Two improvements to detect duplicates in Stack Overflow

Abstract

Stack Overflow is one of the most popular question-and-answer sites for programmers. However, there are a great number of duplicate questions that are expected to be detected automatically in a short time. In this paper, we introduce two approaches to improve the detection accuracy: splitting body into different types of data and using word-embedding to treat word ambiguities that are not contained in the general corpuses. The evaluation shows that these approaches improve the accuracy compared with the traditional method.