A Chinese Document Retrieval Method Considering Text Order Information

Abstract

This paper investigated the use and effect of term positions in text retrieval. The approach models the relevance between text strings by the similarity of text orders. Text similarity measures in our approach captured term ordering and proximity. The experiments showed that incorporating positional information can improve the effectiveness of retrieval results. The main cost of incorporating positional information into a text retrieval system is a larger index space overhead because of the lossless preservation of term occurrences. However, this cost could be compensated by the better retrieval results the approach provided.