A Semantic Similarity Measure Based News Posts Validation on Social Media

At present, Social media networks are widely used for information sharing among billions of people around the globe. However, the credibility of information shared through social media is questionable because the sharing mechanisms used are endless and the initiator of a news item is often unknown. This results in sharing of inaccurate information since the users of social media networks share news posts without varying the authenticity and the accuracy. In order to address this issue, a novel approach to calculate the accuracy level of news posts is proposed in the research paper. The aim of this research project is to provide an accuracy level for a social media news post that is posted as a status update by a user. The proposed system extracts the content of the news item, searches the Internet to find similar articles in reliable online news sources, matches the extracted content with the content of the news sites and generates an accuracy level. In developing the system, Natural Language Processing techniques such as web scraping techniques, web crawling techniques, URL ranking methodologies, automatic text summarization techniques and semantic analysis techniques such as Word2vec and cosine similarity are used. After implementing our system, we have gained a 70% of accuracy of relevancy of news posts on social media with compared to the reliable online news sources.