Subscribe via RSS Feed Connect on Google Plus Connect on Flickr

Glossolalia: Parsey McParseface

Thursday, 19 May, 2016

It’s the week of Pentecost, which is associated (Biblically) with “speaking in tongues,” a phenomenon linguists call glossolalia. So, in honour of all things syntactical, we’re devoting this week’s posts to language. We began with Singlish, followed up with decacorns, continued with Euro English and today we’re venturing into open-sourced language parsing, which is central to creating better voice recognition technologies for our mobile devices.

Google I/O, the company’s annual developer conference, began yesterday and the focus is on machine learning and VR, and how these technologies are being used in its core products. For example, Allo is a new app that merges text messaging with a virtual assistant. When it launches this summer, Allo will “monitor” your conversations and offer relevant information. So, if a friend in Manchester invites you out for an Indian meal, Allo would suggest a nearby Balti house. Useful, innit?

In the build up to I/O, Google released SyntaxNet, its open-source neural network framework, which includes Parsey McParseface, an English language plug-in. SyntaxNet provides a foundation for Google’s Natural Language Understanding systems, such as the voice recognition capabilities of the Google Now intelligent personal assistant. Parsey McParseface is based on machine learning algorithms that analyse sentence structure to understand the role of every word and grammatical element.

parsing

“One of the main problems that makes parsing so challenging is that human languages show remarkable levels of ambiguity,” Google explained in a blog post. “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. A natural language parser must somehow search through all of these alternatives, and find the most plausible structure given the context.”

Google claims Parsey McParseface has achieved 94 percent accuracy interpreting English language news articles. Although not perfect, that’s good enough to be useful in a range of applications, it says.

Note: Despite its popularity, Boaty McBoatface did not became the name of the British government’s new polar research vessel. But it lives on, kind of, in Parsey McParseface, Google’s wry name of its English language parser. Where there’s humour, there’s hope.


Comments (1)

Trackback URL | Comments RSS Feed

Sites That Link to this Post

  1. Glossolalia: Aramaic lessons | Eamonn Fitzgerald Rainy Day | Friday, 20 May, 2016