A Smart Email Client Prototype for Effective Reuse of Past Replies

Abstract

Email communication is widely used in managing customer queries such as complaints and inquiries. With an increasing number of customer emails being received every day, better tools are needed to answer emails efficiently, reuse past efforts and prevent overwhelming and cluttered email backlog. Finding similar past cases is difficult since customer requests might use non-standard terminology. In this paper, we develop a smart email client prototype which helps in replying to an email by providing a list of replies gleaned from the past replied emails. The suggested replies are ranked according to the level of similarity. We have evaluated real-world email data in order to ascertain that reuse is possible, but requires careful retrieval mechanisms. We implement and evaluate a case-based reasoning approach as a methodology to solve the problem by reusing previously written solutions from the past replies stored in the case base. We build a retrieval algorithm that finds similar cases beyond the exact matching, by using text processing and semantics analysis techniques. To optimize retrieval, we apply and cross-evaluate several text analysis techniques such as lexical analysis and synonym expansion, and our evaluation shows that synonym expansion could improve the chance of retrieving a more relevant match even at lower ranks. We evaluate our prototype based on the quality of retrieval results, index size, and processing time elapsed.