Google is stepping up its game of open-sourcing tools for improved language parsing in smart AI, and Parsey McParseface is the latest name on the list.
The near-unpronounceable software may sound laughable, but it is Google’s instrument for understanding English. It is plausible that when they named the module, Google’s engineer got inspired by Boaty McBoatface, the British research vessel.
By going open-source with its parsing, the company gives developers access to its SyntaxNet framework that makes sure sentences are parsed correctly. Parsey McParseface is just the module that connects to the framework and works for the English language.
Google touts that the framework is aptly identifying sentence pieces such as objects, subjects or verbs with an accuracy of 94 percent. Machine learning helps immensely with finding the proper sense of the sentences, and Google expects the 94 percent accuracy to increase after coders deploy it in their own software.
Parsing is an essential tool to ensure that computers understand verbal interaction for search features and more. Seeing how human language comes packed with subtlety and (worse for computers) ambiguity, an algorithm is required to decrypt the most likely solution.
On the official page, the company exemplifies with the following phrase "Alice drove down the street in her car." There are two ways to interpret this. One, you already guessed, that Alice was driving in her car down the street. The second, absurd but grammatically correct, is that Alice’s street was placed inside her car, and she was driving in it.
According to Google, SyntaxNet accesses neural networks and Beam Search in order to figure out which of the two is correct.
"It is not uncommon for moderate-length sentences – say 20 or 30 words in length – to have hundreds, thousands, or even tens of thousands of possible syntactic structures," the company says.
The blog post that announces SyntaxNet going open-source mentions that a parser of a natural language must shuffle through all these possibilities and select the one that best fits the context.
Tapping into the open-source framework opens a myriad of possibilities for developers. By using it, Google’s research gets put to good use in innovative assistants and apps or even polish existing software.
Big names in the industry could use improved parsing abilities. For example, userscomplainthat Facebook Messenger’s chatbots are handling verbal instructions poorly and could use an overhaul.