MITRE-Bedford: Description of the ALEMBIC System as Used for MUC-4
Abstract
The ALEMBIC text understanding system fielded at MUC-4 by MITRE Bedford is primarily based on natural language techniques . ALEMBIC is a research prototype that is intended to explore several major areas of investigation : 1)Error recovery, involving primarily issues of semi-parsing and recovery of plausible attachments; 2) Robustness, involving primarily issues of uncertain reasoning and tractable inference; 3)Self-extensibility, focusing primarily on machine learning of natural language and user - configurable semantics; and 4) System integration, through SGML (the Standard Generalized Markup Language), both at the level of meaning analysis and at the overall application level. This investigation is part of an internally-funded research initiative towards processing open source texts. The system's underlying architecture follows a task breakdown used in several other systems. Processing occurs in three distinct phases: preprocessing, natural language analysis, and application-specific output generation . One of the ways ALEMBIC differs from other MUC systems, however, is in exploiting SGML as the interchange lingua franca between these three processing phases. The intention is to allow system modules whose invocation occurs early in the analysis of a document to record processing results directly in the document through SGML markup . This information then becomes available to subsequent modules as meta-data.
Document Details
- Document Type
- Technical Report
- Publication Date
- Jan 01, 1992
- Accession Number
- ADA460609
Entities
People
- Dennis Connolly
- John Aberdeen
- John Burger
- Marc Vilain
- Susan Roberts
Organizations
- MITRE Corporation