Researchers have developed an intelligent text-mining tool which is tuned to the way humans think, speak and ask questions.
Type "what Steve Jobs said yesterday" into the so-called "FactSpotter" tool and the software will hunt through documents and return a handful of relevant answers, instead of churning out countless articles containing the Apple chief executive's name.
But the FactSpotter software will not be available to the public over the internet or otherwise — it will only be available to customers of document management company Xerox, which developed the text-miner tool.
Jean-René Gain, director and general manager of marketing, strategy and alliances at Xerox, said Xerox will not sell FactSpotter as a stand-alone application but as an embedded app to its customers.
Gain said: "We are not taking on Google with this. It is an aside option to consider but we need this technology to differentiate ourselves [from our competitors]."
Mario Jarmasz, a technology showroom engineer at Xerox, said: "Xerox is not creating a Google killer. This is completely different from searching on Google because we can drill down to certain levels of detail."
The document-heavy legal and litigation market will first be offered the FactSpotter tool, which will be available by 2008.
Xerox also predicts the text-miner software will be useful in other situations where information must be retrieved from a massive database, including corporate and government searches, drug discovery, fraud detection and risk management.
Christopher Dance, laboratory manager at Xerox, said FactSpotter could also be used to manage the vast number of documents which are produced during large mergers and acquisitions.
FactSpotter can hunt for relevant documents at a rate of 2,000 documents per second. Dance explained the next stage of the development process will be to speed up the software.
The tool uses a linguistic engine which analyses the meaning of words and the construction of phrases and sentences to work out exactly what a user is hunting for.
FactSpotter also recognises concepts in a search term. To use the previous example, when a user types in "what Steve Jobs said yesterday", the tool will break down the sentence and recognise "Steve Jobs" as a person and "yesterday" as a time.
Frédérique Segond, parsing and semantics area manager at Xerox, said: "[Xerox is] trying to make a computer understand text like a human being."
Segond added that the FactSpotter tool is the next step for searching documents and uses "Web 3.0" technology which connects data, whereas Web 2.0 apps only collect data.