Facilitating Design-by-Analogy: Development of a Complete Functional Vocabulary and Functional Vector Approach to Analogical Search

Abstract

Design-by-analogy is an effective approach to innovative concept generation, but can be elusive at times due to the fact that few methods and tools exist to assist designers in systematically seeking and identifying analogies from general data sources, databases, or repositories, such as patent databases. A new method for extracting analogies from data sources has been developed to provide this capability. Building on past research, we utilize a functional vector space model to quantify analogous similarity between a design problem and the data source of potential analogies. We quantitatively evaluate the functional similarity between represented design problems and, in this case, patent descriptions of products. We develop a complete functional vocabulary to map the patent database to applicable functionally critical terms, using document parsing algorithms to reduce text descriptions of the data sources down to the key functions, and applying Zipf's law on word count order reduction to reduce the words within the documents. The reduction of a document (in this case a patent) into functional analogous words enables the matching to novel ideas that are functionally similar, which can be customized in various ways. This approach thereby provides relevant sources of design-by-analogy inspiration. Although our implementation of the technique focuses on functional descriptions of patents and the mapping of these functions to those of the design problem, resulting in a set of analogies, we believe that this technique is applicable to other analogy data sources as well. As a verification of the approach, an original design problem for an automated window washer illustrates the distance range of analogical solutions that can be extracted, extending from very near-field, literal solutions to far-field cross-domain analogies. Finally, a comparison with a current patent search tool is performed to draw a contrast to the status quo and evaluate the effectiveness of this work.

A Novel Application of Gamification for Collecting High-Level Design Information

“…In terms of the game content generation, we could explore applying techniques from natural language processing to replace or augment the crowdsourcing approach. For instance, we could apply Latent Semantic Analysis [43] or other vector semantics methods developed for design knowledge retrieval, e.g., [44], to compute functional similarity between a pair of objects. We would need to refine these methods and validate them against the human assessment of functional similarity.…”

Design-by-analogy: experimental evaluation of a functional analogy search methodology for concept generation improvement

“…Goldfire by Invention Machine, a commercially available patent exploration tool, focuses on the goal of understanding competitive landscapes and new areas for technological development, as opposed to facilitating design-by-analogy through function-based cross-domain transfer of knowledge (Goldfire 2012). The main difference between these aforementioned research efforts and our work is our employment of Zipf's law and the functional basis to achieve a similarity and relevance evaluation that relies on a static, converged, computationally derived functional vocabulary for the patent database (Murphy et al 2014). This approach allows for clearer disambiguation to distinguish the different meanings of polysemes.…”

“…Finally, the most relevant patent results are presented to the user. Further details of the methodology can be found in (Murphy 2011;Murphy et al 2014).…”

“…The functional vector space model-based (VSM) analogy search engine was developed in prior work by the authors (Murphy 2011;Murphy et al 2014). The process begins with constructing a controlled and converged vocabulary of functions extracted from a patent database, building on the hierarchical structure of a functional basis, as inspired by the functional modeling approach of Pahl and Beitz (Hirtz et al 2002;Otto and Wood 2001;Stone and Wood 2000;Pahl and Beitz 1996), and five categorical functions of which all functions are more specific instances.…”