Task-oriented and similarity-based retrieval

Abstract

We present two support facilities for identifying design information in a repository. First, task-orientation supports the dynamic composition of repository information with respect to a designer's working context. Second, similarity-based retrieval overcomes the impediment of exact matching. When a requested repository item is not available, it supports the search for closely related items by means of matching criteria and relationships. The algorithm uses domain taxonomies to measure the virtual distance between items, rather than applying artificial numerical measures. Moreover, the criteria for calling repository items similar are determined and formulated by the user in the query, rather than being hard-wired into the retrieval interface. The combination of both features furnishes a visibility concept, that allows designers to identify design information for their specific job at hand.<>