IBM releases UIMA source code as open source

IBM Monday released the source code for its Unstructured Information Management Architecture (UIMA) to encourage independent software vendors (ISV) to use the framework for the creation of complex, enterprise-ready text analytics applications on a standards-based platform.

In an announcement Monday, IBM said the UIMA code is now available as an open-source development project on SourceForge.net.

While traditional content and knowledge management applications today allow users to search for terms, they don't allow searches for concepts or relationships between words in documents, Web sites or other text, said Marc Andrews, a spokesman for content discovery strategy and business development at IBM. Complex text analytics applications from various vendors do provide that kind of analysis, but plugging them into existing search applications can be difficult because of code compatibility issues, he said.

The idea behind UIMA is to have a standards-based platform developers can use to create specialized text analysis applications, which can then be tied in by users with the search applications of their choice. UIMA defines a common, standard interface that enables text analytics components from multiple vendors to work together.

"Customers [have] had to do the integrations themselves because there are no interfaces" between proprietary text analysis applications and search products, Andrews said. "They've had to custom-tie them together," which is often difficult and costly. "UIMA enables them to tie these things together more easily, providing plug-and-play in a common language."

Last August, IBM announced that more than 15 ISVs, including SAS Institute, Cognos, ClearForest and Attensity, had pledged to support UIMA in their text analytics and search products. IBM also introduced its own offering, IBM WebSphere Information Integrator OmniFind Edition, which is based on UIMA.

Text analytics can comb through documents, comment and note fields, problem reports, e-mail, Web sites and other text-based information sources, according to IBM, which worked on the development of UIMA for more than four years.

Several medical institutions are using UIMA to help organize huge amounts of unstructured data that could be useful in medical research, according to IBM.

The Mayo Clinic is using it to help extract and collect data from some 20 million clinical notes in medical records that will be used for research and to improve patient treatments. The Memorial Sloan-Kettering Cancer Center is extracting data on cancer treatments from its records to search for new cancer treatments.

In addition, the International Federation of Pharmaceutical Manufacturers and Associations, a worldwide industry body that represents pharmaceutical companies, recently deployed a portal of clinical trial information that uses the UIMA framework with IBM's OmniFind application to identify medical terms and concepts. That allows doctors, pharmacists, researchers and others to search by disease area or medicine names. The tool even recognizes synonyms across multiple languages. The portal will be used to bring together content from a number of existing clinical trial registries and databases, allowing doctors and patients to review summarized results and find trials they can join, according to IBM.