Abstract: Today in the field of information filtering topic modeling is widely used. Topic modeling helps to generate models which can discover the hidden topics in a document collection and each of these topics are represented by word distribution. There are term-based and pattern-based approaches in information filtering. Patterns are more discriminative than single words. In many pattern-based methods the patterns in the documents are considered. But the pattern which occurs multiple times in a document which has to be filtered is also given equal importance. Another problem which the existing pattern-based methods face is that the semantics of a term in a pattern is not considered. Another limitation is that the distribution of a pattern in a document is not given any importance. To overcome the above mentioned problems a new ranking method which considers the number of frequency of the patterns, distribution of patterns and semantic based pattern representation to estimate the relevance of the documents based on the user information needs is introduced. This helps to remove the irrelevant documents effectively. TREC data collection Reuters Corpus Volume 1 is used to do the extensive experiments to evaluate the effectiveness of the proposed method. The result says that the proposed model works better than the existing pattern-based topic for document modeling in information filtering.
Keywords: Topic Model, Information Filtering, Pattern mining, relevance ranking, user interest model.
Title: SEMANTIC BASED PATTERN TOPIC MODEL
Author: Tincy Chinnu Varghese
International Journal of Computer Science and Information Technology Research
ISSN 2348-1196 (print), ISSN 2348-120X (online)
Research Publish Journals
