Project topics and materials
Get Complete Project Material File(s) Now! »
Event Definition and Characterization
What is meant by the word “event”? has always been a research question leading to several meanings. This term has received substantial consideration across different fields such as philosophy [20] and computer science [2]. From a broader point of view, a real event is considered as something that happens: a happening, an occurrence, an event [126]. This definition has been extended in a philosophical study to characterize events as an abstract concept in which the meaning depends on the target type such as activity, state or action [20]. From technical point of view, an earlier work in Topic Detection and Tracking (TDT) field defines an event “as something that happens at a particular time and place” [2]. This definition puts emphasis on the spatial-temporal aspect, which seems to be adopted by many other researchers [84, 145]. However, while events can happen at a specific time, other events continue over a long period of time. Moreover, associating a specific location to events fails to handle some events which may happen in different venues. These facts have led to other definitions in the literature attempting to cast an event to just a temporal entity [114] or to stress on the geographical dimension [130]. To sum up, by drawing together all these defini tions, three important views appear to identify what an event is. These views are represented by three Ws questions: what, when and where.
Later on, some researchers point out a missing concept that could define an event. They attempt to pay attention to “who” was involved in the event. Although events can happen without participants, it seems important to consider this aspect when it comes to describe the people’s experiences. Thus, the definition in [2] has been extended to “an event is something that has a specific time, location, and people associated with it” [1]. For instance, it has been shown that the “who” view is important to define a historical event which is described by five elements: object, person, location, time and cause [101]. While causality appear in some definitions, it is of less significance to our work since we are not primarily interested in linking events by cause/effect relationships. In [129], the authors proposed a study to compare existing semantic models that attempt to represent events in a structured format.
They propose an interoperable model to represent intersubjective “consensus reality” over all event definitions. Based on this model, we define an event in terms of the four Ws questions as follows:
1. What happened: represented by a set of descriptive terms.
2. Where it happened: associates an event with any number of places.
3. When it happened: associates an event with a specific time or period of time.
4. Who was involved: distinguishes between people having “active” or “passive” role.
Social Websites
Events on the Web exist in two different types: unstructured and structured. On the one hand, unstructured events are mostly represented in form of natural language phrases which require complex parsing and extraction mechanisms. On the other hand, structured events are represented in a well-defined structure that may differ from one site to another. Currently, there exists a large variety of websites that host structured information about past and upcoming events, some of which may display media. In this thesis, we focus on structured events as provided by some popular event websites. In the following, we provide an overview about these sites as well as the platforms which host related media.
Exploratory User Study
The initial motivation behind this thesis lies in an exploratory user study conducted by Fialho et al. [40]. The goal is to understand the event-related activities (e.g., searching, attending, sharing) and to collect insights about existingWeb-based technologies. This study consists of a user survey completed by 28 participants and two focus-group sessions (10 and 25 participants). The questions were elaborated to assess the perceived benefits and drawbacks of using: event directories, media directories, social networks, and a merger of these services. In the following, we describe some highlights of this survey:
Finding and attending an event: Participants reported to discover events mainly through invitations, recommendations, friends’ posts or some traditional media (e.g., news articles, ads, etc). They also refer to previously attended events or venues to find new events, and they use search engines particularly when they knew what to look for. Moreover, it is found that decision about attending an event seems to prioritize some 6. http://www.flickr.com significant constraints such as time, location and price. Social information about which friends will attend an event has also an important role in decision making. Other additional details appear to have slight influence like the case of subjective factors (type, topic, performer). To share their experiences, participants tend to use media directories and social networks by posting comments, photos and few videos.
Use of social directories: According to participants, an event directory or website is the best source to provide a general overview of an event context within a single channel. It also enables a user-friendly event exploration from various views (what, when, where) along with other features (e.g., tickets, comments). However, it appears that the information perceived are often incomplete and insufficient for decision support (lack of media and geographic map). To overcome this issue, media directories have been considered as one valuable outlet that better illustrate the event context based on visual information. Similarly, social networks seem to be precious channel to enrich the event context by some features such as attendance, opinions and invitations. Besides, some other functionalities have been mentioned to be desirable for reducing the information overload. For example, it is of great importance to support recommendation of events based on friend’s attendance and user interests. Another functionality is to better visualize events by improving search features (e.g., geographic map) and enriching descriptions (e.g., price, attendance).
Recapitulation: To sum up, lack of coverage of event directories and frustration of being locked in isolated sites are the recurrent issues perceived during the study. Participants recognized that there is a need to access several social channels to gather information. One participant reported “I don’t like always having to go from one site to another to find out things about the event”. Overall, users advocate the need for a single source to explore events, not by creating another information source, but by centralizing all available information leading to broader coverage. In addition, they highlight the role of photos and videos to provide powerful means of identifying several event characteristics. Media is thus useful to convey the experience and to support decision making. Nevertheless, a common concern of information overload suggests that the environment should avoid cluttered information and provide advanced browsing and personalization mechanisms. Motivated by this study, we decided to build a platform based on the Semantic Web technologies in order to integrate information spread in many silos, and to improve event discovery and content personalization.
