Analyzing semantic dynamics in a quest for truth discovery

Abstract: In an era of data abundance, the task of truth discovery (TD) becomes increasingly complex and crucial. The high availability of text data, particularly on social media platforms, necessitates the integration of TD algorithms with Natural Language Processing (NLP) techniques. This work aims to enhance recent works on TD over text by leveraging state-of-the-art NLP techniques. A key component of our approach is the application of the Anna Karenina principle, a method that suggests ”all good workers share common traits,” while ”every poor worker exhibits unique deficiencies,” making them dissimilar to other workers. We demonstrate that employing large language models (LLMs) to extract answers to specific, well-defined questions from informational text, combined with style transfer techniques to minimize stylistic discrepancies between text data, improves the accuracy of truth discovery.