Facebook’s AI Research (FAIR) Lab is open-sourcing its AI bot-building research, currently available in the form of fastText library. Facebook has been pushing bots on Messenger, where businesses can interact directly with customers, and even send them advertisements and subscription messages.
Now by opening up its resource library for bots, Facebook wants to make sure these bots are smarter, sharper, and will hopefully spam users less.
FAIR Lab is going to open-source fastText, the library system design for text representation and classification. Facebook hopes to share its learning from the field of AI in order to help developers in the field.
Facebook’s blogpost says, “Automatic text processing forms a key part of the day-to-day interaction with your computer; it’s a critical component of everything from web search and content ranking to spam filtering, and when it works well, it’s completely invisible to you.” Facebook says that in their tests, “fastText is often on par with deep learning classifiers in terms of accuracy.”
Facebook claims the fastText library can cut language training from days to a matter of seconds. “Ultimately, we hope that fastText will help us all design better applications and further advance the research in language understanding,” FAIR said in a statement.
The library of code is now available on Github for public use. It will require a compiler with “good C++11 support.” According to Facebook, fastText is much quicker than other learning methods. It can train models “on more than one billion words in less than 10 minutes using a standard multi-core CPU.” The system can also “classify half-million sentences among more than 300,000 categories in less than five minutes,” claims Facebook.
Facebook is pitching bots as the future of Messenger, where businesses can reach out to customers directly. But training computers to understand all forms of human language is not as easy a task, and thus Facebook is hoping that by opening up its research, it will boost these bots and hopefully make them smarter.
With inputs from IANS