Text Rank: A Novel Concept for Extraction Based Text Summarization

Indexing used in text summarization has been an active area of current researches. Text summarization plays a crucial role in information retrieval. Snippets generated by web search engines for each query result is an application of text summarization. Existing text summarization techniques shows that the indexing is done on the basis of the words in the document and consists of an array of the posting lists. Document features like term frequency, text length are used to assign indexing weight to words. Hence indexing weights of the document words are used to calculate the sentence similarity value between document words which remains independent on context. The word based index seems to be less efficient due to information retrieval problems like polysemy and Synonymy. Thus the significance of term for building the index is reduced and the emphasis is laid on the context of the document. This paper proposes an indexing structure in which index is built on the basis of context of the document rather than on the terms basis. While doing so we have also used novel concept of Lexical association (semantic association) between document words to calculate the similarity between sentences using computed indexing Weights. The proposed concept of sentence similarity measure has been used with the graph-based ranking method to create document graph and get summary of document.