An Empirical Study on Learning to Rank of Tweets

Proceedings of the 23rd International Conference on Computational Linguistics (Coling 2010), pages 295–303, Beijing, August 2010 An Empirical Study on Learning to Rank of Tweets 1Yajuan Duan* 2Long Jiang 2Tao Qin 2Ming Zhou 2Heung-Yeung Shum 1Department of Computer Science and Technology University of Science and Technology of China 2Microsoft Research Asia {v-yaduan,longj,taoqin,mingzhou,hshum}@microsoft.com Abstract Twitter, as one of the most popular micro-blogging services, provides large quantities of fresh information including real-time news, comments, conversation, pointless babble and advertisements. Twitter presents tweets in chronological order. Recently, Twitter introduced a new ranking strategy that considers popularity of tweets in terms of number of retweets. This ranking method, however, has not taken into account content relevance or the twitter account. Therefore a large amount of pointless tweets inevitably flood the relevant tweets. This paper proposes a new ranking strategy which uses not only the content relevance of a tweet, but also the account authority and tweet-specific features such as whether a URL link is included in the tweet. We employ learning to rank algorithms to determine the best set of features with a series of experiments. It is demonstrated that whether a tweet contains URL or not, length of tweet and account authority are the best conjunction.1 1 Introduction Twitter provides a platform to allow users to post text messages known as tweets to update their followers with their findings, thinking and comments on some topics (Java et al., 2007). * The work was done when the first author was intern at Microsoft Research Asia The searched tweets are presented by Twitter in chronological order except the first three, which are ranked by considering popularity of tweets in terms of the number of retweets. This ranking method, however, has not taken into account the content relevance and twitter account; inevitably, a large amount of pointless tweets (Pear Analytics, 2009) may flood the relevant tweets. Although this ranking method can provide fresh information to tweet users, users frequently expect to search relevant tweets to the search queries. For example, consider someone researching consumer responses toward the iPad. He or she would like to find tweets with appropriate comments such as iPad is great or you can find many useful features of iPad, rather than tweets with irrelevant comment, even if they are most recent or popular. Moreover, neither Twitter’s current chronological order based ranking nor the recently introduced popularity based ranking can avoid spam. A developer can accumulate hundreds of thousands of followers in a day or so. At the same time, it is not difficult for spammers to create large quantities of retweets. By contrast, content relevance ranking can effectively prevent spammers from cheating. Different from ranking tweets through chronological order and popularity, a content relevance strategy considers many characteristics of a tweet to determine its ranking level. Thus it is difficult for spammers to break the ranking system by simple methods such as increasing retweet count or number of followers. In this paper, we propose a method to rank the tweets which outputs …

Content truncated to 500 words