SentenceRank — A graph based approach to summarize text

Abstract

We introduce a graph and an intersection based technique which uses statistical and semantic analysis for computing relative importance of textual units in large data sets in order to summarize text. Current implementations consider only the mathematical/statistical approach to summarize text. (like frequency, TFIDF, etc.) But there are many cases where two completely different textual units might be semantically related. We hope to overcome this problem by exploiting the resources of WordNet and by the use of semantic graphs which represents the semantic dissimilarity between any pair of sentences. Ranking is usually performed on statistical information. The algorithm constructs semantic graphs using implicit links which are based on the semantic relatedness between text nodes and consequently ranks nodes using a ranking algorithm.