Login

Publications  •  Project Statistics

Glossary  •  Schools  •  Disciplines
People Search: 
   
Title/Abstract Search: 

Dissertation Information for Karen Kukich

NAME:
- Karen Kukich

DEGREE:
- Ph.D.

DISCIPLINE:
- Library and Information Science

SCHOOL:
- University of Pittsburgh (USA) (1983)

ADVISORS:
- Roger Flynn
- Donald Shirey

COMMITTEE MEMBERS:
- Terry Noreault
- Jaime Carbonell
- Allen Kent
- Michael Lesk

MPACT Status: Fully Complete

Title: KNOWLEDGE-BASED REPORT GENERATION: A KNOWLEDGE-ENGINEERING APPROACH TO NATURAL LANGUAGE REPORT GENERATION

Abstract: This report discusses a new technique for automatically generating natural language reports from machine-readable databases. The governing principle of the technique is that natural language report generation is a knowledge-intensive activity. Accordingly, three epistemological premises are adopted: (1) a knowledge-engineering premise, which states that the knowledge requirements for natural language report generation are domain-dependent and that production systems provide an idea formalism for representing both semantic and linguistic knowledge; (2) a macro-level premise, which states that macro-level knowledge structures and processes, such as whole messages, a phrasal lexicon, and a clause-combining grammar, provide an effective and efficient level for representing and processing knowledge; and (3) an integrated-knowledge context-dependent grammar principle, which states that the generation of fluent natural language reports requires a grammar in which a variety in types of knowledge are integrated into highly context-dependent rules.

A prototype system named Ana has been implemented for the domain of stock market reports. It is currently able to input data from the Dow Jones News Service database and output three-paragraph summary reports of the day's market activity.

A critical analysis of the system's architecture and function focusses on two questions: does it work? and is it practical? To answer the workability question, samples of the system's output are analyzed for accuracy, fluency and completeness. General findings are that the reports are highly accurate and fluent and potentially complete. To answer the practicality question, the extendibility, tailorability, portability , and cost-effectiveness of the system is studied. General findings are that extendibility and tailorability are major strengths of the technique, that portability remains an open question until a research project is undertaken to transport a knowledge-based report generator to another domain, and that storage and processing costs are very low but implementation costs, mainly, the time and services of a skilled knowledge engineer, are very high.

MPACT Scores for Karen Kukich

A = 0
C = 0
A+C = 0
T = 0
G = 0
W = 0
TD = 0
TA = 0
calculated 2008-01-31 06:22:25

Advisors and Advisees Graph

generating graph, please reload

Students under Karen Kukich

ADVISEES:
- None

COMMITTEESHIPS:
- None