I’m giving a talk this Friday (September 26, from 6 – 8 PM) on my current research into interest detection on Twitter. The talk is sponsored by the NY Chapter of the IEEE Systems, Man, Cybernetics Society and is at at Long Island University, Brooklyn Campus, Room HS 118. (Directions)
Abstract
This talk introduces graph analysis and how it can be used to infer a Twitter user’s interests. Activities such as marketing, promotion, and outreach all require a level of intimacy with an individual’s interests to make a lasting, positive impression. Few organizations have the resources to achieve personalized service and often have to rely on lower resolution demographic data coupled with surveys. Social media data heralds a new era of personalization, where any organization can access an individual’s interests.
Model API
One of my goals for the talk is to solicit feedback on a model API that queries user interests directly. Essentially, the API will give you the same data that is available at Me Meme but in raw JSON. I’m working on a client library in R that would connect directly to the API. The primary use case I’m thinking of is extracting the interests of users that engaged with a particular hashtag.
I’ll be writing more about this idea after I give the talk. Any thoughts/comments are welcome.
SocialFunction() said:
Really wish I could attend!
Is it community finder-related? As in trying to uncover nodes’ preferences(homophily) by looking at their ties?
LikeLike
Brian Lee Yung Rowe said:
Yes, quite so. Are you doing research in this area? Would love to hear about your work.
LikeLike
SocialFunction() said:
I did some research in the area but nothing too thorough. At the time I would run the community finder algorithm proposed here (http://arxiv.org/pdf/0803.0476.pdf). Then do a bit of text mining within each community (user desciption, profiles, recent tweets, etc.) to put to view shared attributes, topics of interest and the like.
I just completely failed to find a way to scale it and increase reproducibility; it was all very tedious.
Will we be able to see your talk online? Maybe a transcript? I’d really like to see the approach you take! Thanks Brian.
LikeLike
Christine said:
Hi Brian,
Came across this post poking around R + Twitter blog posts…have you made any progress on the idea of a model API? I’ve worked a bit with the NY Times API that provides data in JSON, but I’ve yet to get back into the Twitter data-scraping world (haven’t done much since 2012, right before they changed their API)…
LikeLike
Brian Lee Yung Rowe said:
Christine, Hi I’m actually in the middle of releasing an API for a number of models. If you are interested in beta testing, please let me know! Brian
LikeLike
Christine said:
Brian – quite interested!
LikeLike
Audun said:
If you need more beta testers, I would be very interested to try it out. Best, Audun
LikeLike