In recent years, microblogs have become an important source for reporting real-world events. A real-world occurrence reported in microblogs is also called a social event. Social events may hold critical materials that describe the situations during a crisis. In real applications, such as crisis management and decision making, monitoring the critical events over social streams will enable watch officers to analyze a whole situation that is a composite event, and make the right decision based on the detailed contexts such as what is happening, where an event is happening, and who are involved. Although there has been significant research effort on detecting a target event in social networks based on a single source, in crisis, we often want to analyze the composite events contributed by different social users. So far, the problem of integrating ambiguous views from different users is not well investigated. To address this issue, we propose a novel framework to detect composite social events over streams, which fully exploits the information of social data over multiple dimensions. Specifically, we first propose a graphical model called location-time constrained topic (LTT) to capture the content, time, and location of social messages. Using LTT, a social message is represented as a probability distribution over a set of topics by inference, and the similarity between two messages is measured by the distance between their distributions. Then, the events are identified by conducting efficient similarity joins over social media streams. To accelerate the similarity join, we also propose a variable dimensional extendible hash over social streams. We have conducted extensive experiments to prove the high effectiveness and efficiency of the proposed approach.

Event detection over twitter social media streams

Groups play an essential role in many social websites which promote users' interactions and accelerate the diffusion of information. Recommending groups that users are really interested to join is significant for both users and social media. While traditional group recommendation problem has been extensively studied, we focus on a new type of the problem, i.e., event-based group recommendation. Unlike the other forms of groups, users join this type of groups mainly for participating offline events organized by group members or inviting other users to attend events sponsored by them. These characteristics determine that previously proposed approaches for group recommendation cannot be adapted to the new problem easily as they ignore the geographical influence and other explicit features of groups and users. In this paper, we propose a method called Pairwise Tag enhAnced and featuRe-based Matrix factorIzation for Group recommendAtioN (PTARMIGAN), which considers location features, social features, and implicit patterns simultaneously in a unified model. More specifically, we exploit matrix factorization to model interactions between users and groups. Meanwhile, we incorporate their profile information into pairwise enhanced latent factors respectively. We also utilize the linear model to capture explicit features. Due to the reinforcement between explicit features and implicit patterns, our approach can provide better group recommendations. We conducted a comprehensive performance evaluation on real word data sets and the experimental results demonstrate the effectiveness of our method.

/pdf/combining-latent-factor-model-with-location-features-for-3shxqo9d3l.pdf

Combining latent factor model with location features for event-based group recommendation

Many recommenders aim to provide relevant recommendations to users by building personal topic interest profiles and then using these profiles to find interesting contents for the user. In social media, recommender systems build user profiles by directly combining users' topic interest signals from a wide variety of consumption and publishing behaviors, such as social media posts they authored, commented on, +1'd or liked. Here we propose to separately model users' topical interests that come from these various behavioral signals in order to construct better user profiles. Intuitively, since publishing a post requires more effort, the topic interests coming from publishing signals should be more accurate of a user's central interest than, say, a simple gesture such as a +1. By separating a single user's interest profile into several behavioral profiles, we obtain better and cleaner topic interest signals, as well as enabling topic prediction for different types of behavior, such as topics that the user might +1 or comment on, but might never write a post on that topic. To do this at large scales in Google+, we employed matrix factorization techniques to model each user's behaviors as a separate example entry in the input user-by-topic matrix. Using this technique, which we call "behavioral factorization", we implemented and built a topic recommender predicting user's topical interests using their actions within Google+. We experimentally showed that we obtained better and cleaner signals than baseline methods, and are able to more accurately predict topic interests as well as achieve better coverage.

/pdf/improving-user-topic-interest-profiles-by-behavior-35h4sewafo.pdf

Improving User Topic Interest Profiles by Behavior Factorization

This paper presents a new image hashing that is designed with tensor decomposition (TD), referred to as TD hashing, where image hash generation is viewed as deriving a compact representation from a tensor. Specifically, a stable three-order tensor is first constructed from the normalized image, so as to enhance the robustness of our TD hashing. A popular TD algorithm, called Tucker decomposition, is then exploited to decompose the three-order tensor into a core tensor and three orthogonal factor matrices. As the factor matrices can reflect intrinsic structure of original tensor, hash construction with the factor matrices makes a desirable discrimination of the TD hashing. To examine these claims, there are 14,551 images selected for our experiments. A receiver operating characteristics (ROC) graph is used to conduct theoretical analysis and the ROC comparisons illustrate that the TD hashing outperforms some state-of-the-art algorithms in classification performance between the robustness and discrimination.

