Google just announced that they’re open-sourcing their new language parsing model.
“Language understanding is the next big uncracked nut in AI,” Jason Freidenfelds, a global communications representative for Google, told the Observer in an email. “If we can really solve natural language, it’ll improve everything from the Google app (understanding your questions as well as the meaning of all the text in the world), to Inbox (suggesting smart replies), to products yet to be invented.”
This is big news that will benefit computer scientists around the world, and therefore, the field of artificial intelligence as well. But while the model—which can understand the functional role of each word in a given sentence and diagram this out automatically—is serious technical business, the name Google chose for it is anything but. The new language parsing model has been named Parsey McParseface.
“We were having trouble thinking of a good name, and then someone said, ‘We could just call it Parsey McParseface!’ So… yup,” Mr. Freidenfelds said.
Although it sports a silly name, Parsey McParseface is the most sophisticated model out there. In terms of accuracy, it closes in on 96 to 97 percent accurate, which beats the previous record of 94 percent (also held by Google).
“This suggests that we are approaching human performance—but only on well-formed text. Sentences drawn from the web are a lot harder to analyze,” a post on Google’s research announcing the release reads.
The model is part of the overall framework SyntaxNet and will be released on TensorFlow. Just last month, we reported that Google released the feature that TensorFlow users had been begging for since the company made it open-source in November. Now, developers and researchers are able to run machine learning on more than one machine simultaneously, shortening the training process for some models from weeks to hours.