Startup Luka has developed an app that uses artificial intelligence (AI) to make restaurant recommendations based on short instant messaging conversations.

Luka’s restaurant recommendations make use of AI

There are so many startups operating in the restaurant sector these days that the generic name ‘foodtech’ has been coined to describe the range of services they provide. Some of these firms focus on organising express home delivery and there are a number of foodtech startups – Postmates, Sprig, Munchery being perhaps the best known – that pride themselves on arranging delivery services in record time without comprising on meal quality. The other major area of foodtech is all about providing recommendations for restaurants on the basis of customer opinion and location. The best known service here is Yelp, which offers an app to help you find bars, restaurants and other food and drink establishments in the vicinity and enables customers to grade them on quality. Founded in 2004 in San Francisco, Yelp has become the number one platform for searches on local businesses and also for publishing crowd-sourced reviews on them. One consequence has been that many restaurants now try hard to improve their Yelp grading in order to boost their appeal to customers.

It is a fact that nowadays customer reviews are a key factor among online users in choosing a place to eat. However the Yelp service is essentially browsing. There is no conversation and the app does not take your likes/dislikes or location into account, so the advice it offers is not at all customised. By contrast, fellow San Francisco startup Luka offers a recommendation service based on real-time conversation and makes use of AI to ensure more personalised advice.

Conversation-based recommendations

Luka, formerly called IO, processes your request and sends back recommendations based not only on your request as framed but also taking into account your recorded history and known preferences. You can obtain tailored recommendations produced by the algorithm developed by Luka on some 2,000 restaurants currently registered in the app, although for the moment the service is only available in the San Francisco Bay area. The interface displays a conversational exchange that looks a lot like an SMS chat with a friend.

You can phrase your question to the app in any way you like, such as: “Is there a good restaurant you could recommend for this evening?” Or ‟Which restaurant would you suggest for a date tonight?” Luka will send you an immediate recommendation for restaurants in the neighbourhood, based on price, menu, Yelp recommendations, Foursquare check-ins, photos, and so on. Meanwhile the algorithm is learning all the time. Luka takes on board your likes/dislikes – whether you are a strict vegetarian or passionate meat-lover – thus saving you the hassle of browsing the Web yourself to match your needs for an eatery. Having received Luka’s initial recommendations you can then whittle down the range of choice by asking more precise, detailed questions. All-in-all Luka’s ambition is to be the ‘foodie friend’ whose advice we would all wish to obtain when looking for the right restaurant.

By Arthur de Villemandy