An algorithmic approach to GitHub exploration | Hacker News Reader

An algorithmic approach to GitHub exploration

blog.algorithmia.com

26 pointsdoppenhe14 comments

This post highlights that there are indeed some significant untapped opportunities in mining GitHub user and repository data. As I was working on the 2nd Edition of Mining the Social Web last year, I observed the very same thing and introduced an entire chapter that models GitHub as a interest graph. (Think: users are interested in projects and programming languages by extension.) The IPython Notebook with all of the sample code is available with all of the other source [1] but really just begins to scratch the surface with some rudimentary centrality techniques. Like any other interest graph, the possibilities are fairly endless.

[1] http://nbviewer.ipython.org/github/ptwobrussell/Mining-the-S...

absolutelly. Everything is up and running on Algorithmia inclduing some graphing algorithms - would love to collaborate on such exploration if you are interested. diego (at) algorithmia.com

I'll definitely reach out. It would be fun to jointly tackle a fun problem together.

Tried with /JuliaLang/julia and got garbage results - my guess is that the build instructions in the README dominate. Trying something like /JuliaOpt/Optim.jl, which has a very on-topic README, faired slightly better but still had some bizzare things like /sergiotapia/go-style-guide

Yeah some of the results are not great, a lot of READMEs are purely build info or license - we tried to a degree to filter those out. Hopefully you at least discovered one repo that was relevant and hadn't seen before - so much hidden treasure in GitHub.

I like the idea of your project, but it seems like the algorithmic database version of wikipedia that you plan to profiteer off of?

Words like marketplace, crowdsourced, and open platform played well in 2005 but now they kinda smell like a scam.

Another related site that attempts to do a similar thing: http://kare.progger.io

thanks for sharing, I tried it out but i keep getting erros. Do you know the owner of the project ?

I am the owner of the project. What errors are you getting?

404 Not Found

The resource could not be found.

/search/bootstrap

We can't determine which repo to give results for just by the name of the repo because multiple owners could have the same repo name (i.e. bob/bootstrap and jack/bootstrap). So we need both the owner and repo. This works:

http://kare.progger.io/search/twbs/bootstrap

Strangely, they didn't mention what kind of topic algorithm they are using. Is it LDA based?

yes this one is LDA based.

I had the similar feeling.

Reply on news.ycombinator.com