At its core, MindServer provides a base platform that enables the automated analysis and topic-based classification of both structured and unstructured data; as well as provides semantics-aware information retrieval capabilities from that analyzed information that is based not strictly on keywords, but instead on the latent concepts or topics discovered within the document content.

The technology that drives the MindServer product line is proprietary to Recommind, developed by their Chief Scientist, Dr. Thomas Hofmann. Called "Probabilistic Latent Semantic Analysis" (PLSA), the technology seeks to understand the concepts within documents (information provided by the vendor notes that over 200 document/application types are supported for analysis), creating a statistical model from a document's contained information based on the analysis of word co-occurrences and identification of repeatable contexts, topics, or concepts within which certain word groupings occur. Because the technology utilizes the actual content of the document provided to perform this analysis, it does not require pre or manual training (i.e., the manual feeding or maintenance of lexicons or external taxonomies), though such tools can be incorporated by the platform to enhance its results if they are available. Because it does not require knowledge of language-specific lexical semantics to perform its analysis, the technology is said to support multiple languages natively (a white-paper from the vendor notes that the base PLSA technology "...is applicable to any language, as long as the language can be tokenized").

In addition to these concept-based categorization and analysis capabilities, an additional feature of the PLSA technology is its ability to recognize semantic similarities between words that often occur in the same contexts. This recognition goes beyond basic synonym recognition (relating "car" to "automobile," for example), to automatically relate words that are not synonyms but nonetheless frequently occur in relation to one another (such as "automobile" and "driving"). Semantic recognition is accomplished through the use of numeric computations that indicate the probability of specific words to occur in a particular concept. Such analysis and computations enable the future retrieval of the analyzed data based on concept-based analysis of the key words provided; i.e., if a user searches for "automobile" they can retrieve not only documents that contain the word "automobile," but those that contain concepts strongly related to "automobile," even if the retrieved documents don't actually contain that specific word.

On top of this core technology, Recommind offers the MindServer product line; and its two primary components: MindServer Categorization, for the automated categorization of document collections; and MindServer Retrieval, for the searching and returning of information. And beyond these core components, additional product bundles and modules are offered for more specific information management environments; including MindServer Legal, which includes Searching, "Matters and Expertise," and categorization components; and MindServer Media, which focuses specifically on categorization. An additional module, "Projects & Expertise," contains search and expertise location components, and focuses specifically on the locating of projects and people within an organization.

Other product features include support for parallel processing in multi-core and distributed environments, XML-compatibility, and an API (Java, CORBA/DCOM) allowing for the interfacing of the platform with custom applications.

Among the key features in the new release of MindServer is the ability to perform "Litigation Holds" of retrieved documents for legal purposes; i.e., the ability to lock-down a retrieved document set, either directly from within MindServer (for those platforms with a specific MindServer connector), or, more likely, by passing the search result set to separate applications or databases to perform the actual lock down. This Litigation Hold feature is available in the search components of both MindServer Legal and Projects & Expertise.

Other new features in the latest release include support for 64-bit hardware, and customized connectors for document and knowledge-management products from vendors including EMC/Documentum, LexisNexis, and Microsoft.

MindServer is available now and is priced per processor; including the Litigation Hold feature described above, pricing starts at $75,000/processor.

Visit the Recommind Web site for further information.