TOKYO -- Yahoo Japan has begun providing to clients, and even its rivals, a free service that uses artificial intelligence to determine if comments posted in Japanese on online forums are appropriate to its clients.
The service will then allow the client companies to delete the posts or warn writers if they are deemed to be defamatory. Three companies, including social media-based news website NewsPicks, are the first since Yahoo launched the service in September to adopt the cutting-edge technology that can make those analyses within a second.
In the competitive field of technology, it is rare for any company to open up its AI capabilities for free. More than 20 companies have consulted with Yahoo, which is owned by Z Holdings, over this technology, and it expects the number of clients to increase.
The AI analyzes words and phrases in Japanese comments to rate them on a scale of one to 10 on how "constructive" they are and their relevance to the subject or topic. Yahoo has been using this in-house deep-learning technology, which draws conclusions from vast amounts of data, on its own online news service since 2018.
The AI is also able to decide if a post is objective or biased, and rate those comments accordingly. About 60,000 comments on Yahoo News evaluated by 3,000 monitors were distilled and incorporated in the creation of this technology.
Based on the AI analyses, client companies can automatically highlight high-rated posts or remove low-rated content. For now, it will be used to "help detect posts that violate the terms of service," said a representative for NewsPicks.
Yahoo News provides around 7,000 stories per day that attract over 300,000 comments from users of its service. Yahoo changes the rankings of the comments based on their scores, and deletes about 20,000 posts a day that violate its terms of service. Experts decide how to deal with comments that are not obvious and require attention.
Currently, many companies employ staff to monitor posts. Some companies that want to introduce AI to do the job said they lacked the resources to do so properly.
Across the world, social networking websites have become home to slanderous, discriminatory comments and conspiracy theories that can incite hate. According to Japan's Ministry of Justice, based numbers from human rights organizations, there are now more than 1,500 cases of online abuse, fake news, and conspiracy theories, per year since 2015 that are in the process of resolution. The ministry partnered with Google this spring to crack down on such online behavior.
Z Holdings President Kentaro Kawabe has said that the company will create standards for the industry. Along with the free service, Yahoo will ask companies to share information of AI analyses, in the hope of accumulating more data that will help it to develop more advanced technologies for general use.
There are some worries that the technology can become a form of censorship and undermine freedom of expression but Keio University Professor Tatsuhiko Yamamoto said at a meeting of experts set up by Yahoo in 2020: "We will seek thorough transparency so that frank expressions do not disappear."
Yahoo said it will disclose AI algorithms and other information in a way that is easy for consumers to understand.