Energy strikes it rich with knowledge portal

Bob Wells knew he had a problem when his employees were spending more time searching for data than actually using it to do their jobs.

The problem was understandable. The 2,500 workers at the Energy Department's Office of Civilian Radioactive Waste Management must comb through more than a million electronic and paper documents as part of their research into sites being considered as permanent storage facilities for spent nuclear fuel and high-level radioactive waste.

The challenge is to determine if any information in those records is relevant to a particular license application, which means worrying about both context and content. That's a tall order, considering that, given the volume of records, staff members frequently were unsure if they were finding all the records needed to do their searches.

If ever there was a knowledge management problem, this was it.

So Wells, the office's chief information officer, tapped software vendor Autonomy Corp. to help develop what he describes as a knowledge management portal.

The system, which went online in late December in a limited release, reduces the workload significantly by creating a virtual repository of all the electronic documents available to the office and providing a powerful search engine to automate the process of finding and analyzing those documents.

The portal, which cost $900,000 to develop, is not comprehensive because not all documents are available in digital form yet. But Wells said he expects immediate benefits.

"There are a million records to reprocess manually, and we're looking for significant effects because there are a lot of experts with a lot of documents," he said. "If Autonomy captures 50 percent of documents, we only have 50 percent left for those folks to look at. We're expecting some serious savings."

No Mere Search Engine

Autonomy specializes in software for searching, analyzing and organizing information. One tool is a natural-language search engine that not only searches for keywords but also can identify words that are not identical but represent the same idea.

That was a critical feature for Wells. "Our keyword searches were not working," he said. "For example, if you looked for "hydrology,' documents discussing "water movement' were not picked up."

But it's more than a search engine. The software also can analyze documents with "hydrology" or similar terms to determine how significant a role those terms play in those documents. So rather than returning hundreds of documents where "hydrology" appears, the portal returns only the "most important and relevant pieces of information," said Rita Joseph, vice president for the public sector at Autonomy.

The system also has the ability to categorize documents, identify staff members who might be interested in those documents based on their past searches and even deliver the documents to them (see box below).

"They were sitting on a vast warehouse of information and didn't have an idea of what it was," said David Appelbaum, head of product marketing at Autonomy. "Our information portal brings order to that chaos, and data [to the workers], so they can easily find what they need and have a forum for identifying colleagues who are experts."

Autonomy has developed such tools for many federal customers, including Defense Department agencies, but this one stands out for the amount of content and the size of the documents, Joseph said.

"The content was different than news feeds in vocabulary and sensitivity," she said. "Also, they were geographically dispersed, and there are different types of repositories, [such as] e-mails with attachments to documents, document imaging, huge PDF files and huge Word documents."

The records management component of the portal is based on ForeMost software from Provenance Systems Inc., based in Arlington, Va., and Ottawa. The National Archives and Records Administration is considering a standard based on the same software.

In September, the DOE office began working with its East Coast personnel to identify the pockets of information available in its various offices. The knowledge assessment, completed in December, yielded some interesting results. "We found repositories we didn't even know were out there," Wells said. "The user community didn't know the information was there internally. We didn't know it was within our realm."

A few small groups piloted the portal to gather feedback before the formal launch at the end of December. "We used that feedback, along with the knowledge assessment, to determine the location of information that the user community was searching for," Wells said.

About 200 staff members in the Washington, D.C., office now use the portal, but the remaining 2,300 users in Las Vegas won't begin with the system until this summer, when their knowledge assessment, hardware purchases, deployment and training are complete.

The Payback

Wells said the typical return on investment for a knowledge and records management portal is anywhere from 7 percent to 20 percent, but his office used 1 percent as a baseline just to be safe.

"We expect 5 to 10 percent gains in productivity," Wells said. "Searching for information took a long time, and if a person left, we never captured that tacit knowledge. We had to re-learn where the repositories of information were, and there are deadlines for license applications."

There is no benchmark for measuring the time, financial or knowledge savings, but the first month's feedback has been positive and is already driving future changes to the system.

Currently, users must e-mail their suggestions and concerns to administrators, but a "feedback mechanism" will be added to the portal in the future.