Scraping and Analysing Google Maps Reviews

Even on a premium account, the Google Places API is limited to delivering only the top five reviews of any location.

I was interested in more in depth analysis of the reviews, to assess factors which would cause someone to give a positive or negative review/high or low star rating. This can aid businesses and government agencies to find out what patrons love or hate about their venue/location. Results can be applied in marketing and making the location more enjoyable for visitors.

This was achieved by webscraping all the reviews from any target location using Selenium (slower than using an API as it needs to scroll down and wait for contents to load). The reviews were further evaluated using natural language processing and common themes were found.

For example, themes relating to waiting times, quality of service, and food quality were common for cafes and restaurants and mentioned in both positive and negative reviews. However, cleanliness(or lack thereof) tends to only be mentioned in negative or 1 star reviews.

For public places, common themes were air conditioning, presence of helpful information booths and staff, and general comment on architecture. Reviewers mentioned construction works and waiting times frequently in a negative light.