Info 202 F12

A lot of us love twitter. But it does become overwhelming sometimes. The substantial stream of continuous tweets makes it impossible to peruse all of them. The home timeline accomodates a fixed number of tweets and it is difficult to recall tweets past that certain threshold number of home timeline tweets. This seems to be an obvious issue and a number of attempts have been made solve this. 

From Twitter's in-house feature called 'Lists', which allows users to organize the people they follow, to third party applications like 'TweetDeck' that allows for a lot more categorization, a lot of people have tried to solve this issue. Thirst, attempts to solve the same issue with a fresh new approach. 

Unlike the previous attempts, which were trying to categorize tweets into static categories (i.e. predefined set of classification schemes), Thirst uses advanced NLP techniques to generate dynamic categories for the tweets currently on user's Twitter homepage. A new way to look at Twitter, the app uses its own natural-language processor to filter and organize tweets into an easily digestible format that will change the way you look at your Twitter stream for good. Popular topics are displayed in a grid-like format on your iPhone screen. The more popular the topic, the higher it will appear on your screen and the larger its icon will be. Icons are all topic-specific images, pulled from Wikipedia and other sources on the web. Tapping on a topic will bring up a page where you can read a short description of the subject, see where it is showing up in the news, and get a look for what people are saying about it on Twitter.

To understand the fine distinction,let us consider an example. Consider, Barrack Obama's Twitter Account. With 'List' approach it would be reasonable to classify him into 'Political-News' given the context of most of his tweets. But during a particular event, like say Olympics, his tweets may fall under the categories of 'Sports-News'. This is where the dynamic natural-language processing based classification comes into picture. In Thirst, the tweet would appear in the category that is related to the content and context of the tweet, even if it may have originated from the same user account. As I am writing this I can see '@SeemaHari' in 'Cal' category on Thirst for her tweet about helicopters over Berkeley, as well as in #i202 category about her search related tweet.

Once inside the category, Thirst has 'topic filters' like 'TimeLine', 'Public', 'News & Media' etc ,which unlike how most filters work, can help to show more relevant content in that subject from the best tweeters. So, currently one of my categories is 'Shane Warne' who is a famous cricketer who I follow. The 'Timeline' topic filter shows me but when I click on 'Public' it would show me tweets mentioning 'Shane Warne'. 

The app is totally customizable using topic filters to show you the most relevant content in that subject from the best tweeters. And like Twitter, it’s essentially live streaming. So if one of your friends tweets about an Apple product, his Twitter avatar will automatically show up in the Apple category.

For the interface also, they use indicators like size of 'Subjects' icon and their relative listing order on page to indicate to users which are the most relevant and happening topics on their timelines. The more popular the topic, the higher it will appear on your screen and the larger its icon will be. Icons are all topic-specific images, pulled from Wikipedia and other sources on the web. Tapping on a topic will bring up a page where you can read a short description of the subject, see where it is showing up in the news, and get a look for what people are saying about it on Twitter.

Thirst was started by two recent Berkeley graduates, Anuj Verma and Kunal Modi.

Links:

--------------

[APPLICATION] : http://www.thirst.co

[NEWS] : http://mashable.com/2012/08/01/thirst-iphone-app/

[NEWS] : http://www.tapscape.com/thirst-twitter-app-bridges-social-media-and-newspapers/

[NEWS] : http://techcrunch.com/2012/05/16/thirst-aims-to-slake-your-hunger-for-relevant-twitter-content/

------------

For interest, they used their processing to parse election related tweets and here are the generated infographics:

http://media.tumblr.com/tumblr_mc1y8hGclU1r2c6u1.jpg

http://media.tumblr.com/tumblr_mcde1jsDKU1r2c6u1.jpg

Details of their findings can be found on their blog: 

http://blog.thirst.co