A new approach for finding semantic similar scientific articles

Abstract

Calculating article similarities enables users to find similar articles and documents in a collection of articles. Two similar documents are extremely helpful for text applications such as document-to-document similarity search, plagiarism checker, text mining for repetition, and text filtering. This paper proposes a new method for calculating the semantic similarities of articles. WordNet is used to find word semantic associations. The proposed technique first compares the similarity of each part two by two. The final results are then calculated based on weighted mean from different parts. Results are compared with human scores to find how it is close to Pearson’s correlation coefficient. The correlation coefficient above 87 percent is the result of the proposed system. The system works precisely in identifying the similarities.