Accelerating Innovation Through Analogy Mining

Abstract

The availability of large idea repositories (e.g., patents) could significantly accelerate innovation and discovery by providing people inspiration from solutions to analogous problems. However, finding useful analogies in these large, messy, real-world repositories remains a persistent challenge for both humans and computers. Previous approaches include costly hand-created databases that do not scale, or machine-learning similarity metrics that struggle to account for structural similarity, which is central to analogy. In this paper we explore the viability and value of learning simple structural representations. Our approach combines crowdsourcing and recurrent neural networks to extract purpose and mechanism vector representations from product descriptions. We demonstrate that these learned vectors allow us to find analogies with higher precision and recall than traditional methods. In an ideation experiment, analogies retrieved by our models significantly increased people's likelihood of generating creative ideas.

Analogy Mining for Specific Design Needs

“…Recent work by Hope et al [15] proposes to find analogies among free-text consumer product descriptions by learning to predict an overall representation of a product's purpose (what it is good for) and mechanism (how it works). It uses annotators to mark words related to the purpose/mechanism of the product, and weighs the Glove [23] values of those words to assign an overall purpose/mechanism representation for each document.…”

“…We use a corpus of product descriptions from [15]. The products in the corpus are from Quirky.com, an online crowdsourced product innovation website.…”

“…and mechanism (how does it work?). Given the vector representations, the algorithm of [15] finds products with similar purpose but different mechanisms, that might serve as analogies.…”

“…One specific finding that deserves further discussion is the high performance of the OverallGlove condition in terms of relevance. Our initial prediction was that this condition would perform poorly on relevance, since, like the OverallPurpMech method from [15], it operates on the whole product description as opposed to a specific focus query. Cognitive theories of analogy suggest one possible explanation.…”

Challenges in Finding Metaphorical Connections

“…Beyond poetry, helping people find connections between two domains has far-reaching applications from science education (Glynn, 1991) to product design (Hope et al, 2017). This is a hallmark of human intelligence that can be computationally supported.…”