Goal
Make it easier for people to understand what is good or bad in reviews by summarising content with the help of machine learning
Problem:
It’s very difficult for people to understand the positive and negative aspects of a property at a glance. A property may have a very high overall rating, but certain aspects mentioned in reviews (like location or breakfast) may be negative. People need to read through many reviews to form their own opinion about it.
Process:
The first step of this project was carried on in collaboration with my data scientist and product manager. We utilised machine learning to annotate thousands of real guest reviews for different categories (location, breakfast, etc.) and their most commonly talked about topics. This allowed us to know whether a particular category is considered good or bad for our guests and what is the reason behind it. Starting from the data analyzed, my role was to generate sentences for each topic, both in positive and negative sentiment.
Solution:
I create a review summary. The biggest challenge for me was to write phrases that were useful and insightful but at the same time not repetitive and broad enough to include all the feedback written by guests for that particular topic. I read hundreds of real reviews in order to find the right words and use a natural and fluent language. Another challenge was translating these sentences into different languages: I’ve created a clear brief with all the reference material to use the best terminology for every language and culture.
Impact:
The A/B experiment we ran showed that the summary resulted in more people booking higher-rated properties. Behavioural metrics also showed that the properties people were booking better fit their needs.
Company: Booking.com
Team: User Generated Content