Michalis Vazirgiannis

Graph of Words: Boosting Text Mining Tasks with Graphs

The Bag-of-words model has been the dominant approach for IR and Text mining for many years assuming the word independence and the frequencies as the main feature for feature selection and for query to document similarity. Although the long and successful usage, bag- of-words ignores words order and distance within the document -- weakening thus the expressive power of the distance metrics. We propose graph-of-word, an alternative approach that capitalizes on a graph representation of documents and challenges the word independence assumption by taking into account words order and distance. We applied graph-of-word in various tasks such as ad-hoc Information Retrieval, Single-Document Keyword Extraction, Text Categorization and Sub-event Detection in Textual Streams. In all cases the the graph of word approach, assisted by degeneracy at times, outperforms the state of the art base lines in all cases.

Kernel Graph Convolutional Neural Nets

Graph kernels have been successfully applied to many graph classification problems. Typically, a kernel is first designed, and then an SVM classifier is trained based on the features defined implicitly by this kernel. This two-stage approach decouples data representation from learning, which is suboptimal. On the other hand, Convolutional Neural Networks (CNNs) have the capability to learn their own features directly from the raw data during training. Unfortunately, they cannot handle irregular data such as graphs. We address this challenge by using graph kernels to embed meaningful local neighborhoods of the graphs in a continuous vector space. A set of filters is then convolved with these patches, pooled, and the output is then passed to a feedforward network. With limited parameter tuning, our approach outperforms strong baselines on 7 out of 10 benchmark datasets.

BigNet 2018 Chairs' Welcome & Organization

It is our great pleasure to welcome you to the 2018 International Workshop on Learning Representations for Big Networks (BigNet@WWW2018). This is the third edition of the BigNet workshop series, following its inauguration at the 25th ACM International Conference on Information and Knowledge Management (CIKM 2016) and the second edition at the 26th International World Wide Web Conference (WWW 2017). Recent years have witnessed the emergence of network representation learning research. Different from the feature engineering process in conventional analysis, network representation learning, also know as network embedding, aims to learn the latent low-dimensional representations for objects in networks, such as nodes, links, and groups. Its ultimate objective is to encode networks structural properties into the latent representations, benefiting all existing network mining tasks, such as node classification, link prediction, community detection, etc. In the BigNet 2018 workshop, we aim to provide a forum for presenting the most recent advances in network representation learning to unearth rich knowledge.

Error-space representations for multi-dimensional data streams with temporal dependence

AbstractIn many application scenarios, data points are not only temporally dependent, but also expected in the form of a fast-moving stream. A broad selection of efficient learning algorithms exists which may be applied to data streams, but they typically do not take into account the temporal nature of the data.n We motivate and design a method which creates an efficient representation of a data stream, where temporal information is embedded into each instance via the error space of forecasting models. Unlike many other methods in the literature, our approach can be rapidly initialized and does not require iterations over the full data sequence, thus it is suitable for a streaming scenario. This allows the application of off-the-shelf data-stream methods, depending on the application domain.n In this paper, we investigate classification. We compare to a large variety of methods (auto-encoders, HMMs, basis functions, clustering methodologies, and PCA) and find that our proposed methods perform very competitively, and offers much promise for future work.

Special focus on natural language processing and social computing

Natural language processing (NLP) and social network analysis (SNA) technologies are among the most active research and development areas due to rapid advancement of the Internet as well as the worldwide proliferation of social media. We are living in an increasingly networked world. People, information and other entities are connected via World Wide Web, email networks, instant messaging networks, mobile communication networks, online social networks, etc. These online networks grow fast and possess huge amount of recorded information, which presents great opportunities in understanding the science of these networks, and in developing new applications for these networks. However, new challenges have to be met — the networks are huge and information is noisy, and they demand new methodologies in analyzing these networks, and in developing theories and applications for the big networked data.

Introduction to Special Issue on Social Media Processing ( TIST - SMP )

We are living in an increasingly networked world. People in and across organizations are connected via the World Wide Web, email networks, instant messaging networks, mobile communication networks, online social networks, and so on. These online networks grow quickly and possess vast amounts of recorded information, which presents unprecedented opportunities for analyzing and developing applications. The networks also bring new challenges. For example, the networks are enormous and their information is noisy, which calls for new methodologies, theories, and applications to analyze and understand. This special issue is a forum for recent advances in data mining/machine learning on social media processing, including but not limited to Web search and information retrieval, Web mining, social network analysis, semantic Web, natural language processing, text mining, and computational advertising. We are particularly interested in the articles on how to handle crisis and disaster situations with the help of social Web and mining. The special issue attracted 42 submissions covering a wide range of research topics. After two rounds of reviews (a few papers underwent more rounds of reviews), we selected seven high-quality papers for publication in this special issue. The selected papers span various topics, such as user behavior modeling, community detection, and social media content mining. There are several interesting topics of research in the submissions. The first line of research focuses on the detection of specialized user groups or user characteristics from social media. Tu et al. presented a PRISM framework for identifying user professions, using both social media contents and community network information. Chen et al. examined an interesting problem of detecting target predefined user groups from mobile social networks based on specific communication patterns. Huang et al. developed a method to distinguish residents from visitors using spatiotemporal check-in data. Finally, Fu et al. proposed an interesting algorithm that detects spam users by leveraging and quantifying users’ carefulness. The second type of research focuses on community detections. Chikhaoui et al. explored the potential of detecting communities of authorities and offered findings for understanding the influence in dynamic social networks. Another research topic is related to social content. Glenski and Weninger presented an approach for automatically rating and ranking social media posts and comments. Li et al. developed a framework for personalizing recommendations in microblogging social networks. Social media processing has been one of the most active fields of research in both academia and industry, and it will continue to grow. We hope this special issue offers a flavor of the state-of-the-art researches on some of the important topics in this field.