Cognition Searches For Understanding

The company's alternative search engine to Google, Microsoft, and Yahoo aims to enable the Semantic Web.
Search doesn't work. That's the mantra of most search startups today.

To some extent, it's something said to solicit funding from venture capitalists and to ward off concerns about some rather well-funded competition in the form of Google, Microsoft, and Yahoo.

Truth be told, many of the failings of search engines reflect the shortcomings of searchers. Those who teach themselves to understand search engines tend to get better results.

The alternative is teaching computers how to understand people. That's the approach Cognition Technologies has taken with CognitionSearch, a new meaning-based linguistic search engine.

Scott Jarus, CEO of Cognition Technologies, says his company has essentially taught CognitionSearch the meaning of some 375,000 English words. That's almost 10 times the average vocabulary of the average college graduate, he adds.

"We're the first step toward the Semantic Web and enabling the Semantic Web to exist," says Jarus.

The Semantic Web, sometimes referred to as Web 3.0 or as a subset of Web 3.0, is a vision of the Internet as a structured database. Today's Internet is largely unstructured information. As a consequence, searching for "house bill" on Google returns results related to legislation as well as a link to a site about Bill Gates' house. That's because Google has no idea whether you're using "bill" in the legislative sense or as a proper noun.

CognitionSearch has no idea either, but at least it has the sense to ask. Though its index of documents is far more limited than Google's -- it currently supports vertical searches related to health, politics, and law -- it will ask which of the six possible definitions of "house" and which of seven possible definitions of "bill" you meant in your query.

Jarus shows how CognitionSearch might be valuable for lawyers. Using a database of Enron e-mail messages and his company's enterprise search interface (rather than the online site), he queries for "energy bill."

A keyword match finds 8,290 documents, compared with 934 using CognitionSearch tuned to the specific legislative meaning of the query. Lawyers or paralegals conducting such a search could save a significant amount of time with almost one-tenth as many documents to review.

More significantly, Jarus points out that 153 of the documents CognitionSearch returned were not in the keyword search list because they did not contain the words "energy" or "bill." Rather, they contained synonyms that mapped to the queried terms, such as "utility regulation." In litigation, that's the sort of thing you don't want to miss.

CognitionSearch is being offered to enterprise customers in litigation support and the life sciences. The online search site is more of a technology demonstration than a replacement for Google. But that may change in time. Cognition intends to add more content on a regular basis.

Editor's Choice
Brandon Taylor, Digital Editorial Program Manager
Jessica Davis, Senior Editor
Cynthia Harvey, Freelance Journalist, InformationWeek
Terry White, Associate Chief Analyst, Omdia
John Abel, Technical Director, Google Cloud
Richard Pallardy, Freelance Writer
Cynthia Harvey, Freelance Journalist, InformationWeek
Pam Baker, Contributing Writer