One of the biggest differences between the human animal and all the other is how we managed to develop a language that uses syntax. Grammar might not seem important to some people, but it is one of the few things that separates us from animals. Like the internet would say, grammar is love, grammar is life.
Many things influence the way grammar works, and the fact that many language constructs are highly dependent on context makes reproducing human speech very difficult. A team of researchers realized just how difficult this task is, as Google’s Parsey McParseface helps computers learn English.
With its name inspired by the recently named British research vessel Boaty McBoatface, the now open-source tool for better language parsing in advanced AI can basically help computers learn the English language. It helps them differentiate between different meanings of a sentence, based on context.
For example, as presented on the company’s official website, take the sentence “Alice drove down the street in her car.” It can be interpreted in one of two ways, depending on whether you actually understand how the language works and the context of the phrase.
The first way to look at the phrase is that Alice was driving in her car down the street. It’s a simple sentence to us, because we know it makes sense. An AI could very easily have trouble with the way the sentence is phrased, and it could very realistically interpret it as a street being placed inside Alice’s car, and that she is driving down it.
What Parsey McParseface does is to ensure that computers understand this difference, that they understand verbal interaction. This subtlety and ambiguity so typical of all the languages on the planet would greatly confuse a computer without a parsing tool to help it out.
By going open-source with its parsing tool, Google is basically giving developers access to the SytaxNet framework. This is what actually helps computers understand our language. Parsey McParseface is just the module that connects to the aforementioned framework. It also works for the English language.
And the company is very confident in its framework, claiming the tool is able to identify with a 94 percent accuracy the differences between how verbs, subjects, or objects should be interpreted. And once coders start deploying it in their own software, this accuracy will only grow higher.
This is owed to the machine learning algorithms used by Parsey McParseface. These algorithms ensure that every time the program runs into a new interpretation, it will learn from it. Every time Parsey McParseface is used, it will only get smarter and better at speaking the English language correctly.
Image source: Pixabay
Latest posts by Alan O’Leary (see all)
- Woman Found Alive After Missing for 42 Years - Oct 30, 2017
- October Will Welcome The Draconid Meteor Shower And The Orionids - Oct 6, 2017
- Scientists Are At A Loss After Unearthing A Porpoise Grave - Sep 22, 2017