A tool "ORES" that enables Wikipedia to judge whether the quality has not been lowered by editing / modifying articles using AI is open source released

The online encyclopedia "Wikipedia" is used by users all over the world thanks to the openness that anyone can edit and modify freely, but avoid the degradation of article quality due to malicious edits because of its openness I can not. For this reason, it is essential to check and modify by Wikipedian which checks the contents of modification and sometimes deletes and rewrites, but it is an open source program that utilizes AI "Objective Revision Evaluation Service (ORES)"The Wikimedia Foundation that runs Wikipedia reveals it on the official blog.

Artificial intelligence service gives Wikipedians 'X-ray specs' to see through bad edits «Wikimedia blog

Traditionally, the Wikimedia Foundation has developed and provided software that helps Wikipedian exercise editorial rights for the purpose of securing the quality of articles, such as whether there is no so-called "vandalism" acts in the article, It was. There is "AntiVandalBot" (AVB) as a representative thing "vandalism destroyer bot", and it is clear from the following article about the background of AVB development.

How has Wikipedia continued to eradicate vandalism such as alteration and seizure? - GIGAZINE

This time, the newly developed software "ORES" displays numerical values ​​as to whether the contents of edits are to lower the quality of articles from the viewpoint of grammar, context, etc. when articles are edited. It is important to note that it is not a judgment on the "truthfulness of contents" which is a matter of quality in grammar and context but difficult to check.



Then "false" is quantified to about 0.05, true is about 0.95. It seems there is no degradation in quality by this editing.

On the other hand, if the edit ID is "638307884" ...

"Fear of quality deterioration ant" because false is dominant. It is OK if Wikipedian checks for the first time here.

ORES is a program using existing published AI technology, and has open API as open source. There seems to be a lot of tools for editors that adopted AI technology so far, but based on reflection on the fact that many of the brand new Wikipedian, which were hard to be attached, were not used much, it is simple to diagnose by mere numerical value I heard he did. At the time of writing, ORES supports Wikipedia articles in 14 languages ​​including English, Spanish, German and French, but the Japanese version has not yet been released.

in Software, Posted by darkhorse_log