Show simple item record

dc.contributor.advisorAamodt, Agnarnb_NO
dc.contributor.advisorGundersen, Odd-Eriknb_NO
dc.contributor.authorKokkersvold, Idanb_NO
dc.date.accessioned2014-12-19T13:31:05Z
dc.date.available2014-12-19T13:31:05Z
dc.date.created2010-09-02nb_NO
dc.date.issued2007nb_NO
dc.identifier346877nb_NO
dc.identifierntnudaim:3450nb_NO
dc.identifier.urihttp://hdl.handle.net/11250/250222
dc.description.abstractConversational Cased-Based Reasoning (CCBR) is an interactive, dialogoriented method within Cased-Based Reasoning (CBR) that provides a mixed-initative dialog for guiding users to construct their problem description incrementally through a question-answering sequence. A query case is created as the problem description and will often contain one or several explicit initial features provided by the user. The query case is used to retrieve a set of candidate cases, and a group of informative features are identified to generate discriminative questions. In order to design and implement a CCBR question answering system, it is important to find an appropriate question selection strategy and conversation process. Our study includes an overview of different methods to be used in a question selection strategy, and some existing application of CCBR. Which question selection strategy to apply in a CCBR system depends of the application's domain. Factors such as purpose and real life conversation style will influence the design of a CCBR question answering system. Based on our research, we designed and partly implemtend a CCBR question selection system called CCBR Diagnose. The CCBR Diagnose is planned to be a CCBR tool to use within pallatative care, and is based on the CBR Core System developed by Odd Erik Gundersen. The CCBR Diagnose is designed and implemented as a knowledge-poor CCBR application, but should in future work be extended to be knowledge-intensive. Information gain is implemented for selection of a feature to be ask about, and has improved the system effiency in form of shorter conversation length. The system uses information from the knowledge model to make sure that the questions is comprehensible. By including knowledge-intensive methods in the question selection strategy, the system will improve performance by reducing the cognitive load on the patient, and is highly recommended for the final tool. A knowledge-intensive selection strategy can improve the conversation by avoid asking implictly answered questions and presenting a more coherent sequence of questions. Additional, the system is missing a graphical user interface. Such an interface can be based on the already developed computerzed version of the paper-based questionarries, but should also follow our guidelines. Clearly more work lies ahead in order to achieve the final CCBR system for pallatative care, but future work can be based on our research contributions and our implemented CCBR Diagnose can be us a basis for further development of a final tool within pallatative care. We have proven that CCBR improves the conversation in form of shorther conversation length, and we have suggested how additional improvements can be achieved by supplying the implemented question selection method with knowledge-intensive methods. Which knowledge-intensive strategy to apply is up to future work to decide.nb_NO
dc.languageengnb_NO
dc.publisherInstitutt for datateknikk og informasjonsvitenskapnb_NO
dc.subjectntnudaimno_NO
dc.subjectSIF2 datateknikkno_NO
dc.subjectIntelligente systemerno_NO
dc.titleConversational CBR for an Adaptive Question Answering Systemnb_NO
dc.typeMaster thesisnb_NO
dc.source.pagenumber129nb_NO
dc.contributor.departmentNorges teknisk-naturvitenskapelige universitet, Fakultet for informasjonsteknologi, matematikk og elektroteknikk, Institutt for datateknikk og informasjonsvitenskapnb_NO


Files in this item

Thumbnail
Thumbnail
Thumbnail

This item appears in the following Collection(s)

Show simple item record