Abstract

This paper explores the idea of keyword extraction from conversations, the goal of using these keywords to retrieve, for each short conversation text file, a small number of possibly relevant documents, which can be recommended to the participants. However, even a short conversation contains different types of words, which are absolutely related to several topics; Therefore, it is difficult to infer precisely the information needs of the conversation participants. The existing system proposed a diverse keyword extraction technique which extracts the keyword from the meeting conversation transcripts and recommends the document to the participants. So, in this paper we first propose an algorithm to extract keywords from the output of preprocessing process where string is processed to its basic meaning by following the basic four activities. Then, we propose a feature extraction method to extract multiple topically differentiated queries from this keyword set, in order to maximize the chances of making at least one relevant recommendation to participants. The proposed methods are evaluated in terms of relevance with respect to conversation fragments from the conversation text file. The results shows that our system improves over previous methods that consider only word frequency or topic similarity, and represents a promising solution for a document recommender system to be used in conversations.