/pdf/robust-image-hashing-with-tensor-decomposition-4jv0j1a0ps.pdf

Robust Image Hashing with Tensor Decomposition

Crowdfunding has gained a widespread popularity by fueling the creative minds of entrepreneurs. Not only has it democratized the funding of startups, it has also bridged the gap between the venture capitalists and the entrepreneurs by providing a plethora of opportunities for people seeking to invest in new business ventures. Nonetheless, despite the huge success of the crowdfunding platforms, not every project reaches its funding goal. One of the main reasons for a project's failure is the difficulty in establishing a linkage between it's founders and those investors who are interested in funding such projects. A potential solution to this problem is to develop recommendation systems that suggest suitable projects to crowdfunding investors by capturing their interests. In this paper, we explore Kickstarter, a popular reward-based crowdfunding platform. Being a highly heterogeneous platform, Kickstarter is fuelled by a dynamic community of people who constantly interact with each other before investing in projects. Therefore, the decision to invest in a project depends not only on the preference of individuals, but also on the influence of groups that a person belongs and the on-going status of the projects. In this paper, we propose a probabilistic recommendation model, called CrowdRec, that recommends Kickstarter projects to a group of investors by incorporating the on-going status of projects, the personal preference of individual members, and the collective preference of the group . Using a comprehensive dataset of over 40K crowdfunding groups and 5K projects, we show that our model is effective in recommending projects to groups of Kickstarter users.

/pdf/probabilistic-group-recommendation-model-for-crowdfunding-4d6xalubdp.pdf

Probabilistic Group Recommendation Model for Crowdfunding Domains

The explosion of multimedia content in social media networks raises a great demand of developing tools to facilitate producing, sharing and viewing media content. Flickr groups, self-organized communities with declared common interests, are able to help users to conveniently participate in social media network. In this paper, we address the problem of automatically recommending groups to users. We propose to simultaneously exploit media contents and link structures between users and groups. To this end, we present a probabilistic latent topic model to model them in an integrated framework, expecting to jointly discover the latent interests for users and groups and simultaneously learn the recommendation function. We demonstrate the proposed approach on the dataset crawled from Flickr.com.

http://net.pku.edu.cn/~cuibin/Papers/2012%20IR.pdf

Recommending Flickr groups with social topic model

Estimating the number of distinct flows, also called the cardinality, is an important issue in many network applications, such as traffic measurement, anomaly detection, etc. The challenge is that high accuracy should be achieved with line speed and small auxiliary memory. Flajolet-Martin algorithm, LogLog algorithm, and HyperLogLog algorithm form a line of work in this area with improving performance. In this paper, we propose refined versions of these algorithms to achieve higher accuracy. The key observations are (1) the “leftmost” hash functions used by these algorithms can be generalized to reach higher accuracy, (2) the amendment coefficient can be highly biased in some certain streams or datasets so dynamically setting the amendment coefficient instead of using the one derived in pure math can lead to much better accuracy. Experimental results show great improvement of accuracy and stability of the refined versions over original algorithms.

Fine-grained probability counting for cardinality estimation of data streams

Sketches are being extensively used in a large number of real world applications to estimate frequencies of data items. Due to the unprecedented increase in the amount of Internet data and a relatively slower increase in the size of on-chip memories, existing sketches are becoming increasingly unable to keep the accuracy of the frequency estimates at an acceptable level. In this paper, we design a new sketch, called FID-sketch, that has a significantly higher accuracy and a much smaller on-chip memory footprint compared to the existing sketches. The key intuition behind the design of the FID-sketch is that before inserting an item, unlike prior sketches, it first estimates the current value of the frequency of that item stored in the sketch, and then increments as few counters as possible instead of incrementing a pre-determined fixed number of counters. We carried out extensive experiments to evaluate and compare the performance of FID-sketch with existing sketches on multi-core CPU and GPU platforms. Our experimental results show that our FID-sketch significantly outperforms the state-of-the-art with 36.7 times lower relative error. We have released the source code of our proposed sketch and other related sketches that we implemented at Github [21].

Hao Wang

Papers

Recommending Flickr groups with social topic model

Fine-grained probability counting for cardinality estimation of data streams

FID-sketch: an accurate sketch to store frequencies in data streams