A system created by MIT researchers, the Massachusetts Institute of Technology, could be used to automatically update factual inconsistencies in Wikipedia articles, reducing the time and effort employed by human editors who are currently performing this activity manually.

The millions of articles in Wikipedia, explain the researchers of MIT, need constant changes, to be updated with new information. Currently, people around the world work as volunteers to make these changes.

In a paper presented at the AAAI Conference on Artificial Intelligence, MIT researchers described a text generation system that identifies and replaces specific information in the relevant Wikipedia phrases, maintaining a style similar to what humans would use in writing

The basic idea of researchers is that humans type in an interface an unstructured phrase with up-to-date information, without having to worry about style or grammar. The system would then perform the search in Wikipedia, identify the appropriate page and the obsolete sentence and rewrite it with a human

In the future, researchers say, there is the possibility of creating a fully automated system that identifies and uses the latest information on the web to produce rewritten sentences in the corresponding Wikipedia articles, to reflect up-to-date information.

There are many other bots that make automatic changes to Wikipedia, the researchers point out. However, these bots usually work in mitigating vandalism or on some information that is strictly specified in predefined templates.

The MIT model, however, solves a more difficult problem of artificial intelligence: given a new piece of unstructured information, the model automatically changes the sentence in a human way. The other bots are basically based on rules, while the MIT system requires reasoning on contradictory parts in two sentences and generates a coherent text.

The system can also be used for other text generation applications. In their paper, researchers also used it to automatically synthesize sentences in a popular fact-cecking dataset that helped reduce bias without having to manually collect additional data.

Leave a Reply

Your email address will not be published.

You May Also Like