Searching for design examples with crowdsourcing

Nikita Spirin, Motahhare Eslami, Jie Ding, Pooja Jain, Brian P. Bailey, Karrie Karahalios

Abstract: Examples are very important in design, but existing tools for design example search still do not cover many cases. For instance, long tail queries containing subtle and subjective design concepts, like "calm and quiet", "elegant", "dark background with a hint of color to make it less boring", are poorly supported. This is mainly due to the inherent complexity of the task, which so far has been tackled only algorithmically using general image search techniques. We propose a powerful new approach based on crowdsourcing, which complements existing algorithmic approaches and addresses their shortcomings. Out of many explored crowdsourcing configurations we found that (1) a design need should be represented via several query images and (2) AMT crowd workers should assess a query-specific relevance of a candidate example from a pre-built design collection. To test the utility of our approach, we compared it with Google Images in a query-by-example mode. Based on feedback from expert designers, the crowd selects more relevant design examples.