University of Minnesota
University Relations
myU OneStop

Go to unit's home.

Home | Seminars and Symposia | Past seminars/symposia: Monday, November 7, 2005

Toward machine understanding of unconstrained natural language text


William Schuler
Department of Computer Science

Monday, November 7, 2005
3:00 pm

402 Walter Library

Download slides (pdf 49 KB) This talk will explore the state of the art in semantically deep statistical modeling techniques for extracting large databases of propositions from unconstrained natural language texts. Deep semantic processing (also called interpretation, or Natural Language Understanding) has long offered the promise of extracting very specific relations from unconstrained text, in contrast with the very general information provided by text mining approaches such as latent semantic indexing, but the reality of current deep semantic approaches is that they are severely constrained by the lack of semantically-annotated training data for broad-coverage domains. This talk will describe current efforts to define and then simplify the task of annotating deep semantic information, in order to train robust statistical models capable of understanding unconstrained natural language text.