information retrieval system depends on its weighting scheme. Weighting of a term can be seen in two aspects, local and global. For each type of weighting scheme, generally, single terms are considered. Term dependency is quite natural in a document. Word pairs or phrases can better describe a document in place of single terms. In the present paper an attempt has been made to study and quantify the dependency of terms to each other. Term dependency has been utilized to define a local weighting scheme for word pairs. Utility of the proposed weighting scheme has been shown by arbitrarily choosing some documents and extracting relevant word pairs from the documents.