George Valkanas

How the live web feels about events

Microblogging platforms, such as Twitter, Tumblr etc., have been established as key components in the contemporary Web ecosystem. Users constantly post snippets of information regarding their actions, interests or perception of their surroundings, which is why they have been attributed the term Live Web. Nevertheless, research on such platforms has been quite limited when it comes to identifying events, but is rapidly gaining ground. Event identification is a key step to news reporting, proactive or reactive crisis management at multiple scales, efficient resource allocation, etc. In this paper, we focus on the problem of automatically identifying events as they occur, in such a user-driven, fast paced and voluminous setting. We propose a novel and natural way to address the issue using notions from emotional theories, combined with spatiotemporal information and employ online event detection mechanisms to solve it at large scale in a distributed fashion. We present a modular framework that incorporates all of our key ideas and experimentally validate its superiority, in terms of both efficiency and effectiveness, over the state-of-the-art using real life data from the Twitter stream. We also present empirical evidence on the importance of spatiotemporal information in event detection for this setting.

Understanding Within-Content Engagement through Pattern Analysis of Mouse Gestures

The availability of large volumes of interaction data and scalable data mining techniques have made possible to study the online behaviour for millions of Web users. Part of the efforts have focused on understanding how users interact and engage with web content. However, the measurement of within-content engagement remains a difficult and unsolved task. This is because of the lack of standardised, well-validated methods for measuring engagement, especially in an online context. To address this gap, we perform a controlled user study where we observe how users respond to online news in the presence or lack of interest. We collect mouse tracking data, which are known to correlate with visual attention, and examine how cursor behaviour can inform user engagement measures. The proposed method does not use any pre-determined concepts to characterise the cursor patterns. We, rather, follow an unsupervised approach and use a large set of features engineered from our data to extract the cursor patterns. Our findings support the connection between gaze and cursor behaviour but also, and more importantly, reveal other dependencies, such as the correlation between cursor activity and experienced affect. Finally, we demonstrate the value of our method by predicting the outcome of online news reading experiences.

Location Extraction from Social Networks with Commodity Software and Online Data

Location is prevalent in most applications nowadays, and is considered a first class citizen in social networks. Locational information is of great significance since it can be used to map information from the online back to the physical world, to contextualize information, or to provide localized recommendations through Location-Based Services (LBS), and can be extended to trajectories and itineraries by adding a time-dimension. Despite all that, location extraction in social networks usually relies on GPS-enabled devices, that provide accurate geodetic coordinates. Nevertheless, most users provide general textual information about their surroundings, such as the city they live in, county or state (or equivalents), without using a GPS-enabled device, which is still valuable information for a number of applications. In this paper, we tackle the problem of extracting location information, usually referred to as geocoding, from additional user-provided content. Instead of using sophisticated and complex algorithms, which are common in online map services but require heavy development and tuning, we rely on software and data which are available online and public ally accessible. We discuss the particularities of geocoding in online social networks and present a simple, lightweight, yet efficient approach for location extraction in such a setting. We finally evaluate our approach experimentally on a large corpus of Twitter users.

Efficient and domain-invariant competitor mining

In any competitive business, success is based on the ability to make an item more appealing to customers than the competition. A number of questions arise in the context of this task: how do we formalize and quantify the competitiveness relationship between two items? Who are the true competitors of a given item? What are the features of an item that most affect its competitiveness? Despite the impact and relevance of this problem to many domains, only a limited amount of work has been devoted toward an effective solution. In this paper, we present a formal definition of the competitiveness between two items. We present efficient methods for evaluating competitiveness in large datasets and address the natural problem of finding the top-k competitors of a given item. Our methodology is evaluated against strong baselines via a user study and experiments on multiple datasets from different domains.

Efficient and adaptive distributed skyline computation

Skyline queries have attracted considerable attention over the last few years, mainly due to their ability to return interesting objects without the need for user-defined scoring functions. In this work, we study the problem of distributed skyline computation and propose an adaptive algorithm towards controlling the degree of parallelism and the required network traffic. In contrast to state-of-the-art methods, our algorithm handles efficiently diverse preferences imposed on attributes. The key idea is to partition the data using a grid scheme and for each query to build on-the-fly a dependency graph among partitions which can help in effective pruning. Our algorithm operates in two modes: (i) full-parallel mode, where processors are activated simultaneously or (ii) cascading mode, where processors are activated in a cascading manner using propagation of intermediate results, thus reducing network traffic and potentially increasing throughput. Performance evaluation results, based on real-life and synthetic data sets, demonstrate the scalability with respect to the number of processors and database size.

