Google has just made its natural language processing tool open source, letting developers take advantage of it to create newer experiences for users. The company has a huge AI model named SyntaxNet, and the English component of it is what’s being made open source now.
It’s called Parsey McParseface as a homage to the recent incident in the UK which saw Internet users successfully voting to have a $288 million sea vessel named Boaty McBoatface. Google claims that McParseface is able to correctly identify subjects, objects, verbs and other grammatical components of sentences as expertly as trained human linguists. And it does so with 94 percent accuracy.
It actually relies on SyntaxNet’s neural-network framework in order to analyze the linguistic structure of a sentence or statement that you type or speak on Google search. It then parses the functional role of each word in a sentence to bring out the true meaning of what you want to say.
Also See: Google team asks for gender equality in emoji, proposes fresh changes
Google has given a pretty simple example to help us understand how Parsey McParseface works. The sentence is ‘Alice saw Bob.’ Now a normal person will easily recognize what this sentence says, but artificial intelligence has to parse it by deconstructing it, and this is where it can go wrong.
It could very well recognize the word ‘saw’ as a noun, thinking it to be the tool used for cutting things. It’s these kind of scenarios that Google’s Parsey’s excels in by recognizing correctly what the inputted natural language wants to convey.
By making this tool open source, Google has taken a first step towards better AI language parsing all across the software world.
