Abstract

Currently, the synonyms are a problem to retrieved answer from question answering systems. A fuzzy based similarity is one method that many researchers used to solve this problem. This paper applied the fuzzy method with TF-IDF weighting in considering the alphabet of words in order to analysis a similarity between words. Our corpus consists of five hundred answers collected from reliable medical resources. Several fuzzy conditions were investigated to find out the best condition for answering the question. To evaluate our proposed method, thirty frequently asked questions are tested and compared to experts answers. The results showed that the acceptable answers were discovered on 80% words similarity above (fuzzy degree is greater than 0.8) with 80.09% or more of precision.