10.17192/z2007.0481
Mustafaraj, Eniana
Eniana
Mustafaraj
Knowledge Extraction and Summarization for Textual Case-Based Reasoning: A Probabilistic Task Content Modeling Approach
Wissenextraktion und Zussamenfassung für Textuelles Fallbasiertes Schliessen: Ein probabilistischer Ansatz für inhaltsbasierte Modelierung
Philipps-Universität Marburg
2007
Natural language processing
Fallbasiertes Schliessen
Case-based reasoning
Data processing Computer science
Informatik
Freisleben, Bernd (Prof. Dr.)
Künstliche Intelligenz
2013-11-05
2007-08-01
en
DoctoralThesis
https://archiv.ub.uni-marburg.de/diss/z2007/0481
urn:nbn:de:hebis:04-z2007-04813
https://archiv.ub.uni-marburg.de/diss/z2007/0481/cover.png
application/pdf
https://rightsstatements.org/vocab/InC-NC/1.0/
Case-Based Reasoning (CBR) is an Artificial Intelligence (AI) technique that
has been successfully used for building knowledge systems for tasks/domains where different knowledge sources are easily available, particularly in the form of problem solving situations, known as cases. Cases generally display a clear
distinction between different components of problem solving, for instance, components of the problem description and of the problem solution. Thus, an existing and explicit structure of cases is presumed. However, when problem solving experiences are stored in the form of textual narratives (in natural language), there is no explicit case structure, so that CBR cannot be applied directly.
This thesis presents a novel approach for authoring cases from episodic textual
narratives and organizing these cases in a case base structure that permits a
better support for user goals. The approach is based on the following fundamental ideas:
- CBR as a problem solving technique is goal-oriented and goals are realized by
means of task strategies.
- Tasks have an internal structure that can be represented in terms of
participating events and event components.
- Episodic textual narratives are not random containers of domain concept
terms. Rather, the text can be considered as generated by the underlying
task structure whose content they describe.
The presented case base authoring process combines task knowledge with Natural
Language Processing (NLP) techniques to perform the needed knowledge extraction
and summarization.