Re:Show HN: 13M opinions on 48k products extracted from Redd...
by ivihernandez on Tuesday November 03, 2026 @08:05AM
(#50351516)
Dang, thats a lot of data!! Good job. Tough to organize so much data across so many categories. Not sure how you were able to create different pages for different things (e.g. fantasy books and travel)
1) the first thing you see on your UI for a city is the nearby attractions, instead of things to do in the city itself (i.e. Nearby ranked)
2) the second thing one sees is the confidence, e.g. medium confidence for Bucharest. But I am not sure what "medium confidence" means. I infer it is the mean of the sentiment, but might be good to complement it with the distribution of sentiment over, say the last year. And perhaps use a name other than "confidence". You have something like this further below, but it does not specify the time range.
3) for the sentiment over time, I think this is a great great idea, because ratings can shift quickly and I don't necessarily want 5 year old things to have an impact right now. Having said that, I recommend to add label to all axis and in this case, perhaps just draw the time series of the median sentiment. I find the graph too complex with many colors, time series and bars at the same time.
I looked specifically at the travel category. A couple of thoughts (https://www.trustmebro.reviews/p/bucharest-a8a4880c):
1) the first thing you see on your UI for a city is the nearby attractions, instead of things to do in the city itself (i.e. Nearby ranked)
2) the second thing one sees is the confidence, e.g. medium confidence for Bucharest. But I am not sure what "medium confidence" means. I infer it is the mean of the sentiment, but might be good to complement it with the distribution of sentiment over, say the last year. And perhaps use a name other than "confidence". You have something like this further below, but it does not specify the time range.
3) for the sentiment over time, I think this is a great great idea, because ratings can shift quickly and I don't necessarily want 5 year old things to have an impact right now. Having said that, I recommend to add label to all axis and in this case, perhaps just draw the time series of the median sentiment. I find the graph too complex with many colors, time series and bars at the same time.