
You’ve got a big family dinner coming up. How can you be sure that the restaurant you’ve picked out is the perfect place to host it? Why, hop on Zomato of course! There are loads of reviews there to help confirm your impeccable taste in dining establishments. But there are so many things to consider – is the lighting perfect? Which dish is unmissable? Will they refill your water glass often enough?
These struggles are real, and we’ve all faced them at one time or another. We realised that all the information in reviews needed to be packaged and presented in an easier way, to help our users make better and faster choices. And this is where our new Review Highlights feature comes in.
What we’re doing
In a nutshell, making your life a whole lot easier – we’re crystallising all the opinions in a restaurant’s reviews into a single section. One glance will now tell you exactly what people loved there, and what to look out for when you visit, be it stunning rooftop seating or a risotto that’ll knock your socks off. Aggregated ratings (that we like to call “sentiment scores”) for food, service, and ambience, will also give you a better idea of how diners feel about each of these decision criteria.

How we’re doing this
By combining natural language processing, sentiment analysis, and a bit of engineering wizardry. The first step was identifying every relevant verb and noun that we could use, by analysing every review on the platform. Once we had the starting point for our (ever increasing) word bank, it was a small matter of working our magic under the hood, and we were good to go. Allow us to illustrate.
Let’s use pizza as an example – because why use anything else? Assume that about 50 people who ate at a particular restaurant had only nice things to say about their pizza (and that the word “bae” was used a LOT). And maybe 4 others didn’t particularly care for how it tasted. We weigh the positive mentions against the negative ones, while also factoring in how often each word was mentioned in the reviews, and boom! “Pizza” at that restaurant now has a score.

We did this for every keyword, with the highest scoring ones for each restaurant taking pride of place on their Zomato pages. But wait, what about the aggregated sentiment score we spoke about earlier? Pretty much the same as above, but only for all food related words, rather than specific dishes. Rinse and repeat for service and ambience, and the review highlights are good to go.
Sometimes, the ambience or service at a restaurant may not be spoken about enough for it to have a score. So if you happen to see missing categories on the page, don’t worry – they’ll join the party once people start talking about them. The obvious exception here would be places that only offer takeaway and delivery. The section will be dedicated to highlighting what you need to know about their food and service, as the ambience isn’t relevant.

Every year, diners worldwide collectively waste about 15,000 hours trying to figure out what to eat [citation needed]. That lost time is never coming back, and you deserve to be spending it with your loved ones, or doing whatever it is that makes your life awesome. Besides, faster decision-making means you can be eating sooner, and who doesn’t want that?
This feature is currently live in India, Australia, New Zealand, UAE, and Malaysia, on iOS and mobile web. But we’ll be spreading the joy to all our markets, as well as other platforms, very soon. Watch this space, and until then, happy eating!