Unsupervised Hidden Topic Framework for Extracting Keywords (Synonym, Homonym, Hyponymy and Polysemy) and Topics in Meeting Transcripts

Keyword is the important item in the document that provides efficient access to the content of a document. It can be used to search for information or to decide whether to read a document. This paper mainly focuses on extracting hidden topics from meeting transcripts. Existing system is handled with web documents, but this proposed framework focuses on solving Synonym, Homonym, Hyponymy and Polysemy problems in meeting transcripts. Synonym problem means different words having similar meaning are grouped and single keyword is extracted. Hyponymy problem means one word denoting subclass is considered and super class keyword is extracted. Homonym means a word can have two or more different meanings. For example, Left might appear in two different contexts: Car left (past tense of leave) and Left side (Opposite of right). A polysemy means word with different, but related senses. For example, count has different related meanings: to say number in right order, to calculate. Hidden topics from meeting transcripts can be found using LDA model. Finally MaxEnt classifier is used for extracting keywords and topics which will be used for information retrieval.