Browsing through Swiggy’s vast catalogue of restaurant options can be overwhelming, especially when you are absolutely famished. Sometimes you know the type of food you want to have – sweet, savoury, or spicy – but can’t find the dish that fits your cravings perfectly. With a new feature called neural search, the food delivery app aims to make food hunting an effortless process. With neural search, users will be able to type out conversational-style prompts and “open-ended queries” and receive highly specific recommendations accordingly, making it possible to find food without having to remember specific keywords or dish names. For instance, typing in queries like, “I just finished my workout. Show me healthy lunch options” or “Show me vegan-friendly starters” will fetch personalised suggestions. In a video demonstration, Swiggy shows the neural search feature plugged into the existing search feature on the app’s home screen. Normally, searching for “ice creams for diabetic patients” returns random ice cream outlets since Swiggy only matches your keywords against the restaurants and dishes in its database. But with the power of generative AI, Swiggy is able to ‘understand’ that query and suggest sugar-free ice cream. Swiggy says that the feature is powered by a neural search capability that uses a Large Language Model (LLM) to understand food-related terminology and queries. The LLM has been trained on Swiggy-specific search data and can handle more than 50 million items in its food catalogue. The company claims that the neural search capability can provide accurate and relevant results in real-time and that it has been developed in-house to enable faster innovation and adaptation. Going by the demo video, neural search has seemingly entered the advanced stages of testing already. Swiggy says it will be in pilot by September, with a wider rollout to follow in an unspecified time frame. The feature will also be able to help shop for groceries on Instamart. Support for regional languages is in the cards too. But that’s not all. Swiggy is experimenting with generative AI applications within its restaurant discovery Dineout feature. The platform is also collaborating with a third party to develop a GPT-4 powered chatbot to answer customer queries. And finally, an LLM-based conversational assistant for restaurant partners to self-serve on various processes is in the works.
