Thinking With Containers: A Multi-Agent Retrieval Approach for the Case-Based Semantic Search of Architectural Designs

To provide the retrieval of information, that can be considered useful during the design conceptualization process, with advantages of distributed artificial knowledge, an approach, that distributes retrieval-related and knowledge maintaining tasks among autonomously working and case-based self-learning agents and agent groups, can be used. In this work we present the distributed retrieval system MetisCBR for the architectural design domain, where agents work in groups (containers) on resolving of user queries built with a semantic description model Semantic Fingerprint. The main aim of our approach is to carry out a basis for a considerable retrieval tool for architects, where the combination of case-based reasoning and multi-agent methods helps to achieve valuable and helpful search results in a comprehensive building design collection.