SkyDiver: a framework for skyline diversification

Skyline queries have attracted considerable attention by the database community during the last decade, due to their applicability in a series of domains. However, most existing works tackle the problem from an efficiency standpoint, i.e., returning the skyline as quickly as possible. The user is then presented with the entire skyline set, which may be in several cases overwhelming, therefore requiring manual inspection to come up with the most informative data points. To overcome this shortcoming, we propose a novel approach in selecting the k most diverse skyline points, i.e., the ones that best capture the different aspects of both the skyline and the dataset they belong to. We present a novel formulation of diversification which, in contrast to previous proposals, is intuitive, because it is based solely on the domination relationships among points. Consequently, additional artificial distance measures (e.g., Lp norms) among skyline points are not required. We present efficient approaches in solving this problem and demonstrate the efficiency and effectiveness of our approach through an extensive experimental evaluation with both real-life and synthetic data sets.

Debugging applications created by a Domain Specific Language: The IPAC case

Nowadays, software developers have created a large number of applications in various research domains of Computer Science. However, not all of them are familiar with the majority of the research domains. Hence, Domain Specific Languages (DSLs) can provide an abstract, concrete description of a domain in terms that can easily be managed by developers. The most important in such cases is the provision of a debugger for debugging the generated software based on a specific DSL. In this paper, we propose and present a simple but efficient debugger created for the needs of the IPAC system. The debugger is able to provide debugging facilities to developers that define applications for autonomous mobile nodes. The debugger can map code lines between the initial application workflow and the final code defined in a known programming language. Finally, we propose a logging server responsible to provide debugging facilities for the IPAC framework. The IPAC system is consisted of a number of middleware services for mobile nodes acting in a network. In this system a number of mobile nodes exchanged messages that are visualized for more efficient manipulation.

Twitter Floods when it Rains: A Case Study of the UK Floods in early 2014

Twitter is one of the most prominent social media platforms nowadays. A primary reason that has brought the medium at the spotlight of academic attention is its real-time nature, with people constantly uploading information regarding their surroundings. This trait, coupled with the services data access policy for researchers and developers, has allowed the community to explore Twitters potential as a news reporting tool. Finding out promptly about newsworthy events can prove extremely useful in crisis management situations. In this paper, we explore the use of Twitter as a mechanism used in disaster relief, and consequently in public safety. In particular, we perform a case study on the floods that occurred in the United Kingdom during January 2014, and how these were reflected on Twitter, according to tweets (i.e., posts) submitted by the users. We present a systematic algorithmic analysis of tweets collected with respect to our use case scenario, supplemented by visual analytic tools. Our objective is to identify meaningful and effective ways to take advantage of the wealth of Twitter data in crisis management, and we report on the findings of our analysis.

A UI Prototype for Emotion-Based Event Detection in the Live Web

Microblogging platforms are at the core of what is known as the Live Web: the most dynamic, and fast changing portion of the web, where content is generated constantly by the users, in snippets of information. Therefore, the Live Web (or Now Web) is a good source of information for event detection, because it reflects what is happening in the physical world in a timely manner. Meanwhile, it introduces constraints and challenges: large volumes of unstructured, noisy data, which are also as diverse as the users and their interests. In this work we present a prototype User Interface (UI) of our TwInsight system, which deals with event detection of real-world phenomena from microblogs. Our system applies i) emotion extraction techniques on microblogs, and ii) location extraction techniques on user profiles. Combining these two, we convert highly unstructured content to thematically enriched, locational information, which we present to the user through a unified front-end. A separate area of the UI is used to show events to the user, as they are identified. Taking into account the characteristics of the setting, all of the components are updated along the temporal dimension. We discuss each part of our UI in detail, and present anecdotal evidence of its operation through two real-life event examples.

An architecture for detecting events in real-time using massive heterogeneous data sources

The wealth of information that is readily available nowadays grants researchers and practitioners the ability to develop techniques and applications that monitor and react to all sorts of circumstances: from network congestions to natural catastrophies. Therefore, it is no longer a question of whether this can be done, but how to do it in real-time, and if possible proactively. Consequently, it becomes a necessity to develop a platform that will aggregate all the necessary information and will orchestrate it in the best way possible, towards meeting these goals. A main problem that arises in such a setting is the high diversity of the incoming data, obtained from very different sources such as sensors, smart phones, GPS signals and social networks. The large volume of the incoming data is a gift that ensures high quality of the produced output, but also a curse, because higher computational resources are needed. In this paper, we present the architecture of a framework designed to gather, aggregate and process a wide range of sensory input coming from very different sources. A distinctive characteristic of our framework is the active involvement of citizens. We guide the description of how our framework meets our requirements through two indicative use cases.