CDR Project

Dedal AI:  Dedal with Automated Indexing

DedalAI is a search engine that indexes and retrieves design information conceptually rather than by word frequency, as many popular search engines do. Search and retrieval of design experience is an important component of the design process -- “all design is redesign.” The conceptual DedalAI approach has been shown to increase the recall* of information up to 45% over simple Boolean keyword search.

DedalAI is based on the Dedal system and works by searching for two parts of a query: a descriptor, which is a generic design concept, and a subject, which is the specific part of a design model. A list of descriptors and an example model is shown below.

In the original Dedal implementation, descriptors and subject were indexed by hand, a time consuming and knowledge intensive process. DedalAI increases the usability of Dedal by automatically indexing design texts such as PENS notebooks. It accomplishes this by iincorporating a descriptor thesaurus and a subject framework. The descriptor thesaurus consists of words related to each of the descriptors, and the subject frame includes parts of a device model. For example, if a query is <Alternative> of mirror, DedalAI would search for instances of where both <Alternative>, or one of its descriptor synonyms, and mirror, or a subject in its frame, such as CCD Camera, appeared.

This use of a descriptor thesaurus with subject frames increases the recall of information retrieval significantly without any loss of precision. Future work involves further exploration in indexing of subject frames to maximize the precision* as well.

*  Recall and precision are standard measures of information retrieval performance. Recall indicates how much good information is retrieved, while precision shows how much retrieved information is good.

Publications

Maria C. Yang and Mark R. Cutkosky. Automated Indexing of Design Concepts for Information Management. International Conference in Engineering Design, Tampere, Finland, August 1997 (Accepted for presentation and publication in proceedings)

Maria C. Yang and Mark R. Cutkosky. Automated Indexing of Device Models for Information Management. 1997 Design Engineering Technical Conferences 9th International Conference on Design Theory and Methodology. Sacramento, California, September 1997. American Society of Mechanical Engineers. (Submitted, working copy available for download)