Advertisement

Google makes its most powerful language parser open source

Now anyone can add voice recognition to their app.

The days of us communicating with our computers by using our fingers are nearing an end. Google announced on Thursday that it is releasing its language parsing neural network framework, SyntaxNet, as an open source system. The released code includes everything you need to train it using your own data set, though Google is also releasing a version already fluent in English: Parsey McParseface.

SyntaxNet Natural Language Parser

These programs, like other language parsers, form the basis of Natural Language Understanding (NLU) systems. Parsers are what allow applications like Google Now and Siri to understand the words you are either speaking or typing, label each word by its syntax and then discern your intent from what you've said/typed. The problem is, sentences become more complex and harder to decipher the longer they are. A sentence comprised of just 20 words may have hundreds of different syntax interpretations. Luckily, Parsey McParseyface is ranked as the most accurate parsing model currently available with a comprehension accuracy of about 95 percent.