Ask HN:interest in api where given a Twitter login, you get their interests? | Hacker News Reader

Ask HN:interest in api where given a Twitter login, you get their interests?

30 pointstoisanji20 comments

I'm working on a service that crawls through all the data for a given twitter user and figures out what they are interested in. I have an early stage prototype that shows some of the information that can be returned. I am planning to offer this as an api for other websites to use and integrate so they can better personalize their offering. Would something like this be interesting to you guys? What other features and information would you like? Currently I am working on getting more accurate results and then getting more in depth interests (X likes biking, swimming vs X likes sports)

Here is an example site I built using the api:

http://badgeu.com/t/davemcclure

http://badgeu.com/tags/show?name=vc

You can also see what interests your own twitter profile gets by putting in your twitter login at http://badgeu.com If anyone is interested in using the api, please email me at jason@socmetrics.com

Clickable Links:

http://badgeu.com/t/davemcclure

http://badgeu.com/tags/show?name=vc

where I see the real value in this, is suggesting similar users. It seems like twitter suggestions are based on your social graph, but I believe there to be much more value in tweet content analysis. Creating simple badges allows you to process once, and then reduce the analysis required when build your similar users suggestions eg.instead of doing a full text analysis between users you can use these simple binary flags to correlate using Jaccard index or Tanimoto coefficient.

What I was missing is explanations for the badges: what actually got me each badge? Like dannytatom, I would have loved to know what a "Fisher Price First Tweet" is. Especially if you're talking about "winning badges", I would love to know what I need to do to win a badge (are you going to have badges that are "upgrades" from previous badges?)

One of my badges is "Fisher Price First Tweet," but I have no idea what it means and unlike the others it doesn't link anywhere. What is it?

Also, I laughed that one of my favorite topics is #fuckisuck. :(

All of my 'favourite topics' are hashtags I've only used once (and in jest).

My results were interesting and relatively accurate.

I don't get what the second link does.

I could imagine somebody finding it useful, but I can't think of anything I'd do that would use it.

The second link is just a convenience page showing what all the top VCs get for badges.

Ok, gotcha. I tried changing vc to some things that I knew were common hashtags and it didn't find any results. Maybe /lists/ would be a better url than /tags/

For me, it seems to be fairly accurate, but mostly based on recent tweets (vs. my entire history of tweets). Is there a limit to the number of tweets it will actually use to evaluate?

It currently only evaluates recent tweets for resource reasons, but I do plan to evaluate all of a single users's tweets. We can see the peaks and dips of when someone is talking about something.

This is very neat. Quite surprising results too, which is great.

No idea whether you can monetize this, or whether you should, but it's certainly fun. Kudos.

I noticed on less geeky folk that there aren't many "badges". Might want to work on that.

Yeah, nice noticing that, I've been testing this with mostly tech people, but will be testing this more with "mainstream" people.

Seems like a cool service, do you utilize only the "#" tags or all the words in a tweet?

Its not hash tags, it's a combination of several data points

I just emailed you but I got a bounce, "address rejected".

jason@socmetrics.com

This is a cool app. How much time have you put into it?

This was a side project to my main project for 2 months, but I'm planning to move to this full-time if I can find enough interest in this.

How are the badges generated?

Reply on news.ycombinator.com