In a Thursday post on the company blog, Alphabet Inc. subsidiary Google announced the launch of a new open-source neural network framework called SyntaxNet, along with a new English language parser – Parsey McParseface – which has been trained using SyntaxNet.
Advertisement
It is hard for computers to get parsing right because of the amount of ambiguity found in human languages. In fact, Google says it’s not uncommon for sentences 20 or 30 words in length to have up to tens of thousands of possible syntactic structures.
Search engine Google appears to have had some trouble naming its new language parsing model and dubbed it Parsey McParseface. This essentially means AI developers can now use the SyntaxNet technology to make any (supported) computer understand human language more accurately.
A syntactic parser like SyntaxNet is created to look at any human-language sentence, tag each word according to what part of speech it is (i.e., subject, object, etc.), determine the relationships between each word and, from there, decide on the most likely meaning of the sentence.
“For the sake of simplicity, let us consider an example sentence – ‘Alice saw Bob”. For example, you can tackle language understanding with an end-to-end approach just feed the words in and hope that meaning comes out.
Parsey McParseface has a benchmarked 94 per cent accuracy rating, which is probably more than most people. Machines can take things rather more literally, so giving them a neural network to help them crunch parsing permutations helps things along.
As part of the deal, which was aimed at promoting scientific research into revolutionising human-machine communications, Oxford professors, who were behind startups Dark Blue Labs and Vision Factory, joined Google’s DeepMind team.
‘While there are no explicit studies in the literature about human performance, we know from our in-house annotation projects that linguists trained for this task agree in 96-97% of the cases.
Parsing is the process of the software breaking down the sentence in order to understand the words and context.
Advertisement
“Our work is still cut out for us: we would like to develop methods that can learn world knowledge and enable equal understanding of natural language across all languages and contexts”, the company said.
