Standard Statistical and Graph based Automatic Keyword Extraction

Abstract

Automatic extraction of terms from a document is essential in the current digital era to summarize the documents. For instance, instead of go through the full documents, some of the author's keywords partially explain the discussions of the documents. However, the author's keywords are not sufficient to identify the whole concept of the document. Hence the requirement of automatic term extraction methods is necessary. The major categories of automatic extraction approaches falls mainly on some techniques such as Natural Language Processing, Statistical approaches, Graph Based approaches, Natural Inspired algorithmic approaches, etc. Even though there are numerous approaches available the exact automatic keyword extraction is a major challenge in areas, that reveals around documents. In this paper, a comparative analysis of Keyword extraction between standard Statistical approaches and Graph based approaches has been conducted. In standard statistical approaches, the terms are extracted on the basis of physical counts and in the Graph based approach, the documents are automatically constructed as graphs by applying centrality measures during the keyword extraction process. The results of both approaches were compared and analyzed.