A Messenger chatbot well-versed with perfumes

Now that the most popular social platforms (Facebook Messenger, Skype, Whatsapp, Telegram, WeChat...) have opened their doors to them, conversational robots (chatter robots, or chatbots), able to exchange with a human on a precise matter, are gradually becoming more familiar to the largest possible number of people. Main advantage: they are instantly accessible, from the messaging that most owners of smartphones use every day, without it being necessary to download any additional application.

Sommelier du Parfum belongs to this new generation of tools much appreciated by millennials. Its objective: guide its users in the choice of a perfume, for oneself or to offer. "It is actually the first question asked by the robot," explains Samuel Fillon, its creator. The chatbot then uses several other criteria to define the "olfactory profile" of the person for whom the perfume is intended: man or woman, other favourite perfumes, types of smells that we like (citrus, woody, etc.).

"The vocabulary used by the chatbot is very accessible, jargon-free. Everyone understands what a citrus note is, while it is not the case when talking about hesperidia notes," highlights Samuel Fillon, who started this project during his studies in Machine Learning at the University of Tokyo, as part of his training as a civil engineer (Ponts et Chaussées school). "Before my departure for Japan, I found myself a bit lost when wanting to offer a perfume to my brother for his birthday. I knew he liked the smells of some of the shower gels he used, but that was of little help in the shop in front of the salesperson for whom my inquiry was obviously too vague."

In fact, the chatbot uses a database of more than 30,000 references of crowdsourced perfumes retrieved from data available on the web. Each of these 30,000 references is indexed according to its female/male positioning, its main olfactory notes, its retail price, its appreciation by perfume critics, its creator, etc. Knowledge definitely out of reach for ordinary human beings! Not to mention that, in the course of its conversation, the chatbot learns to better know the tastes of his interlocutor.

Once the olfactory profile and the tastes of the recipient of the perfume are defined, the chatbot proposes a series of recommendations: "for next summer, winter, for work or in the evening." Detailed information is available for each proposal, and a store locator enables to spot the closest location to test the suggested products.

Perfume experts will probably feel confused by the robot’s proposals, but unquestionably, it brings an element of fun to the relation with the customer, and yet without breaking the human relationship because it facilitates access to the perfume market. A tool, which will appeal to both retailers and brands. It can be adapted to the specific needs of different players and also gives access to many consumption data.