Cognition Releases the Largest Semantic Map of English Language

Today Cognition, creator of an already much-discussed natural language processing technology, is announcing release of its new semantic map which is the largest one ever created for English language. For you to imagine the size of the map, it includes more than 10 million semantic connections, over 4 million semantic contexts, more than 536,000 word senses, 75,000 concept classes, 7,500 nodes in the classification scheme, and 506,000 word stems:

It took the scientific team in Cognition 23 years (and over 300 person years) to build this map which is a timeframe unheard of for a web startup and should give you an additional hint at how large-scale the development behind the technology is.

To summarize the announcement, I think it will be no exaggeration to say that after those 23 years of research Cognition has basically taught a computer to understand the meaning of virtually all the words and phrases in the English language - and from what I have seen on their website where they demonstrate examples of how the technology performs and processes queries it is absolutely true and definitely no exaggeration.

And while for us in the technology blogosphere anything “semantic” is immediately associated with “search”, search is actually not the focus for Cognition as they have described so many applications for the map when I talked to them yesterday that I am completely amazed about all the opportunities here. The possible fields of application for the newly released semantic map include personalizing and filtering content (something that must be able to significantly improve relevance of contextual advertising, for example), reducing information noise and producing more relevant search results as well as helping to enhance machine translation.

Unlike Powerset, Cognition is not building a search engine of its own - instead they have chosen to develop the technology and the semantic map to empower developers of other applications where the technology could significantly improve performance and user experience. The idea is that Cognition serves as a language understanding tool (or add-on) for new and existing software and web applications. The map can be integrated into such applications via an API if the developers see potential for their products when enhanced with a natural language processing technology and semantic map provided by Cognition.

For us to get a general idea of how Cognition performs, they offer a few examples on their website - search Wikipedia content, search the database of court decisions for some legal information you may need or search the database of health-related information. I have performed a few queries myself and I was surprised to see how accurately Cognition determines the meaning of a word based on its context - even for the most ambiguous words.

In general, I am more than impressed by what I’ve seen and the opportunities for a whole range of various enterprise and end-user applications that could be dramatically refined using Cognition technologies. The only question that I now have is if Cognition will remain independent with its valuable technology or if some big guys (Google anyone?) will rush to acquire the company to use the technology exclusively in their products.

Here is a link to two video demos of how Cognition works if you want to understand it better.