What shoes go with that dress? Or this pair of jeans? Artificial intelligence is now answering those questions automatically for online shoppers in China, thanks to an algorithm developed by web giant Alibaba.
The system recommends entire personalised outfits to users as they browse, mixing ensembles from recently viewed items and other items judged to coordinate well with them. A live trial of the tool has already recommended outfits to more than 5 million users.
The aim is to enhance the experience for customers by encouraging more fashionable shopping behaviour, says Wen Chen at Alibaba.
Advertisement
To build the system, Chen and her colleagues assembled a dataset of more than 1 million sample outfits created manually by staff at Alibaba鈥檚 online shopping site Taobao. An algorithm used this dataset to learn what clothing items are compatible with one another.
Once knowledge of what goes with what had been gained, the outfit generation tool was able to select compatible clothing for items recently clicked on by users. Endless combinations can be continuously created as the customer shops.
In a live test on Taobao鈥檚 iFashion designer clothes app, the suggested outfits were clicked on about 25 per cent of the time, compared to a 15 per cent click through rate from other approaches.
The system can also build up a sense of customers鈥� fashion profiles based on their clicks, which could potentially be used for targeting customers with other items.
Online shoppers are often faced with a 鈥渢yranny of choice鈥� that can be overwhelming, says retail adviser Doug Stephens. Offering suggestions of outfits that genuinely look good would therefore be welcome.
鈥淚t鈥檚 a great problem to solve, given that one of the consumer鈥檚 most significant apparel challenges is the coordination of items 鈥� particularly online where there鈥檚 an absence of human assistance,鈥� he says.
Reference: arXiv,
