Search engine speaks Maori

By ADAM GIFFORD

The Department of Conservation has bought an Australian-developed "state of the art" search engine which enables visitors to its website to ask natural language questions in English or Maori.

Chief information officer Channa Jayasinha says the 80-20 Discovery search engine also powers DoC's intranet, where it is expected to reduce administration costs and lead to greater user satisfaction and productivity among the department's 1500 staff.

"The homegrown search engine we had on the website wasn't working, and on the intranet we had multiple search tools depending on the application."

Mr Jayasinha says the cost of the new tool, including licences and installation fees, was $95,000.

Maintenance costs are minimal because of the way the algorithms in the software look for word patterns and concepts in the stored text, removing the need to manually create keywords or indexes or otherwise cleanse data.

The technology is language independent, so it is able to handle the large number of documents in Maori on the site.

David Gillespie, from 80-20 Software, said the algorithms which drive Discovery were discovered at Telstra's pure research division and licensed to 80-20 for commercial development.

The privately owned company, whose shareholders include Intel Corporation, GE Capital and Australian venture capital fund Allen and Buckeridge, had previously turned some Telstra technology into the Microsoft Exchange-based 80-20 DME (Document Management Extensions) document management tool also used by DoC and the Public Trust Office.

"Most companies are faced with exponential growth of documents, and they don't have the time or resources to do the stuff needed to keep track of what's in them," Mr Gillespie said.

"The value of Discovery is you don't need to do any of that stuff. A DoC or a Telstra can now get a conceptual search engine which can point to a diverse range of content and file formats without anything needing to be done manually."

A research report on categorisation and search software done by United States investment bank Merrill Lynch said the Darwin concept retrieval technology underpinning Discovery used neural networks to infer relationships and correlations without explicit coding.

"Neural networks sort through raw information to divine complex interrelationships between elements," Merrill Lynch said.

"Algorithms are used to 'learn' the relationships in the data. Learning rules enable the network to derive knowledge from available data and apply that knowledge to the categorisation process.

"We regard the [80-20's] technology as state of the art."

80-20 Discovery is an enterprise product charged for on a per-processor basis and requiring significant processing power.

But some of the same technology is in 80-20 Retriever, a $US50 version downloadable from the website which indexes everything on a PC for swift retrieval.

Department of Conservation

80-20