Coruscation

Wordnik, which has raised $12.8 million in venture financing, plans to use its vast database of words and word associations at the site and in many business partnerships to be announced this year, said Joe Hyrkin, the president and C.E.O.

The products will be similar to recommendation engines, but more powerful, he said. If you like a particular book, for example, Wordnik can recommend a similar one based on its understanding of words used to describe the book, he said.

"We're not just using tags and descriptors," he said. "Our system understands and identifies matches at a concept level."

The company is already providing many other word-based services, including one used on the Web site of The Times to define words in articles. Wordnik is also providing a financial glossary for SmartMoney.com.

Geoffrey Nunberg, a linguist at the School of Information at the University of California, Berkeley, who talks about language on "Fresh Air," the NPR program, appreciates Wordnik's breadth. "There's a lot of useful information here," he said. (He has also written commentaries on language for The Times.)

But he thinks that hands-on lexicographers could fine-tune the entries.

"The idea that you can pull lexicographers out of the loop and have an algorithm to mediate between me and the English language is goofy," he said. "Without hand citations done by trained people, you get a mess."

To illustrate his point, he noted flaws in a number of Wordnik's definitions. The first definition of "davenport," for instance, in three of the fives sources used by Wordnik is a kind of small writing desk. "It hasn't meant that since Grandma was a girl," he said.

People use a dictionary to find out what is correct, and what is incorrect, he said. "If I were a journalist looking to see if a word was being used correctly," he said, "I wouldn't put my eggs in the Wordnik basket."