Improving the retrieval performance by using distance-based bigram

In this paper, we discussed a new scheme of forming and weighing a term called a distance-based bigram. In this scheme, the distance between two words is considered for a new term and a new weighting. This new scheme is applied to the vector formation in the process of the vector space model with other standard term forming schemes: unigram and bigram. The tested domains are English and Thai medical corpora. The results show that our proposed method performs well for the Thai corpus under the condition that only a few returned documents are needed. Within the first ten percent of recall, our method improves the precision over the standard unigram by nearly 30%.