"Enterprise information environments are so complex that no knowledge worker can be expected to understand and navigate the myriad information sources, systems, interrelationships, and business dimensions of the enterprise's intellectual capital," said Bob Tennant, CEO of Recommind. "The MindServer 4.2 platform enables an organization's most important assets to be utilized by employees and customers for improving performance and driving business objectives faster and more effectively."

The MindServer platform delivers automated search, retrieval, and categorization of text in documents, presentations, e-mails, Web pages, and newsfeeds—and now data from other applications—with what the company says is human-level accuracy. The technology behind it is Recommind's patented Probabilistic Latent Semantic Analysis (PLSA) algorithms, which were developed during research led by Recommind's chief scientist, Dr. Thomas Hofmann, and Dr. Jan Puzicha, Recommind's chief technology officer. A leading expert in the area of machine learning, Hofmann's ability to apply sophisticated mathematical models to real-world problems has resulted in the core of Recommind's technology. Puzicha is an expert in real-time information optimization and statistical modeling.

Tennant called PLSA the company's "secret sauce." He said that standardized academic tests have actually shown an average gain of 20 percent in relative precision of retrieval for PLSA over search technologies that use a "vector space" retrieval model. The conceptual search capability means that users can retrieve relevant documents that don't contain the exact keywords used for a search. To see this in action, visit the Recommind-powered search engine at MedlinePlus.gov.

MindServer now provides centralized access via a simple (Google-like) search box. It allows users to zoom in on search results using relevant filters—faceted refinement. It now offers new Web 2.0 usability with its enhanced "smart filtering." Users drill into result sets with the simple AJAX (Asynchronous Javascript and XML)-based user interface. Dynamic, on-the-fly computation of result sets helps users immediately analyze and focus in on the information they are looking for across millions of documents. The platform is an enterprise-class solution that can scale to serve millions of users and search terabytes of data.

Projects & Expertise, an add-on module for MindServer, is an upgraded enterprise version of Matters & Expertise, which launched in January 2005 and was specifically targeted to the legal profession. Using its capabilities, a user can find individuals with specific skills based on the projects they have worked on and the documents they have authored.

Oz Benamram, practice resources attorney for Morrison & Foerster, has been working with Recommind over the past year to implement enterprise search for the law firm, which has 19 offices and more than 2,500 employees. Benamram said that in seeking an enterprise search system the firm was guided by the e-commerce model. Retailers help customers select what they want by zeroing in on what is relevant—the law firm wanted to provide a similar system to enable its attorneys to easily and quickly find contextually relevant content. The best e-retail sites use faceted search to target important attributes of products, and so this capability was part of the law firm's requirements.

"As a global firm, we recognized the need for a single information access platform that enables us to search and discover meaningful information from massive volumes of documents, case files, e-mail messages, and other objects," said Benamram. The company interviewed 25 vendors—from enterprise search systems (including Google, Autonomy, and FAST), to specialized legal search vendors, to providers of social network and expert systems, and more. Three finalists were chosen based on a graded list of criteria; proofs of concept were done with two vendors and Recommind's product was chosen.

Contextual search and application integration were key requirements according to Benamram. MindServer sits on top of the company's document management system, e-mail archive, SQL databases, CRM system, and intranet portal. Other differentiators in favor of Recommind were its scalability, flexibility in development, price, and more efficient use of server space (MindServer required about one-third the space of other vendors). Furthermore, explained Benamram, the company's data is multidimensional and linked. The system needed to account for the "who, what, where, how, and why" related to a document. "Recommind lets users drill down in a meaningful way and also drill up to find additional related information."

[Note: Benamram is discussing his firm's implementation of Recommind's search-based solution in a presentation this Wednesday at ESS. Recommind is also a sponsor of the 2-day event and is participating in the Enterprise Search Showcase.]

Recommind's customers include the National Library of Medicine, RLG, and Bertelsmann AG. Founded in 2000 by researchers from the University of California-Berkeley, MIT, and Brown University, Recommind is privately held and said to be profitable. The company is headquartered in San Francisco, and has offices in New York, London, U.K., and Bonn, Germany.

The company has been focused on the legal market but with the launch of its enhanced platform is poised to expand into other enterprise markets. Tennant said, "We're number one in the legal market. We will definitely expand in the government sector." He said that other attractive markets include retail, financial services, and pharmaceuticals. The company also plans to add new personalization capabilities in a future release.