Paper Details

The system perform the extraction of keyword its address the problem for conversation for each shortconversion section. A less number of potentially significant documents with the goal of using the information retrievedwhich can be recommended to participant. Using automatic speech recongnization system introduce error among themwhich are potentially related to various topic, even short fragment contains a variety of word. Therefore, it iscomplicated to infer specifically the information needs the discussion of participants. The use of topic modelingtechniques and of a sub modular reward function which favors diversity in the keyword set, for making to match thepotential diversity of topic and reduce ASR noise. Then, paper propose a method to derive several topically dividedqueries from this keyword set, in order to make the most of the chances of building at least one significantrecommendation when using these queries to search over the English Wikipedia. The Fisher, AMI, and ELEAconversational corpora, rated by various human judges by using proposed methods are calculated in terms ofsignificance with respect to conversation fragments from. The scores show that our proposal improves over previousmethods that consider only word frequency or topic correspondence, and represents a promising solution for a documentrecommender system to be used in conversations.

Yogesh Daundkar, Shubham Potdukhe, Virendra Saini, Snigdha Swaroop, Prof. Pragati Kinage, “Document Recommendation for Conversation using Keyword Extraction with Clustering”, International Journal of Advance Engineering and Research Development (IJAERD), Vol. 3, Issue 3, pp. 622-625, March 2016.