Author Description

Information Retrieval(IR) systems are gaining importance due to wide range of applications like recommender systems, search engines, etc., however, most of the IR systems use statistical methods built on top of bag-of-words approach for text retrieval. Graph-of-words approach is an alternative to bag-of-words approach that uses graph theoretic methods to rank keywords and related documents. We represent text documents as graphs whose vertices correspond to the unique terms belonging to the document. The edges represent co-occurrences between the terms. The underlying assumption is that the terms that co-occur have some sort of semantic relationship that can be harnessed for IR systems. The significant terms can be extracted using graph centrality measures. In this book, we have proposed a novel graph-of-words indexing technique using eigenvector scores that uses case separation for Gujarati language. We compared the performance of IR systems of our approach over the classical bag-of-words approach, mean average precision (MAP) values obtained in our experiments show that our approach has shown significant improvement over classical approaches.