Show HN: Using semantic tagging to improve hotel reviews

repustate.com

7 points·waterside81··2 comments

Wow, it looks good. I might use this next time I'm looking for a hotel. I would love this as an addon in Chrome, so I can use it while browsing hotels. Unfortunately, most of my trips lately have been to Russia, and you don't cover those.

One nit pick, the 'staff' filter seems to cut sentences in half, usually ending in the word 'staff'. The actual comments about the staff seem to follow the word 'staff' in the reviews, thus it failed on the few reviews that I looked at. Still, there were over 400 reviews for the particular hotel that I looked at, so it may have worked better in other reviews.

I'd also be curious if it gave more weight to more recent reviews. Staff can change quite a bit in two years.

Yeah, this is just a proof-of-concept that our intern made. It can definitely be extended to cover any & all hotel reviews that are publicly accessible.

As for the staff issue - the classification goes by semantic chunk, rather than by sentence, since sentences can cover different semantic topics e.g. "The staff were great, but the beds were too hard." But I agree, sometimes it looks a bit ... off.

Your weighting idea is also a good one too. Maybe introduce some sort of decay factor and weight the scores accordingly. In fact, that might be good feedback for the hotel to see how their ratings have changed over time.

Thanks for the feedback.

Reply on news.ycombinator.com