Textual Similarity Detection from Sentence

Abstract

In computer science, textual similarity used for detecting the similarity between words, terms, sentences, paragraph, and document. In natural language processing, sentence similarity performs the tasks such as document summarization, word sense disambiguation, short answer grading , and information retrieval. The lexical overlapping approach evaluates the similarity between the sentence and finds whether a sentence pair is semantically equivalent or not. Existing methods are used for checking the similarity of long text documents. These methods process sentences in high-dimensional space and are not much efficient, requires human input and also not adaptable to some application domains. Semantic textual similarity methods improved in two areas -(a) in the semantic relation between the words and (b) in semantic resources to reduce the dimension. The proposed architecture uses the two methods for directly computing the similarity between very short texts of the sentence and long text sentences. The Weighted Overlap Approach based proposed method provides a nonparametric similarity by comparing the similarity of the rankings for an intersection of the senses in both the sentences. The Cosine similarity based proposed method identify all distinct words from the sentences. In the proposed work the similarity detection methods are focused to check the synonyms similarity between the sentences.