Dec 23, 2021 (NHK) - Yahoo Japan has released the first report on its actions against guideline-violating posts, such as defamatory comments, in the comment section of its news site.
The firm uses artificial intelligence, in addition to human monitors, to automatically spot and delete comments that violate its rules.
The firm published the number of posts deleted and the reasons for removal for the first time in a bid to ensure transparency of its measures. This is in response to recommendations from experts, as concerns have been raised over excessive deletions.
The report says Yahoo Japan removed about 350,000 posts out of some 10.5 million in the month of March this year.
It also says that of the posts removed in the one-year period through March, 71 percent were automatically deleted by AI, while 22 percent were deleted by human patrols. Seven percent were removed by humans after being pointed out by outside parties.
Of the posts removed by humans, 65.4 percent were judged to be unpleasant, and 7.9 percent were thought to be unrelated to articles.
Of the posts deleted because they were judged to be unpleasant, 84.1 percent contained slander or excessive criticism, and 11.7 percent had ethnic slurs or hate speech.
The report says the automatic deletion by AI is based on scores given by a deep-learning model.
Professor Yamamoto Tatsuhiko of Keio University headed the expert panel that made the recommendations to Yahoo Japan.
Yamamoto said it is necessary to constantly review the post monitoring system, and to regularly study content and matters to be made transparent.
Source: ANNnewsCH
