Weakly supervised joint sentiment-topic detection from text

Sentiment analysis or opinion mining aims to use automated tools to detect subjective information such as opinions, attitudes, and feelings expressed in text. This paper proposes a novel probabilistic modeling framework called joint sentiment-topic (JST) model based on latent Dirichlet allocation (LDA), which detects sentiment and topic simultaneously from text. A reparameterized version of the JST model called Reverse-JST, obtained by reversing the sequence of sentiment and topic generation in the modeling process, is also studied. Although JST is equivalent to Reverse-JST without a hierarchical prior, extensive experiments show that when sentiment priors are added, JST performs consistently better than Reverse-JST. Besides, unlike supervised approaches to sentiment classification which often fail to produce satisfactory performance when shifting to other domains, the weakly supervised nature of JST makes it highly portable to other domains. This is verified by the experimental results on data sets from five different domains where the JST model even outperforms existing semi-supervised approaches in some of the data sets despite using no labeled documents. Moreover, the topics and topic sentiment detected by JST are indeed coherent and informative. We hypothesize that the JST model can readily meet the demand of large-scale sentiment analysis from the web in an open-ended fashion.

Minority influence and argument generation

The theories of Moscovici (1980) and Nemeth (1986) concerning the cognitive processes underlying minority influence are examined in an argument generation paradigm. While Moscovici (1980) argues that minority influence increases the generation of arguments for and against the minority position, Nemeth (1986) proposes that minorities induce divergent thinking which leads to the generation of a wider range of arguments which are more original. In the first study, subjects read a minority text and then generated arguments concerning the minority issue within a specified time. The second study was similar to the first and included a condition where minority influence followed partial sensory deprivation (being placed in a dark, soundproof room for 45 minutes) which was predicted to decrease cognitive effort. Contrary to Moscovici, in neither study was there evidence that a minority led to more arguments being generated compared to a control condition (no influence). However, in one study, a minority led to more arguments being generated in the minority than in the majority direction. However, as predicted by Nemeth, in both studies a minority resulted in a wider range of arguments being generated than those proposed in the minority's message and these were rated by independent judges as being more original. Finally, as predicted, partial sensory deprivation led to a narrower range of arguments which were focused more upon issues raised in the minority text.

Timeline generation with social attention

Timeline generation is an important research task which can help users to have a quick understanding of the overall evolution of any given topic. It thus attracts much attention from research communities in recent years. Nevertheless, existing work on timeline generation often ignores an important factor, the attention attracted to topics of interest (hereafter termed "social attention"). Without taking into consideration social attention, the generated timelines may not reflect users' collective interests. In this paper, we study how to incorporate social attention in the generation of timeline summaries. In particular, for a given topic, we capture social attention by learning users' collective interests in the form of word distributions from Twitter, which are subsequently incorporated into a unified framework for timeline summary generation. We construct four evaluation sets over six diverse topics. We demonstrate that our proposed approach is able to generate both informative and interesting timelines. Our work sheds light on the feasibility of incorporating social attention into traditional text mining tasks. Copyright © 2013 ACM.

An angle-based interest model for text recommendation

Building an interest model is the key to realize personalized text recommendation. Previous interest models neglect the fact that a user may have multiple angles of interests. Different angles of interest provide different requests and criteria for text recommendation. This paper proposes an interest model that consists of two kinds of angles: persistence and pattern, which can be combined to form complex angles. The model uses a new method to represent the long-term interest and the short-term interest, and distinguishes the interest on object and the interest on the link structure of objects. Experiments with news-scale text data show that the interest on object and the interest on link structure have real requirements, and it is effective to recommend texts according to the angles.