There’s a new version of Lexalytics’ Salience Text Analytics Engine: Some of its key new capabilities in Version 6 are enabled by underlying Syntax Matrix technology that the vendor has been working on for the last 12 to 18 months.
Syntax Matrix, explains vp of product and marketing Seth Redmore, takes on the job of doing efficient chunk parsing, so that customers who can be dealing with hundreds of millions of documents a day can maintain that scale without sacrificing accuracy or performance. “What [chunk parsing] means is that we can tear apart a sentence to understand quickly how all the phrases in the sentence relate to each other,” he says, much as Salience’s existing Concept Matrix technology leverages Wikipedia to help it tell what entities are related to each other and how closely.
Deep learning infuses the Syntax Matrix, which is trained on billions of words to support its rich approach to extracting phrases, each with some 200 different features associated with it. “With deep learning we extract so many different features and understand all the interrelationships between them,” he says, providing users with the chunks of the sentence that are most interesting to them, and what they mean so they can take action. “Sentiment Matrix lets us tear apart these sentences in a grammatically meaningful fashion and do it in such a way that you can build other stuff on top of it,” he says.