Twahpic (pronounced twa:pik), a portmanteau of Twitter and topic, is a Microsoft Research demonstration project that combines semi-supervised topic modeling with twitter feeds to allow people to explore what a set of tweets (from a user, matching a query) is all about.
Twahpic’s topics are generated by analyzing about a week’s worth of tweets from Twitter, then building special models that encapsulate the language used to describe various concepts. Often, these concepts clearly coalesce around nouns such as people and places. Topics can also be less well defined, such as positive adjectives, or greeting words. It then identifies whether each topic is social, status, substance, or style.
Twahpic or Twitter Topic Modeling is thus a project aimed at modeling tweets. According to Microsoft, “Twahpic” shows what tweets on Twitter are about. Twahpic uses Partially Labeled Latent Dirichlet Analysis (PLDA) to identify 200 topics used on Twitter. This is really amazing in terms of results it provides, because before this I have never found something which can really model my interests along with the substance, social, status and style.
What impresses me even more is that it is hosted on Windows Azure. Yes, it is a cloud based app which substantiates the investment Microsoft is doing in Cloud Services.
Let me give you a glimpse of it:
Visit Twahpic and enter the name of the person of whose tweets you want to model or just write any query. Here we have taken the example of @TheWindowsClub tweets.
You can see what we tweet about! This is just an interpretation of our tweets which helps folks in understanding our content. It can also be fun trying to search with keywords like fashion, movies, clothes etc. I am sure you will enjoy this!