Improving the performance of semantic graph-based keyword extraction and text summarization using fuzzy relations in Hindi Wordnet

Abstract

In this study, a Fuzzy Semantic Graph-based approach is proposed to extract keywords and generate extractive text summaries from Hindi text documents. Hindi Wordnet is used as a knowledge source to construct the semantic graph. As the semantic relations defined in Hindi Wordnet are crisp, they do not capture the semantic relationship as a matter of degree. Due to that, many terms are represented as not being related, while these can share some meaningful relationship as per real-life scenarios. To overcome this curb of Hindi Wordnet, the paper presents several fuzzy semantic associations between such terms by assigning a value ranging from 0 to 1 to such relations. While constructing the semantic graph to represent documents using Hindi Wordnet semantic relations, the terms sharing fuzzy semantic relations are also added to enhance the quality of the graph. The experiments are done to extract potential keywords and to generate a good content summary. It is observed that such semantics generate a more accurate summary and produce prospective keywords for the document. The performance of the proposed approach fuzzy-based semantic graph is compared to semantic graph-based approach for keyword extraction and text summarization. The keywords extracted and the summary generated by the proposed approach is match up to human extracted keywords and human-generated text summary. The proposed approach results are evaluated using precision, recall, and f-measure. Different outcomes of generated text summaries are evaluated using the ROUGE matrix. The results of the proposed approach are pretty encouraging.