Enhanced Neuro-Fuzzy Based Information Retrieval Technique | Original Article

ABSTRACT:

ABSTRACT

In the proposed Neuro-fuzzy based model, fuzzy system have been selected to enhance the capability of document retrieval process. To obtain this, fuzzy parameter or variables that can describe main features of the document are the term Frequency ratio (tfr), Document frequency ratio (dfr), Term frequency (tf), ratio of the number of search terms that occurs in one document to the length of the search term, Inverse document frequency (idf). In the proposed Neuro-fuzzy model for Information Retrieval, three fuzzy values have been used which are Low, Medium and High to represent fuzzy linguistic values. Various membership functions were tested for fuzzy linguistic value to identify the relevancy in document for query term. The query term is tested in three datasets and computed performance of methods using confusion matrix. The performance of the proposed model is compared with existing models, such as cosine similarity, L-shape, and S-shape membership functions. The comparison is based on precision (p), recall (r), and accuracy (a) parameters. Three standard text datasets, viz. Movie Review, Polarity and ACL IMDB (Large movie review) are used to evaluate the comparative models.