Meta AI builds tool to check Wikipedia sources using machine learning
Articles on the Internet encyclopedia 'Wikipedia' are required to cite what kind of references are used. However, due to the characteristics of Wikipedia that anyone can edit, the source of information may be completely unrelated to the main text, or the source may be deleted and the quality of the article may be degraded. As one of the ways to solve these problems, researchers at Meta have proposed using AI to check the integrity of information sources.
Improving Wikipedia Verifiability with AI | OpenReview
Meta wants to supercharge Wikipedia with an AI upgrade | Digital Trends
A system called ' Side ' devised by Mr. Fabio Petroni, who belongs to the AI team of Meta, analyzes the link of the citation source by neural network-based scanning, and is it consistent with the content written on Wikipedia? can be investigated effectively.
Side is learning using a dataset of 4 million citations on Wikipedia and can automatically scan hundreds of thousands of citations at once. In addition to examining the similarity of words between the article and the source, you can also examine the similarity in sentence units by dividing the content of the source into sentences.
In an experiment conducted by Mr. Petroni and others, it was found that humans prefer the source selected by Side with a probability of 70%. ``What we've made is a proof of concept, it's not really usable at the moment,'' Petroni said. There is,' he said.
``Side will help you find better references in addition to finding fraudulent citations,'' said technology media Digital Trends. He also said that, at least at this time, no attempt has been made to assess the quality of the sources cited, and that such tools are needed to increase the credibility of articles.
Related Posts:
in Web Service, Posted by log1p_kr