What would be some interesting uses of analyzing data from HN favorites?

Question

Every HN user's list of favorites is public. Given that it's a list of topics that people find interesting, I imagine some people, especially advertisers, would find it useful.That's an obvious one though. What kind of wacky ideas have you had for this kind of data?

armchairhacker · Accepted Answer

A lot of people don&rsquo;t use favorites. I don&rsquo;t, and out of the few random accounts i&rsquo;ve looked at, none of them had any favorites either. Doesn&rsquo;t invalidate the data but it does skew towards people who do have favorites.Another source of data is articles where people comment. You could also do ML to analyze whether the comments are positive or negative, guess the commenter&rsquo;s age and demographic, etc.If someone comments on a particular topic, especially if it&rsquo;s not as popular, it shows they&rsquo;re interested in the topic. Maybe look at the writing styles and content of people interested in certain topics. Are people interested in X topic more likely to be negative? etc

toast0 · Answer

I'm not sure favorites are all that useful. Personally, I've used them to bookmark comments I'd like to go back to, and/or when someone says something really nice.I don't favorite submissions (or at least, I haven't yet), so I dunno if that would be useful. Very sparse data eitherway.

jschveibinz · Answer

I would propose that first and foremost, the best application would be to improve the quality of HN. It would be interesting to see changes over time from various angles. For example, the churn rate of users, the repetition of posts, the number of commercially-biased posts (I.e. advertisements), etc. I am relatively new to the forum, so I&rsquo;m sure that veteran users would have even better suggestions.

manx · Answer

One could use it to train a recommender system like matrix factorization. The resulting latent vectors for stories/users can be clustered to find similar stories/users.

tedmiston · Answer

Aggregating the most favorited favorites might lead to uncovering some hidden gems.

What would be some interesting uses of analyzing data from HN favorites?

Every HN user's list of favorites is public. Given that it's a list of topics that people find interesting, I imagine some people, especially advertisers, would find it useful.
That's an obvious one though. What kind of wacky ideas have you had for this kind of data?

I'm not sure favorites are all that useful. Personally, I've used them to bookmark comments I'd like to go back to, and/or when someone says something really nice.
I don't favorite submissions (or at least, I haven't yet), so I dunno if that would be useful. Very sparse data eitherway.

One could use it to train a recommender system like matrix factorization. The resulting latent vectors for stories/users can be clustered to find similar stories/users.

Aggregating the most favorited favorites might lead to uncovering some hidden gems.