In this demo we present WikipEvent, an exploratory system that captures and visualises continuously evolving complex event structures, along with the involved entities. The framework facilitates entitycentric and event-centric search, presented via a user-friendly interface and supported by temporal snippets from corresponding Wikipedia page versions. The events detected and extracted using di erent mechanisms are exposed as freely available Linked Data for further reuse.
Exploratory search systems help users to search, navigate, and discover new facts
and relationships. We detect and extract events from di erent sources and merge
these events into a unique event repository. Each event is described primarily in
terms of (i) a list of entities (Wikipedia pages) participating in the event, (ii) a
textual description of the event, (iii) start and end dates of the event, and (iv)
the extraction method used to obtain the event. We further classify entities as
people, organizations, artifacts, and locations by exploiting the class hierarchy
de ned in YAGO2 [
In this demo, we showcase two popular usecase scenarios. First, we show
that WikipEvent can be used in order to explore the events in which particular
entities have been involved. Secondly, we show the suitability of WikipEvent to
explore the evolution of entities based on the events these are involved in. Both
these scenarios can be additionally surveyed based on the temporal dimension.
In addition to the events that are presented using a timeline1, we adopt a
versioning approach introduced previously[
WikipEvent facilitates the understanding of events and their related
entities on a temporal basis. Instead of exploration of isolated knowledge bases,
WikipEvent takes advantage of the complimentary nature of sources
contributing to the underlying event repository.
1 Interface: http://wikipeventdemo.l3s.uni-hannover.de/WikiEventEntity/
We extract events from three sources; Wikipedia Current Events portal[
The resulting repository contains more than 2:6 million events, extracted from the di erent sources as shown in Table 1. The contribution from the sources is skewed due to two reasons. Firstly, they cover very di erent time periods: YAGO2 contains events and temporal facts spanning over thousands of years, Current Events captures events since 2001 only, and, for performance reasons, the Co-References method has been restricted to edits in 2011. In addition, Co-References exclusively analyses entities of type politician, while YAGO2 and Current Events contain almost all the Wikipedia pages. To facilitate comparisons across entities occurring in all three sources, in the rest of this section we will consider only those events that occurred in 2011 and involved politicians.
Source Total Politicians Politicians 2011
All 2,629,740 50,168 1,401
YAGO2 2,578,547 42,399 360 Current Events 50,951 7,527 799
Co-Reference 242 242 242 Table 1: Number of events within the event repository, split by di erent sources.
Co-References is able to detect events with di erent duration and granularity (from a wrestling match to the Egypt Revolution). YAGO2 mostly contains high-level and well-known events represented through temporal facts regarding entities, often lacking textual descriptions. Current Events portal contains daily events which have a self explanatory textual description and are reliable, thanks to the high level of control within Wikipedia. The complimentary nature of the di erent sources in terms of complexity (number of participants), duration, and granularity of events is evident. Due to these reasons, also the schemas used to represent events across di erent sources are distinct, yet overlapping. While certain properties (for instance for time points) are overlapping yet follow di erent conventions, we lifted the events from di erent sources into a uni ed dataset following Linked Data principles and deploying a joint RDF schema.
Exposing Events Data as RDF.
We have exposed the WikipEvent events data through a public Linked Data
WikipEvent: Temporal Event Data for the Semantic Web interface using D2R Server3, enabling URI dereferencing via content negotiation and providing a public SPARQL endpoint. This data can be accessed and queried via http://wikipevent.l3s.uni-hannover.de/ and using our SPARQL endpoint4.
We represent events through established event RDF vocabularies, to facilitate reuse, interpretation and linking of our data through third parties. In particular, we use properties from the LODE ontology5 to map di erent properties pertaining to events in our dataset, for instance, property lode:atPlace is used as predicate for stating venues of events. Figure 1 presents an example of an event and its entailing properties in our event repository. 3
WikipEvent can help students, scholars, historians, or journalists by facilitating
temporal search focussed on either the entities at hand or the events. Thus, the
WikipEvent framework can be used to satisfy two primary scenarios -
entitybased and event-based information needs. Results are presented through a
userfriendly interface, that supports faceted search, query-path tracing, query
completion, temporal settings, and is abridged with events' sources as well as lters
for related entities. The underlying versioning system [
Entity-based Search. Users may want to learn about entities of interest with respect to their temporal participation in events. For example, journalists might aim at studying political a liations of individuals, and the campaigns 3 http://d2rq.org/ 4 http://wikipevent.l3s.uni-hannover.de/snorql/ 5 http://linkedevents.org/ontology/ they participated in. Existing systems however, make it cumbersome to easily access this information. The WikipEvent interface overcomes this challenge by presenting a timeline of events that an entity is involved in. Additional lters for relevant entities help users to navigate through the retrieved results. Figure 2 presents an example of an entity-based search on WikipEvent.
Event-based Search. Historic events are a subject of interest to a wide array of people, ranging from students to archivists. WikipEvent facilitates a free-text search for events. Events relevant to a given query are presented in the form of a continuous timeline (Figure 3), while highlighting the entities involved. WikipEvent enables users to sift through event related information on a temporal basis, in order to learn more about the events and the participating entities.
To gain a complete understanding of the WikipEvent framework, we point the reader to a demo video 6.