Cognition: Natural Language Processing for Archiving, E-Discovery, Compliance
Cognition offers natural language processing software. This is useful in various ways, including: search through electronic archives, e-discovery, and compliance.
- Natural language processing
- Customer integrates with its own technology
- Cross industry applications
- However, the company has particular success with legal, healthcare, publishing verticals
- Eg, Lexis/Nexis offers Concordance, a widely used litigation system. Cognition's software helps users be able to search a large corpus conceptually, rather than on a key word basis. When integrated into eDiscovery (Legal) applications, precision and recall of salient documents significantly improves, which helps lower e-discovery costs and provides a much clearer understanding of the content
- Based on amount of data being indexed
- List prices range from $15K, to $500K for unlimited use
- After first year, 20% maintenance fee
- Founded originally in 1984; until 2003 Cognition focussed on custom software development for US government, large corporations, in the natural language processing field
- In 2004 Cognition introduced its first commercial product
- Privately held
- Angels provided funding until 2007, a total of about $4M. In 2007 Cognition raised $2.7M in a Series A round, investors include Tim Draper (of Draper Fisher Jurvetson fame)
- Still at early revenues stage, not profitable
- 12 employees
- When searching over a body of electronic material, the most common approaches are to search for keywords, or patterns ("regular expressions"), or to use email header information such as To or FROM addresses; or to search for simple combinations thereof ("ands" and "ors" and "nots")
- There's a need for technology that searches in a more flexible way, corresponding to human concepts
- That's what Cognition offers
- Has obvious value for searching archives, e-discovery, and compliance
- Natural language processing is a hard problem; many companies claim, erroneously, to do this
- While doing Artificial Intelligence research at Stanford in the 1970s, your editor wrote a natural language front end for a system that checked qualitative probability theorems.
- It was fun. Plus it paid his way through college.
- Plus he discovered that Chomsky was wrong: generative grammars aren't sufficient to characterize the richness of ordinary language. ... David Ferris