Zhaoyun Ding

Exploring sentiment parsing of microblogging texts for opinion polling on chinese public figures

Microblogging websites such as twitter and Sina Weibo have attracted many users to share their experiences and express their opinions on a variety of topics, making them ideal platforms on which to conduct electronic opinion polls on products, services and public figures. However, conventional sentiment analysis methods for microblogging messages may not meet the demands of opinion polls for public figures. Therefore, in this study, we focus mainly on the problem of sentiment analysis for opinion polling on Chinese public figures. We propose a sentiment parsing-based architecture, which represents and labels opinion targets and their corresponding sentiments jointly to avoid the mismatching of them, for opinion poll of public figures using microblogs. Furthermore, we formulate sentiment parsing of microblogging sentences as a sequence labeling problem and adapt different Recurrent Neural Network (RNN) models to train and infer the model. Our experimental results demonstrate that the proposed sentiment parsing-based methods achieve better performance than conventional sentiment score-based methods in opinion polling on public figures using microblogs.

Measuring the spreadability of users in microblogs

Message forwarding (e.g., retweeting on Twitter.com) is one of the most popular functions in many existing microblogs, and a large number of users participate in the propagation of information, for any given messages. While this large number can generate notable diversity and not all users have the same ability to diffuse the messages, this also makes it challenging to find the true users with higher spreadability, those generally rated as interesting and authoritative to diffuse the messages. In this paper, a novel method called SpreadRank is proposed to measure the spreadability of users in microblogs, considering both the time interval of retweets and the location of users in information cascades. Experiments were conducted on a real dataset from Twitter containing about 0.26 million users and 10 million tweets, and the results showed that our method is consistently better than the PageRank method with the network of retweets and the method of retweetNum which measures the spreadability according to the number of retweets. Moreover, we find that a user with more tweets or followers does not always have stronger spreadability in microblogs.

An Influence Strength Measurement via Time-Aware Probabilistic Generative Model for Microblogs

Micro-blogging such as Twitter provides users a platform to participate in the discussion of topics and find more interesting friends. While this large number can generate notable diversity and not all influence strengths between users are the same, it also makes measure the influence strength more accurately, that not only rated as binary friendship relations, challenging and interesting. In this work, we develop a time-aware probabilistic generative model to estimate the influence strength by taking the time interval, relationship of following, and the post content into consideration. In particular, the Gibbs sampling is employed to perform approximate inference, and the interval of time and the multi-path influence propagation is incorporated to estimate the indirect influence strength more microscopically according to the propagation of words. Comprehensive experiments has been conducted on a real data set from Twitter, which contains about 0.26 million users and 2.7 million tweets, to evaluate the performance of our proposed approach. As indicated, the experimental results validate the effectiveness of our approach. Furthermore, we also observe that the influence strength ranking by our model is less correlative with the method which ranks the influence strength according to the number of common friends.

A WORD POSITION-RELATED LDA MODEL

LDA (Latent Dirichlet Allocation) proposed by Blei is a generative probabilistic model of a corpus, where documents are represented as random mixtures over latent topics, and each topic is characterized by a distribution over words, but not the attributes of word positions of every document in the corpus. In this paper, a Word Position-Related LDA Model is proposed taking into account the attributes of word positions of every document in the corpus, where each word is characterized by a distribution over word positions. At the same time, the precision of the topic-words interpretability is improved by integrating the distribution of the word-position and the appropriate word degree, taking into account the different word degree in the different word positions. Finally, a new method, a size-aware word intrusion method is proposed to improve the ability of the topic-words interpretability. Experimental results on the NIPS corpus show that the Word Position-Related LDA Model can improve the precision of the topic-words interpretability. And the average improvement of the precision in the topic-words interpretability is about 9.67%. Also, the size-aware word intrusion method can interpret the topic-words semantic information more comprehensively and more effectively through comparing the different experimental data.

Predicting Social Unrest Events with Hidden Markov Models Using GDELT

Proactive handling of social unrest events which are common happenings in both democracies and authoritarian regimes requires that the risk of upcoming social unrest event is continuously assessed. Most existing approaches comparatively pay little attention to considering the event development stages. In this paper, we use autocoded events dataset GDELT (Global Data on Events, Location, and Tone) to build a Hidden Markov Models (HMMs) based framework to predict indicators associated with country instability. The framework utilizes the temporal burst patterns in GDELT event streams to uncover the underlying event development mechanics and formulates the social unrest event prediction as a sequence classification problem based on Bayes decision. Extensive experiments with data from five countries in Southeast Asia demonstrate the effectiveness of this framework, which outperforms the logistic regression method by 7% to 27% and the baseline method 34% to 62% for various countries.

