An update of Recommind Inc.s MindServer knowledge management tool provides enhanced information categorization and retrieval capabilities.
MindServer 2.1, which was released late last month, categorizes unstructured data contained in documents such as presentations, e-mail, Web pages, news feeds and contracts by automatically identifying concepts that describe a document, regardless of the language or subject, said Recommind officials, in Berkeley, Calif.
Version 2.1 offers a new feature called text extraction, which can identify specific entities buried within text, such as personal names, organizations, places, dates and phone numbers. The software can then analyze relationships between those entities.
MindServer 2.1 includes categorization tools that enable automated or interactive classification and taxonomy management and new administrator tools for setting up and tuning searches.
The Research Libraries Group Inc. is using MindServer 2.1 to help its customers find information from the more than 110 million records in the organizations database, mostly research data from universities and large libraries.
“[MindServer] does a couple of things we didnt even think were possible,” said Jim Michalko, CEO of RLG, in Mountain View, Calif. “It really is language-independent. If you issue a search in English, it can recommend results in German or Japanese, related to your search query. A lot of results can be buried that people otherwise wouldnt know how to retrieve.”
Michalko said MindServer 2.1 is able to discern concepts within search results so that even if a search request is very broad, the results will be distilled into categories.
“Lets say you type in a word like Iraq; MindServer will extract the results into associated concepts, whether youre looking for information on ancient Mesopotamia … or the political situation today. People can more easily find what theyre really looking for,” Michalko said.
Pricing for MindServer 2.1 starts at $75,000, although some implementations can cost as much as $1 million based on the number of records to be searched and categorized, Recommind officials said.