The first paper was on cyber bullying. The paper is entitle “Expert knowledge for automatic detection of bullies in social networks”, by Maral Dadvar, dolf Trieschnigg and Franciska de Jong of Twente University. This paper was one of the runners-up for the best paper award. Maral Dadvar was presenting. Cyberbullying is a serious social problem in online environments and social networks. In this knowledge elicitation study a multi-criteria evaluation system was introduced for using knowledge of human experts to analyse YouTube users’ behaviour and their characteristics. Based on expert knowledge, the system assigns a score to the users, which represents their level of “bulliness” based on the history of their activities. The scores can be used to discriminate among users with a bullying history and those who were not engaged in hurtful acts. The apporach can be used for other social networks as well.
The other two papers of the session were on analysing Twitter streams. Teh second paper of the session was on Tweet-classification. The paper was entitled “User interest prediction for tweets using semantic enrichment with DBpedia” and was written by Denis Lukovnikov Mathias Verbeke Bettina Berendt from KU Leuven. Denis Lukovnikov was presenting. The paper focused on topic-based prediction of interest of individual users to posts in the context of Twitter. Two methods for enriching tweets using DBpedia for the purposes of classification are proposed. The first method incorporates entity linking and uses linked entities in a tweet to improve classification, whereas the second method aims to improve upon the first one by adding information derived from DB- pedia about entities found using the first method. The two methods are evaluated with respect to tweet classification. It was found that Deep tweet enrichment gives better classification results than shallow enrichment. Deep enrichment ex- tracts descriptive entities from the taglines of (observable) entities that were found by linking tweet texts to DBpedia and adds these descriptive entities as features (together with observable entities) whereas shallow enrichment only adds observable entities as features.
The third and last paper of the session was also on Tweet-analysis, although this time the topic was how to derive temporal information from tweets. The paper was entitled: “Predicting time-to-event from Twitter messages” and was written by Hannah Tops (Utrecht), Antal van den Bosch and Florian Kunneman (Nijmegen). The paper describes a system that estimates when an event is going to happen from a stream of microtexts on Twitter that are referring to that event. Using a Twitter archive and 60 known football events, a machine learning classifiers is trained to map unseen tweets onto discrete time segments. The time period before the event is automatically segmented; the accuracy with which tweets can be classified into these segments determines the error of the time-to-event prediction. In a cross-validation experiment the authors observed that support vector machines with χ2 feature selection attain the lowest prediction error of 52.3 hours off. In a comparison with human subjects, humans produce a larger error as well, but recognize more tweets as posted before the event; the machine-learning approach more often misclassifies a ‘before’ tweet as posted during or after the event. The topic dealt with predicting the occurrence of Dutch soccer matches. A fair number of questions ensued.
In all, the session was well attended, especially considering that it was one of the last sessions of the conference. The societal relevance of the topics and the quality of the papers may have contributed to this fact.
