Towards Mitigating Misinformation: A Structured Dataset of Fact-Checked Claims from News Media

• Main Authors: Bhanushali, Oam, Mer, Aditya, Giridhar, Rishikesh, Somaiya, Bhavormi, Manasia, Arish, Singh, Shivam, Sharma, Neha, Panday, Mrityunjoy, Nemade, Milind, Shah, Sejal, Mani, G.S
• Format: Conference Proceeding
• Language: English
• Subjects: Accuracy; Analytical models; ColBert; Data Labelling; Fake news; Measurement; Misinformation; Reliability; Selenium; Support vector machines; Text Analysis; Trend Analysis; Web Scraping; Web sites
• ISSN: 2642-6102
• DOI: 10.1109/TENSYMP61132.2024.10752132

False information has become an unavoidable endeavor in the digital world, endangering public discourse and influencing people's decisions. Misinformation comes from a variety of sources and can travel quickly through online networks, drastically altering public perceptions, influencing political beliefs, and escalating audience prejudices. It is possible to come across several types of false information through text, picture, or video recordings. Misinformation has grown to be a significant problem in a world where memes are common, improved technologies are easily accessible, and mind-sharing is unrestrained. In order to address this widespread issue, researchers have gathered news information from many websites that provide truth-checking services, including India Today and ANI. Python libraries and modules, inclusive of BeautifulSoup and Selenium, were utilized to scrape those web sites and create the labeled dataset called "Factrix" indicating the veracity of the records (e.g., fake, true, half-true, mostly-false). Factrix dataset is intended to train models for the purpose of identifying false information through the use of textual records. These statistics can also be applied to pictorial analysis. Furthermore, these datasets enable trend analysis spanning from 2019 to 2024, providing insights into the evolution of incorrect information through the years.