MIT researchers have created a system that detects factual inconsistencies in Wikipedia articles and updates them automatically. The goal is to reduce volunteer editors' time and effort in manual tasks. According to advanced, there are millions of articles that live from its constant editing and updating, and that can mean dates, names, numbers and locations that may be out of date.
The system created by the researchers detects and replaces outdated phrases, keeping the language similar to the way humans write and edit entries in the open encyclopedia. In practice, artificial intelligence acts as a kind of automatic corrector for more complex phrases. In other words, editors write in the Wikipedia interface unstructured sentences with updated information, without worrying about style or grammar. AI does a search on the network, finds the pages with the misaligned phrases and rewrites them with human rigor.
For the future, researchers say there is potential to build a fully automated system that identifies and uses the latest information from the internet to correct information from articles published on Wikipedia. The famous encyclopedia already uses automated systems in publishing, but in this case to mitigate acts of vandalism or to fill in pre-defined information in templates.
The researchers say the system can be used on other text-based systems to create automated routines. The study indicates that the system can be used to reinforce databases in the search for biased situations in the training of fake news tools.