Last updated: June 26, 2023



Location: Portugal

Farfetch Chat R&D Demonstrations

At its conclusion, FARFETCH Chat R&D pertains several features instrumental to a multimodal conversational recommender system. The videos shown below showcase the system as is.

Support for beauty products

FARFETCH Chat R&D aims to support the user journey as if they were physically visiting a store, only with access to the FARFETCH product catalogue. The FARFETCH product catalogue recently included beauty products, ranging from perfumes, to tanning and hydration serums. In these videos, we showcase the AI agent ability to retrieve meaningful beauty products, given the user’s intents.

Product disambiguation

We are aware of how natural conversation is when purchasing fashionable apparel. There may be a few inquiries pertaining to a variety of aspects of the clothing under consideration or regarding the store visit. It only makes sense the chatbot employs counter measures to guide its users towards the products they are interested in. This video shows an example where disambiguation is triggered in the presence of a product information specific question.

Product information for measurements and materials

One of the more general features driving this conversational recommender is the capability to answer questions regarding a product’s attributes. This video presents such user questions, regarding a product’s material composition and measurements for a given size.

Product search with attribute manipulation

Of course, the conversational agent wouldn’t be a recommender system without recommendations. It is capable of performing product search on the FARFETCH catalogue and parse user requests as they further refine the search process. Finding the best suited products for each and every customer demand is a top priority in FARFETCH Chat R&D.

From the beginning, FARFETCH Chat R&D focused on supporting multimodal conversations. When interacting with the chatbot, users are able to perform product search not only via text, but also image! The video below is an example of a visual search process.

Multimodal search with attribute manipulation

As we are able to search products via free text and custom uploaded images, we can also provide an image and add textual constraints to start the search process with inpromptu refinements.