Annual IBM-Berkeley Day: Search and Mining as an Innovative Technology

06/10/2004 02:14 PM

IBM-Berkeley Day Participants

More than 50 distinguished educators, scientists and business leaders gathered for the fourth annual IBM-Berkeley Day on May 20 at the University of California, Berkeley. This year's theme -- "Mining Your Business!" -- featured discussions focused on the ramifications of very large-scale text mining on technology, business and society.

Created four years ago and still chaired by Jean-Paul Jacob, Almaden emeritus researcher and university relations manager, the purpose of the annual IBM-Berkeley Day is to solidify personal and professional relationships between IBM researchers and UC Berkeley (UCB) faculty and students by spending a day interacting during dynamic, high-level presentations and demonstrations. "It is amazing what can be accomplished when researchers collaborate from different disciplines and from different institutions, such as Berkeley and IBM. This is a prime example of "instant innovation." Last year's event resulted in a new course being taught at Berkeley this fall," said Jacob.

UCB Dean of Engineering, Richard Newton, kicked off this year's event with a call for faculty and students to "embrace new business models to support large-scale text mining."

The morning presentations included talks given by Tom Campbell, UCB Dean of the Haas School of Business and former congressman; Robert Morris, IBM Almaden lab director; Eric Brewer, computer science professor and one of the creators of Inktomi; and Robert Carlson, IBM vice president of WebFountain.

Campbell described the Haas School of Business and its interdisciplinary programs, and commented about the positive impact of innovation on California's economy.

While business and society can benefit from new technologies, Morris challenged the audience to ponder creating value for the world economy without exascerbating conflict. "As we search for new disruptive technologies, we need to pay attention to elements of social resposibility, including security and privacy," he said.

In his keynote speech, "Search Engines as Databases: An Inktomi Retrospective," Brewer gave his thoughts on how very large-scale text data stores could be used for text mining. "The future of search depends on merging traditional database technology and search engines to enable both more powerful queries and combinations of structured and unstructured data," he said.

Carlson provided an overview on the infrastructure of WebFountain, including insightful comments about how this technology can fulfill Brewers' vision. "These new models for supporting very large text data stores hold enormous promise for supporting breakthroughs both in research and commercial applications," said Carlson. He further spoke about how text mining supports innovative applications focused on revenue generation through new insights into customer needs and market environments.

Afternoon sessions covered four major topics highlighting 1) data and the need to understand it better; 2) infrastructure and learning from the world of databases, 3) algorithms and related issues; and 4) user applications.

The WebFountain team demonstrated various user applications that stand to significantly improve understanding of very large data sets. W. Scott Spangler, from Almaden's Text Mining for Collaboration group, showed off eClassifier, a knowledge management toolkit for gaining new insights from unstructured data; Keiko Kurita, WebFountain Solutions Marketing Manager, demonstrated Factiva Insights for Reputation, a powerful new tool built on the WebFountain platform for discovering emerging business issues and social trends affecting an organization's greatest asset -- its reputation; and Allen Cypher, from WebFountain's User Experience group, wrapped up the demonstrations with an information discovery and analysis application that explores billions of web pages to answer specific business questions.

Professor Frederic Gey, UCB research associate and director of Technical Services, spoke about research in extracting data from tables of information on the Internet. UCB Computer Science Professor Stuart Russell provided an overview of research on automatically extracting citation information in his talk titled "Truth and Appearance."

The panel discussion, chaired by Wayne Niblack, manager, Almaden WebFountain Information Summarization, and Ross Nelson, from Almaden's WebFountain Miners/Usability group, described more than a dozen text-mining algorithms in use today in WebFountain, including disambiguation of subjects, geographic name-spotting and associations. In another talk, "Personalization, Mining and Privacy," UCB Professor John Canny described methods for personalization and issues around privacy.

Later, Rakesh Agrawal, IBM Almaden Fellow who leads Intelligent Information Systems Research, presented his views on information sharing, searching and mining while preserving privacy in his talk, "Sovereign Information Sharing, Searching and Mining." UCB Professor Joe Hellestein, who spoke fondly of his internship experiences at IBM Research, provided an overview of Internet security issues and the work underway at Berkeley to help raise awareness on these issues.

As Kevin Mann, Almaden's Global Industry Analyst for Banking and Financial Markets and one of the main organizers of the event, summarized at the end of the day, "the presenters have demonstrated that text mining can provide useful business insight today, and gave us a vision of how new text-mining infrastructures, such as WebFountain, will be disruptive technologies. Nearly instant and comprehensive insights into the global dialogue on the Internet provide rapid sensing of business issues, pushing companies to respond much more quickly than ever before."