A novel statistical and linguistic features based technique for keyword extraction

Abstract

WWW is a decentralized, distributed and heterogeneous information resource. With increased availability of information through WWW, it is very difficult to read all documents to retrieve the desired results; therefore there is a need of summarization methods which can help in providing contents of a given document in a precise manner. Keywords of a document may provide a compact representation of a document's content. As a result various algorithms and systems intended to carry out automatic keywords extraction have been proposed in the recent past. However, the existing solutions require either training models or domain specific information for automatic keyword extraction. To cater to these shortcomings an innovative hybrid approach for automatic keyword extraction using statistical and linguistic features of a document has been proposed. This statistical and linguistic technique based keyword extraction works on an individual document without any prior parameter change and takes full advantage of all the features of the document to extract the keywords. The extracted keywords can than assist in domain specific indexing. The performance of the proposed method as compared to existing Keyword Extraction tools such as Dream web design etc. in terms of Precision and Recall are also presented in this paper.