Thanks a lot Kartik for you appreciation and for taking the time to post a response!
I looked a bit into what you’re suggesting.
It is very hard to do that because:
- You have to set a threshold for the number of users that consider a review useful.
- For less popular movies, this may be a limiting factor — you will have to computer the mean for a small number of ratings, which is dangerous because it might not render a meaningful average.
- You need to collect many ratings for each individual movie, and then average them. This is something very tedious to do.
Maybe an interesting angle would be to determine to what degree the IMDB rating matches the average rating obtained only from the most useful reviews.
For this article/project, I cannot pursue this angle because I analyzed only those ratings that are iconic for each website and easy to access. And the iconic IMDB rating takes into account ratings from all users, and doesn’t do any weighting, as far as I know.