PUblicationMAnagement  PumaLogo 
TitleExperience Knowledge Mechanisms and Representation

Author Thönssen B., Witschel H. F., Hinkelmann K., Martin A.

Deliverable D5.3 focuses on creating a knowledge base repository consisting of cases and a similarity based semantic case retrieval mechanism. We studied thoroughly the state-of-the-art of experience knowledge mechanism and representation. We did in depth domain analysis with Marche Region to identify requirements for experience management in Public Administrations, particularly in the Titolo Unico process. This reconfirmed that Case Based Reasoning (CBR) is an appropriate approach for experience management as it reflects extremely well the way individuals use former, i.e. existing experience knowledge to solve problems at hand. Together with Marche Region we determined case descriptions and case content. In order to generalize the approach we enhanced case metadata by a standard of the National Insititute for Statistics (ISTAT). For implementation we built upon research on ontology based CBR performed in a Swiss national project. For the Learn PAd project the core CBR component of the FHNW ICEBERG Toolkit was reused and adapted to meet the specific requirements and to fit into the Learn PAd platform, i.e. into the ontology and recommender component. The Learn PAd ontology was enhanced by CBR concepts and similarity functions. For evaluation we inserted 12 former cases in the case base, formalized as instances in the Learn PAd ontology. Representatives of Marche Region created a new (fictitious) case and determined manually the three most similar cases from the case base. The similar cases suggested by the CBR were then compared to them. Based on the result we improved the similarity measures (e.g. weights of attributes) and did a second run. With the newly derived weight vector the ranking that the expert expected / recommended was achieved. This confirms the suitability of our similarity model and similarity functions but to avoid the risk of overturning it needs subsequent plausibility check with the expert. After this early evaluation, focussing on the achieved quality of recommended cases further evaluations will be done through simulation and the comprehensive demonstrator assessment.

Subject Case Based Reasoning
CBR cycle
Similarity-based Semantic Case Retrieval Mechanism
Ontology-based CBR
Similarity Models
Similarity Functions
Similar Cases
Adaptation Models
Characterisation Space
Characterisation Elements
CBR Configuration and Data Import

DownloadIcona documento      2016-EC-005.pdf

Icona documento Open access Icona documento Restricted Icona documento No access