Twitter is a so-called “micro-blogging” site that provides an open publish-subscribe system to its users. It has a relatively simple design and because of its text-messaging initial public offering it limits “tweets” to 140 characters. It has grown quickly in recent years, has become somewhat saturated with bots, spammers, trolls and celebrities, but remains popular. Its long-anticipated flotation on the stock market was announced in September 2013.
One aspect of Twitter that is particularly attractive to users and to those in the news media, is its rapid response to world events and to the display of topics that are being discussed increasingly at any given time, the trends. Now, computer scientist Yavuz Selim Yilmaz of the University at Buffalo, SUNY, New York, USA, and colleagues there and at Universit`a degli Studi dell’Insubria, in Varese, Italy, have devised a passive sensing system for emerging trends. Their system can analyze the ever-changing twitter stream of endless updates from users and extract not only positive and negative emotions represented by those tweets but display the most critical trends in a way that the simple keyword count used by Twitter to display its trending topics cannot.
The team suggests that there are at least two important uses of their trending algorithm. The first one, they say, is sensing trends in public opinion by using the emotion-category corpus. The second is sensing trends in location-types in a city by using a location-category corpus. “Our experiments show that the proposed methods are able to determine changes in trends effectively in both application scenarios,” they report in the journal Int. J. Ad Hoc and Ubiquitous Computing.
The researchers explain that their system has three characteristics that make it a viable alternative to simplistic trend counting:
Yilmaz, Y.S., Bulut, M.F.,
Akcora, C.G., Bayir, M.A. and Demirbas, M. (2013) ‘Trend sensing via Twitter’, Int. J.
Ad Hoc and Ubiquitous Computing, Vol. 14, No. 1, pp.16–26.
Twitter trend tracking is a post from: David Bradley's Science Spot
via Science Spot http://sciencespot.co.uk/twitter-trend-tracking.html
One aspect of Twitter that is particularly attractive to users and to those in the news media, is its rapid response to world events and to the display of topics that are being discussed increasingly at any given time, the trends. Now, computer scientist Yavuz Selim Yilmaz of the University at Buffalo, SUNY, New York, USA, and colleagues there and at Universit`a degli Studi dell’Insubria, in Varese, Italy, have devised a passive sensing system for emerging trends. Their system can analyze the ever-changing twitter stream of endless updates from users and extract not only positive and negative emotions represented by those tweets but display the most critical trends in a way that the simple keyword count used by Twitter to display its trending topics cannot.
The team suggests that there are at least two important uses of their trending algorithm. The first one, they say, is sensing trends in public opinion by using the emotion-category corpus. The second is sensing trends in location-types in a city by using a location-category corpus. “Our experiments show that the proposed methods are able to determine changes in trends effectively in both application scenarios,” they report in the journal Int. J. Ad Hoc and Ubiquitous Computing.
The researchers explain that their system has three characteristics that make it a viable alternative to simplistic trend counting:
- Our trend sensing framework utilizes a multi-category corpus to detect and process multiple dimensions in tweets. This method enables expanding representation from binary (‘positive or negative’) options with respect to a single dimension to a continuum of options with respect to multiple dimensions, and provides more granularity in trend sensing.
- Our trend sensing framework combines vector and set space methods to identify the trends accurately. From the experimental results, we find that using these two methods together eliminates false positives and improves the accuracy.
- Our trend sensing framework utilizes a dynamic scoring function to give a synopsis (in terms of a list of prominent words) for the cause of the change in trends.
Yilmaz, Y.S., Bulut, M.F.,Akcora, C.G., Bayir, M.A. and Demirbas, M. (2013) ‘Trend sensing via Twitter’, Int. J.
Ad Hoc and Ubiquitous Computing, Vol. 14, No. 1, pp.16–26.
Twitter trend tracking is a post from: David Bradley's Science Spot
via Science Spot http://sciencespot.co.uk/twitter-trend-tracking.html
No comments:
Post a Comment