In this paper we cluster and analyse temporal user behaviour in online communities. We adapt a simple unsupervised clustering algorithm to an evolutionary setting where we cluster users into prototypical behavioural roles based on features derived from their ego-centric reply-graphs. We then analyse changes in the role membership of the users over time, the change in role composition of forums over time and examine the differences between forums in terms of role composition. We perform this analysis on 200 forums from a popular national bulletin board and 14 enterprise technical support forums.

We present a new algorithm for inferring the home locations of Twitter users at different granularities, such as city, state, or time zone, using the content of their tweets and their tweeting behavior. Unlike existing approaches, our algorithm uses an ensemble of statistical and heuristic classifiers to predict locations. We find that a hierarchical classification approach can improve prediction accuracy. Experimental evidence suggests that our algorithm works well in practice and outperforms the best existing algorithms for predicting the location of Twitter users.

We propose a logistic regression model taking into account two analytically different sets of factors–structure and action. The factors include individual, dyadic, and triadic properties between ego and alter whose tie breakup is under consideration. From the fitted model using a large-scale data, we discover 5 structural and 7 actional variables to have significant explanatory power for unfollow. One unique finding from our quantitative analysis is that people appreciate receiving acknowledgements from others even in virtually unilateral communication relationships and are less likely to unfollow them: people are more of a receiver than a giver.

We use data extracted from many weblogs to identify the underlying relations of a set of companies in the Standard and Poor (S\&P) 500 index. We define a pairwise similarity measure for the companies based on the weblog articles and then apply a graph clustering procedure. We show that it is possible to capture some interesting relations between companies using this method. As an application of this clustering procedure we propose a cluster-based portfolio selection method which combines information from the weblog data and historical stock prices. Through simulation experiments, we show that our method performs better (in terms of risk measures) than cluster-based portfolio strategies based on company sectors or historical stock prices. This suggests that the methodology has the potential to identify groups of companies whose stock prices are more likely to be correlated in the future.

We present a computational framework for understanding the social aspects of emotions in Twitter conversations. Using unannotated data and semisupervised machine learning, we look at emotional transitions, emotional influences among the conversation partners, and patterns in the overall emotional exchanges. We find that conversational partners usually express the same emotion, which we name Emotion accommodation, but when they do not, one of the conversational partners tends to respond with a positive emotion. We also show that tweets containing sympathy, apology, and complaint are significant emotion influencers. We verify the emotion classification part of our framework by a human-annotated corpus.

We propose a dynamic joint sentiment-topic model (dJST) which allows the detection and tracking of views of current and recurrent interests and shifts in topic and sentiment. Both topic and sentiment dynamics are captured by assuming that the current sentiment-topic specific word distributions are generated according to the word distributions at previous epochs. We derive efficient online inference procedures to sequentially update the model with newly arrived data and show the effectiveness of our proposed model on the Mozilla add-on reviews crawled between 2007 and 2011.

There has been significant recent interest in using the aggregate sentiment from social media sites to understand and predict real-world phenomena. However, the data from social media sites also offers a unique and — so far — unexplored opportunity to study the impact of external factors on aggregate sentiment, at the scale of a society. Using a Twitter-specific sentiment extraction methodology, we the explore patterns of sentiment present in a corpus of over 1.5 billion tweets. We focus primarily on the effect of the weather and time on aggregate sentiment, evaluating how clearly the well-known individual patterns translate into population-wide patterns. Using machine learning techniques on the Twitter corpus correlated with the weather at the time and location of the tweets, we find that aggregate sentiment follows distinct climate, temporal, and seasonal patterns.

We describe a system for automatically ranking documents by degree of militancy, designed as a tool both for finding militant websites and prioritizing the data found. We compare three ranking systems, one employing a small hand-selected vocabulary based on group membership markers used by insiders to identify members and member properties (us) and outsiders and threats (them), one with a much larger vocabulary, and another with a small vocabulary chosen by Mutual Information. We use the same vocabularies to build classifiers. The ranker that achieves the best correlations with human judgments uses the small us-them vocabulary. We confirm and extend recent results in sentiment analysis (paltoglou 2010), showing that a feature-weighting scheme taken from classical IR (TFIDF) produces the best ranking system; we also find, surprisingly, that adjusting these weights with SVM training, while producing a better classifier, produces a worse ranker. Increasing vocabulary size similarly improves classification (while worsening ranking).

Tracking Twitter for public health has shown great potential. However, most recent work has been focused on correlating Twitter messages to influenza rates, a disease that exhibits a marked seasonal pattern. In the presence of sudden outbreaks, how can social media streams be used to strengthen surveillance capacity? In May 2011, Germany reported an outbreak of Enterohemorrhagic Escherichia coli (EHEC). It was one of the largest described outbreaks of EHEC worldwide and the largest in Germany. In this work, we study the crowd's behavior in Twitter during the outbreak. In particular, we report how tracking Twitter helped to detect key user messages that triggered signal detection alarms before MedISys and other well established early warning systems. We also introduce a personalized learning to rank approach that exploits the relationships discovered by: (i) latent semantic topics computed using Latent Dirichlet Allocation (LDA), and (ii) observing the social tagging behavior in Twitter, to rank tweets for epidemic intelligence. Our results provide the grounds for new public health research based on social media.

Sentiment classification has been a well-investigated research area in the computational linguistics community. However, most of the research is primarily focused on detecting simply the polarity in text, often needing extensive manual labeling of ground truth. Additionally, little attention has been directed towards a finer analysis of human moods and affective states. Motivated by research in psychology, we propose and develop a classifier of several human affective states in social media. Starting with about 200 moods, we utilize mechanical turk studies to derive naturalistic signals from posts shared on Twitter about a variety of affects of individuals. This dataset is then deployed in an affect classification task with promising results. Our findings indicate that different types of affect involve different emotional content and usage styles; hence the performance of the classifier on various affects can differ considerably.