Hrg. Christina Feilmayr,
"Decision Guidance for Optimizing Information Quality - A Recommendation Model for Completing Information Extraction Results"
Decision Guidance for Optimizing Information Quality - A Recommendation Model for Completing Information Extraction Results
Sprache des Titels:
The information quality in an (intelligent) information system frequently suffers because of erroneous information acquisition from full natural language text documents, which is realized by applying information extraction methods. Notably, incomplete information has serious consequences, because early impreciseness in extraction is propagated to later extraction phases, which consequently leads to erroneous annotations, inaccurate predictions in information analysis and finally to incorrect decisions. Hence, somebody who is working on this information reaches possibly wrong decisions on incomplete and untrustworthy information.
Building an information extraction system is a complex and knowledge-intensive task, in which the complexity of the design process depends on both domain and scenario. In principle, there is no support in designing or refining an existing information extraction system or in selecting appropriate information extraction methods, especially when a specific information quality problem such as incompleteness has to be addressed. Accordingly, there is no existing approach that copes with reduction/elimination of information quality issues, which specifically addresses the characteristics of extraction errors, and which provides recommending suggestions for an automatic refinement.
This research work focuses on improving the completeness of extraction results by applying judiciously selected assessment methods (methods from data mining and linguistics) to information extraction within the principle of complementarity. A system designer is assisted in the reassessing by a recommendation model, which suggests suitable methods for improving completeness in response to a dominant incompleteness problem. In addition, the focus is on the various requirements an assessment method must meet in terms of processability and profitability to guarantee effective operation in a complementarity approach.