Confirmatory Analysis on Influencing Factors When Mention Users in Twitter

Nowadays, Twitter has become an important platform to expand the diffusion of information or advertisement. Mention is a new feature on Twitter. By mentioning users in a tweet, they will receive notifications and their possible retweets may help to initiate large cascade diffusion of the tweet. To enhance a tweet’s diffusion by finding the right persons to mention, in this paper, we propose three factors that probably have impact on tweet’s diffusion. Specifically, these factors are user vulnerability, user’s online status and spatial location. In this paper, the issue ‘whom to mention when tweeting’ is transformed to the issue ‘choosing users who have higher probability to retweet. By analyzing users retweet behaviors, online status and users’ location in Twitter, we confirm these three factors. Experiments were conducted on a real dataset from Twitter containing about 49,253 users and 563,758 tweets in a target community, and results show that these three factors all have significant impacts on retweeting and information diffusion.

Graph-Based Method for Detecting Occupy Protest Events Using GDELT Dataset

Recent years have witnessed a series of occupy protest events all over the world. Detecting and monitoring these events is an important and challenging task in social science research and also can provide reference for governments emergency management. Existing methods mainly solve this problem by document clustering techniques. This paper proposes a novel graph-based occupy protest event detection framework which applies sub graph pattern mining for this task. A wealth of event data about Occupy Wall Street in New York and Occupy Central in Hong Kong from the Global Data on Events, Location, and Tone (GDELT) are utilized in the work. Experimental results on these datasets show that the proposed method can achieve higher detection accuracy with 0.921 on average and MCC value 0.748, outperforming the baseline method.

Exploring Event Evolution Patterns at the Atomic Level

The event evolution mining for news corpus is beneficial for people who are less interested in the set of documents related by a topic rather than the underlying stories. Most of state-of-the-art approaches which derived from the TDT field considered events at the document level, which made different granularity for each event in evolution graph. In this paper, we consider events from a unified perspective by introducing the IE techniques into this task. We propose an unsupervised approach to explore event evolution patterns through extracting atomic event from documents, identifying their co-reference, and measuring their relationship automatically. And then we construct the event evolution graph. Meanwhile, we propose two policies to find a newsworthy subset from a mass of atomic events in a news corpus to simplify the event evolution graph. Finally, we show experimentally that our method which works on a Chinese news corpus can construct the atomic event evolution graph, whose vertexes are standing for important events of the topic, and edges are reasonable relationship between their adjacent events.

Finding the Optimal Users to Mention in the Appropriate Time on Twitter

Nowadays, Twitter has become an important platform to expand the diffusion of information or advertisement. Mention is a new feature on Twitter. By mentioning users in a tweet, they will receive notifications and their possible retweets may help to initiate large cascade diffusion of the tweet. In order to maximize the cascade diffusion, two important factors need to be considered: (1) The mentioned users will be interested the tweet; (2) The mentioned users should be online. The second factor was mainly studied in this paper. If we mention users when they are online, they will receive notifications immediately and their possible retweets may help to maximize the cascade diffusion as quickly as possible. In this paper, an unbalance assignment problem was proposed to ensure that we mentioned the optimal users in the appropriate time. In the assignment problem, constraints were modeled to overcome the overload problems on Twitter. Further, the unbalance assignment problem was converted to a balance assignment problem, and the Hungarian algorithm was took to solve the above problem. Experiments were conducted on a real dataset from Twitter containing about 2 thousand users and 5 million tweets in a target community, and results showed that our method was consistently better than mentioning users randomly.

Mentioning the Optimal Users in the Appropriate Time on Twitter

Nowadays, Twitter has become an important platform to expand the diffusion of information or advertisement. Mention is a new feature on Twitter. By mentioning users in a tweet, they will receive notifications and their possible retweets may help to initiate large cascade diffusion of the tweet. In order to maximize the cascade diffusion, two important factors need to be considered: (1) The mentioned users will be interested the tweet; (2) The mentioned users should be online. The second factor was mainly studied in this paper. If we mention users when they are online, they will receive notifications immediately and their possible retweets may help to maximize the cascade diffusion as quickly as possible. In this paper, an unbalance assignment problem was proposed to ensure that we mentioned the optimal users in the appropriate time. In the assignment problem, constraints were modeled to overcome the overload problems on Twitter. Further, the unbalance assignment problem was converted to a balance assignment problem, and the Hungarian algorithm was took to solve the above problem. Experiments were conducted on a real dataset from Twitter containing about 2 thousand users and 5 million tweets in a target community, and results showed that our method was consistently better than mentioning users randomly.