UNITED STATES (OBSERVATORY NEWS) — The quality of information on Wikipedia can suffer not only because of deliberate distortions or the poor work of editors, but also because it is outdated.
The Massachusetts Institute of Technology has developed a machine algorithm that can update articles in an online encyclopedia.
Artificial intelligence has been used in editing Wikipedia before, but its capabilities were limited to updating information in hard-coded templates or combating vandalism.
The new algorithm is capable of precisely rewriting phrases, updating information in them and at the same time not violating the style of the text, while maintaining a “human” tone. In addition, he is trained to complement the data sets used to identify fake news.
A machine learning-based system is trained to notice the differences between a phrase in a Wikipedia article and a statement that contradicts it with new facts.
Having discovered the contradictions, the algorithm selects in the source text the words that need to be deleted and those that need to be left. After that, the program determines, using simplified phrases in Wikipedia and in the source of new information, how to rewrite the article.
In its current form, development is not yet ready for widespread use – the average assessment of the quality of the text it gives people is 4 on a five-point scale for facts and 3.85 for grammar. This is better than previously created algorithms for writing text, but it still does not allow you to write phrases indistinguishable from human ones.
This article is written and prepared by our foreign editors writing for OBSERVATORY NEWS from different countries around the world – material edited and published by OBSERVATORY staff in our newsroom.
Contact us: [email protected]