The MIT presented a study of a software that uses the AI to correct obsolete information within the pages. Still in implementation, the algorithm - which will be able to write "as a human" - could also be used to find false news.
An artificial intelligence will automatically correct the Wikipedia entries
The MIT presented a study of a software that uses the AI to correct obsolete information within the pages. Still in implementation, the algorithm - which will be able to write "as a human" - could also be used to find false news.
The study was presented by researchers in New York at the Aaai conference on artificial intelligence. Thanks to machine learning, the program compares the sentences of the Wikipedia pages in a reasoned way, that is, in order to trace a possible contradiction between two periods: the one already present on the online encyclopedia and the new one that contains the update. The latter must be typed in by the human operator in an unstructured way, that is, without paying particular attention to the style and grammar of the text. Only in this way does the system trace the old information, change it according to the new input, rewriting it as an editor.
If you consider the popularity and the enormous amount of information in the online encyclopedia, you understand how common it is to find some pages that are not updated. Not surprisingly, before this invention, other bots were developed that deal mainly with the maintenance of Wikipedia. In practice they are dedicated to repelling the attack by spammers and / or to restore and correct expired links. With the new system, researchers aim to make the whole process automatic: in the future there will be no need for the user to report the news. The bot will be able to trace them by looking directly among the latest news on the web, and then inserting them into the texts of Wikipedia.
OKAY, BUT WHERE ARE WE?
At the moment, however, the software is not completely ready for use. In particular, when assessing the accuracy, the model, specify by the MIT, gets 4 points out of 5 on the quality of updates and 3.85 on the accuracy of the grammar. Although not yet fully optimized, it is the one that has achieved the best results in a similar experiment ever conducted so far. According to the institute, an approach of this type must not however be limited to the online encyclopedia: it may in fact also be useful in the fight against false news. The algorithm can in fact be used to identify, in the multiplicity of sources that deal with the same news, that - or those - that constitute a potential anomaly, reporting and possibly unmasking it.
