[septimus111]

This is why there is such “grade inflation” in reviews. The way I read reviews is now: - look at the reviews with lowest ratings. They have the most signal (the walls were paper thin and the bathroom was dirty) as opposed to positive reviews (beautiful room in a great location). You can usually tell if the reviewer is being unreasonable. - search for specific negative keywords in the other reviews - “noisy”, “loud”, etc. Often people will leave a five star review but discreetly complain about something.

[scarab92]

Be mindful of confirmation bias when searching reviews. Finding 10 reviews mentioning noise might be because it's noisy, or it might be that they have thousands of reviews and a small percentage experienced noise

[septimus111]

This is an excellent point! I suppose I could normalize by the number of total reviews. It hasn’t been an issue so far as I am only filtering out cases when there are multiple reports of the same issue, and I care more about eliminating bad options than detecting all good options.