Fact Finder - a Fact Search Engine
Information Retrieval Project - A.Y. 2023/24
Team:
- Mattia Piazzalunga - 851931
- Nicolò Urbani - 856213
- FEVER - Thorne et al.
- PyTerrier - Latest Version
The identification of what is real or artificially created is undoubtedly a crucial issue in contemporary society. The advent of social media has accelerated the dissemination of information, but at the same time, it has also led to the spread of unverified facts, artificially created content, and misinformation that do not correspond to reality, thus introducing artificial biases into society. The task of fact-checking plays a fundamental role in ensuring the quality of information, which extends beyond the boundaries of social media. In a context where news can be distorted, manipulated, or even created ad hoc to influence public opinion, fact-checking emerges as a tool of paramount importance in keeping society informed and critical-thinking. Although fact-checking is a necessity, according to a December 2021 Eurostat survey [1], only 23% of individuals aged 16 to 74 had verified the accuracy of information or content in the three months prior to the survey, despite 47% of them encountering false or dubious information from news websites or social media. Providing accurate information to support or refute a fact, therefore, becomes a fundamental challenge that heavily involves research and can mitigate their impact on society. Can an effective retrieval be part of the solution? In this project, the focus is on retrieving verified facts based on a well-known benchmark dataset: FEVER (Fact Extraction and Verification), Thorne et al. 2018 [2]. It is based on 185,445 manually annotated claims, categorizing them into three categories: Supported, Refuted, or Not Enough Info. FEVER has also become a reference dataset for IR and supports our goal: building a fact retrieval system.