UMIACS Computational Linguistics Colloquium Series, DATE

UMIACS Computational Linguistics Colloquium Series, DATE


Gigahertz Processors, the Internet, and Commercial NLP
Paul Jacobs
IsoQuest

Many of us in the practice of artificial intelligence have a thirst that's hard to satisfy: We want not only to discover breakthrough algorithms that solve the fundamental problems of human understanding, but also to see our discoveries impact the way people use and interact with computers. This desire to have it all often leads to a tension as we overestimate the potential of our advances - often "too good to be true" - in the face of simpler, "brute force" approaches. In this conflict, brute force always wins. Finally, after decades, Deep Blue finally beats the world chess champion. Rather than be jubilant in this incredible, historic event, the world of AI laments because it was brute force - 200 million positions per second - that led to the victory.

The same conflict, and the same opportunity for historic achievement, lies in natural language processing, if we can only learn the lessons of these years. Linguistics can?t beat brute force any more than blackboard algorithms and case-based reasoning can play chess. Therefore, the opportunities we face in applying language technologies - such as information retrieval, spoken language interpretation, and corporate knowledge management - all require us to embrace brute force and very carefully and selectively sprinkle brilliant linguistic methods on top of Moore's law and mountains of data, which we must recognize are the real drivers of our current revolution.

This talk will present examples of real innovations in natural language processing and how they are finding their way into significant advances with real commercial impact. Focusing on information retrieval and knowledge management, we will explore the relationships between NLP research and the commercial systems of today along with the exciting opportunities for real innovation in the coming years. In each area we will bring to bear some examples of companies that successfully apply NLP in valuable applications.


Return to the UMD Computational Linguistics Colloquium Series.