Table of contents :
Abstract
Acknowledgements
Contents
List of Figures
List of Tables
Glossary
1 Introduction
1.1 Context and Motivation
1.1.1 Data Reconciliation
1.1.2 Personalization Techniques
1.2 Thesis Contributions
1.3 Thesis Outline
2 Background
2.1 Events on the Web
2.1.1 Event Definition and Characterization
2.1.2 Social Websites
2.1.3 Exploratory User Study
2.2 Events in Research
2.3 The Semantic Web
2.3.1 Resource Description Framework (RDF)
2.3.2 RDF Schema
2.3.3 Ontology Vocabulary
2.3.4 Linked Open Data
2.4 Evaluation Metrics
2.5 Conclusion
3 Data Aggregation and Modeling
3.1 Data Aggregation
3.1.1 The Notion of the Web Service
3.1.2 REST-based Scraping Framework
3.1.3 Explicit Linkage of Events with Media
3.1.4 Real-time Scraping
3.2 Web Dashboard
3.3 Semantic Data Modeling
3.3.1 Event Modeling: the LODE Ontology
3.3.2 Media Modeling
3.4 EventMedia
3.5 Conclusion
4 Event-centric Data Reconciliation
4.1 Domain-independent Matching of Events
4.1.1 Challenges and Related Work
4.1.2 Similarity Metrics
4.1.3 Domain-independent Matching Approach
4.1.4 Real-time Matching
4.1.5 Experiments and Results
4.2 Matching Semantic Events with Microposts
4.2.1 Challenges and Related Work
4.2.2 RDF Representation of Microposts
4.2.3 NER-based Matching Approach
4.2.4 Named Entity Recognition in Microposts
4.2.5 Use Case and Results: ISWC Conference
4.3 Conclusion
5 Consuming Event-centric Linked Data
5.1 EventMedia Application
5.1.1 UI Challenges
5.1.2 Elda: Epimorphics Linked Data API
5.1.3 EventMedia UI
5.1.4 Discussion
5.2 Enhanced Facebook Event Application
5.3 Confomaton: Conference Enhancer with Social Media
5.3.1 Confomaton Architecture
5.3.2 Confomaton UI
5.3.3 Discussion
5.4 Behavioral Aspects and User Profiling
5.4.1 Behavioral Analysis using Linked Data
5.4.2 User Profiling using Linked Data
5.5 Conclusion
6 Hybrid Event Recommendation
6.1 Challenges and Related Work
6.2 Content-based Recommendation using Linked Data
6.2.1 Items Similarity in Linked Data
6.2.2 Similarity-based Interpolation
6.3 Event Recommendation
6.3.1 Content-based Recommendation
6.3.2 User Interest Modeling
6.3.3 Collaborative Filtering
6.3.4 Hybrid Recommendation
6.4 Experiments and Evaluation
6.4.1 Real-world Dataset
6.4.2 Learning Rank Weights
6.4.3 Experiments
6.5 Conclusion
7 Overlapping Semantic Community Detection in Event-based Social Network
7.1 Challenges and Related Work
7.2 EBSN: Event-based Social Network
7.2.1 EBSN Definition
7.2.2 Spatial Aspect of Social Interactions
7.2.3 User Participation
7.3 SMM-based Community Detection
7.3.1 Graph Modeling
7.3.2 The SMM Approach
7.4 Experiments and Results
7.4.1 Experimental Datasets
7.4.2 Topic Modeling
7.4.3 Performance Metrics
7.4.4 Evaluation
7.5 Conclusion
8 Conclusions and Future Perspectives
8.1 Achievements
8.2 Perspectives
III Appendix
A List of Publications
A.1 Journals
A.2 Conferences and Workshops
A.3 Archived Technical Reports
B Extended Background
B.1 String Similarity
B.1.1 Token-based Functions
B.1.2 Character-based Functions
B.1.3 Hybrid Functions
B.2 Optimization Techniques
B.2.1 Genetic Algorithm (GA)
B.2.2 Particle Swarm Optimization (PSO)
B.3 Recommender Systems
B.3.1 Content-based Recommendation
B.3.2 Collaborative Filtering Recommendation
C Résumé en Français
C.1 Introduction
C.2 Contexte de la thèse
C.3 Collecte et sémantisation des données événmentielles
C.3.1 Collecte et agrégation des données
C.3.2 Modélisation sémantique
C.3.3 EventMedia : un jeu de données événementiel
C.4 Interconnexion de données événementielles
C.4.1 Approche de réconciliation
C.4.2 Évaluation de performance
C.4.3 Réconciliation en temps réel
C.5 Enrichissement d’événements par des micro-messages
C.5.1 Structuration des micro-messages
C.5.2 Lier des micro-messages aux événements
C.5.3 Cas d’usage et évaluation
C.6 Approche hybride pour la recommandation d’événements
C.6.1 Recommandation thématique dans le Web sémantique
C.6.2 Recommandation thématique d’événements
C.6.3 Recommandation basée sur le filtrage collaboratif
C.6.4 Recommandation hybride
C.6.5 Expérimentations et évaluation
C.7 Détection de communautés sémantiques et recouvrantes
C.7.1 Similarité d’événements dans l’espace latent
C.7.2 Clustering hiérarchique et formation de communautés
C.7.3 Évaluation de la qualité des communautés
C.8 Conclusion
Bibliography
