We All Have a Need to Know
A former U.S. President may someday live down the quotation "It depends on what the meaning of the word is is." His sympathizers might include those who devise natural language database query systems.
To a speaker of Spanish, that presidential faux pas is just an observation that English makes one verb do the job of two. Spanish lets a speaker distinguish between es, the "is" that describes a characteristic (my name is Peter), and esta, the "is" that describes a transient state (the sun is shining). Its easy to imagine a database query that depends on which "is" is meant: "Is there a reservation for Mr. Coffee?" could mean, "Is there anything in the reservations database, past, present or future, involving Mr. Coffee?"or it could mean "Is Mr. Coffee expected today?"
A system of this kind should tell you what it thinks you were asking and tell you in a way that anticipates follow-up queries. One natural language system that passes both of these tests comes from iPhrase Technologies, whose product handles queries at the Charles Schwab financial services site. I challenged Schwab.com with the spontaneous query (not a vendor-suggested example) "Show me p/e for technology stocks with market cap more than $1 billion" and got back a table that summarized the results by industry segment with one-click drill-down into each companys details. Very nice.
Perversely, just as we develop the tools to grapple with a worldwide knowledge base, proposed legislation would force all IT devices to collaborate in limiting access to that knowledge. As if the Digital Millennium Copyright Act werent bad enough, the ominous Security Systems Standards and Certification Act would impose digital rights management functions on a huge variety of devices and their userssaying, in part, "It is unlawful to ... traffic in any interactive digital device that does not include and utilize certified security technologies."
Well soon be able to ask a machine "What do you know thats worth knowing?" Unfortunately, its only answer may be "Im sorry, but I cant tell you."