The Allen Institute for Artificial Intelligence in Seattle has unveiled a system that offers extreme compression of long computer science reports. The goal is to reduce the time required to study such literature.
The Semantic Scholar is an AI-powered tool used for scientific research. Thanks to the new summarization function, it looks through many research papers and reduces them to one sentence.
Now the Semantic Scholar database contains more than 10 million articles on computer science. Articles from other disciplines will be added gradually, according to Dan Weld, an information scientist at the University of Washington in Seattle who oversees the database.
The system helps researchers who have had to rely on scanning multiple headlines and often lengthy abstracts, which is especially challenging on mobile devices. After the first tests, the user reaction was positive.
Semantic Scholar has the highest compression ratio of any tool. Science articles average 5,000 words, but the AI tool shortens them to 21 words. The closest competitor, a Semantic Scholar, compresses documents 36 times, and the new system – 238 times.
The Allen Institute team makes its code available for free. They also created a demo site available to everyone.
So far, only works written in English are accepted. But the authors of the program hope to include documents in other languages eventually.