]> 2012-02-10T06:32:29-05:00 Encoding extraction as inferences The analysis of natural-language text involves many different kinds ofprocesses that might be described in multiple ways. One way to describethese processes is in terms of the semantics of their requirements andresults. Such a description makes it possible to view these processes asanalogous to inference rules in a theorem-proving system. This analogy isuseful for metacognition because there is existing theory andinfrastructure for manipulating inference rules. We describe a taxonomyof text extraction tasks that we have represented as inference rules. Wealso describe a working system that encodes the behavior of text analysiscomponents as a graph of inferences. This representation is currentlyused to present browsable explanations of text extraction to a user; infuture work, we expect to perform additional automated reasoning over thisencoding of text analysis processes. inproceedings Proceedings of AAAI Spring Symposium on Metacognition in Computation KSL-04-06 Encoding Extraction as Inferences KSL-04-06 2005 Year Tr id Title Computer science Tag Inference web Relation AAAI Press Publisher Media:KSL-04-06.pdf Paper Key Booktitle Bibtype Deborah L. McGuinness Christopher A. Welty David A. Ferrucci Paulo Pinheiro da Silva J. William Murdock Author Stanford University, USA Address Abstract KSL Technical Report Publication Proceeding Paper Workshop Paper