Bootstrapping the Publication of Linked Data Streams

Abstract

Data Velocity reached the Web. New protocols and APIs (e.g. WebSockets, and EventSource) are emerging, and the Web of Data is also evolving to tame Velocity without neglecting Variety. The RDF Stream Processing (RSP) community is actively addressing these challenges by proposing continuous query languages and working prototypes. Nevertheless, the problem of Streaming Linked Data publication is still an open challenge. In this paper, we present the rst attempt to tackle this challenge by introducing a set of guidelines to publish streaming linked data by reusing existing resources such as TripleWave, R2RML/RML, VoCaLS, and RSP-QL. We design a publication life-cycle that follows the W3C best practices. Besides, we present an example for publishing the Global Database of Events, Language and Tone (GDELT) as a Streaming Linked Data resource. We open-sourced the code of our resource and made it available for public use.

Keywords

• RDF Streams
• Streaming Linked Data
• RDF Stream Processing
• Stream Reasoning

-

1 A new generation of Web applications shows the need to process data reactively and continuously [ 3 ]. This challenge, known as Data Velocity in the Big Data domain, is now critical for the Web too [ 8 ]. The RDF Stream Processing (RSP) tries to process heterogeneous data streams that come from complex domains on-the- y. Indeed, the RSP community2 proposed the a data model, (i.e., RDF Streams) a query model (i.e., RSP-QL) and several resources for processing, and publishing Web streams [ 2, 4, 9 ].

The Global Database of Events, Language & Tone (GDELT) project3 is a family of vast and heterogeneous Web Streams that is considered as the largest open-access Spatio-temporal archive for human society. Its Global Knowledge Graph spans more than 215 years, and connects people, organizations, locations, all over the world. In particular, GDELT complex from a complex domain. It 1 Copyright 2019 for this paper by its authors. Use permitted under Creative Commons License Attribution 4.0 International (CC BY 4.0). 2 https://www.w3.org/community/rsp/ 3 www.gdeltproject.org captures themes, images, and emotions into a single holistic global network. GDELT data can be accessed via Google Big Query or via a number of APIs that run pre-de ned analyses. Therefore, interested researches are forced to either run their analysis via the pay-per-use service or stick with the provided APIs.

GDELT Project provisions, every 15 minutes, many TSV Web Streams: the Event Stream makes use of the dyadic CAMEO format, describing the two actors participating in each event, and the action they perform; the Mention Stream records every mention of an event in the Event Stream over time, along with the timestamp the article was published; the GKG Stream connects people, organizations, locations, events, and more across the planet. In this paper, we present our approach for consuming and analyzing the shared GDELT TSV Streams. In particular, we provide a set of methodological guidelines that explain how to employ existing RSP resources { RML, TripleWave, VoCaLS { to publish and consume Streaming Linked Data. The guidelines and an extended version of the documents included in this paper are available at http://gdelt.stream. 2

Publishing GDELT Streams as Streaming Linked Data In this section, we present our methodological guidelines to publish GDELT, an example of RDF streams, as Streaming Linked Data. Figure 1 depicts the steps of our publication life-cycle that has been designed following the W3C Best practices4 and the guidelines presented by Hyland et al. [ 7 ].

The Step (0) Name Things with (HTTP) URIs aims of designing (HTTP) URIs that identify the relevant resources according to the Linked Data principles [ 7 ], and W3C best practices for URI design5. We designed URIs with the following base http://gdelt.stream/, distinguishing between base vocabulary (vocab), CAMEO (onto/cameo), instances (ist ) and time instants (time).

The Step (1) Model the Streams Domain aims of (i) Understanding and capturing the domain knowledge into an ontological model, and reuse existing authoritative vocabularies. (ii) Identifying related resources, and (iii) Formulating canonical information needs [ 7 ]. This process requires collecting and reviewing applications, documentations and analyzing sample data. The data of GDELT is streamed from a multitude of news sources. The extraction process follows Natural Language Processing techniques and makes use of the Con ict and Mediation Event Observations (CAMEO) Ontology [ 6 ].

We collected all the information regarding the schemas of the streams, and we studied the CAMEO documentations. Then we designed a set of OWL2 ontologies describing the GDELT domain. The CAMEO ontology is a coding scheme

4 https://www.w3.org/TR/ld-bp/ 5 https://www.w3.org/TR/cooluris/#cooluris

Bootstrapping the Publication of Linked Data Streams <> a vocals:StreamDescriptor ; dcat:dataset :eventstream 1

. :eventstream a vocals:Stream ; dcat:title "GDELT Event Stream"^^xsd:string ; dcat:publisher <http://www.streamreasoning.org> ; dcat:description "GDELT Events Stream"^^xsd:string ; vocals:windowType vocals:logicalTumbling ; vocals:windowSize "PT15M"^^xsd:duration ; vocals:hasEndpoint [ a vocals:StreamEndpoint ; dcat:license <https://cc.org/licenses/by-nc/4.0/> ; dcat:format frmt:JSON-LD; dcat:accessURL "ws://examples:8080/events" ] . which is designed for the study of third-party mediation in international disputes. It contains a hierarchical coding-scheme for dealing with sub-state actors, event types, and an extensive taxonomy for religious groups and ethnic groups. For CAMEO, we model event and actors types, into a comprehensive hierarchy.

The Step (2) Describe the Stream aims for providing human-readable and machine-readable representations of the streams [ 7 ] as commonly done for datasets [ 1 ]. To this extent, Tommasini et al. proposed a vocabulary based on DCAT6 named VoCaLS [ 10 ].

We used VoCaLS to describe GDELT streams. Listing 1.1 shows an example of description for the GDELT Event Stream. In VoCaLS, a Web Stream is represented using vocals:Stream, i.e., an unbounded sequence of data items that might be accessible on the Web. A Listing 1.1: VoCaLS GDELT Stream vocals:StreamDescriptor, i.e, description. a HTTP-accessible document that contains stream metadata. Finally, the stream content can be consumed via a vocals:StreamEndpoint, that refer to actual sources using dcat:accessURL property. GDELT does not use a license format, thus we include a license that is compliant with the terms of use. We also linked ontologies and mappings using rdfs:seeAlso. Since the stream is published regularly as 15 minutes batch, we include metadata about the rate, i.e., at lines 6-7 vocals:windowType and vocals:windowSize.

The Step (3) Convert to RDF Stream aims of enabling RDF data provisioning so that data could be enriched with domain knowledge [ 7 ] from Step (1). The conversion to RDF can occur either automatically or via expert modeling [ 7 ]. Approaches based on mapping languages Listing 1.2: Subset of GDELT RML Mapping. decouple conversion and modelling, and support the conversion process using a formal language like R2RML 7 for relational data, and RML for logical sources [ 5 ] sharing document-based formats (e.g., JSON), or semi-structured formats (e.g., CSV). Similarly, Web streams are often not shared as RDF Streams and needs to be converted. In practice, GDELT content is not natively RDF. Therefore, we need to set up a conversion mechanism. For data conversion, we followed the RML approach where CSV data is converted using its mappings. Listing 1.2 shows a sample of <GEM> a rr:TriplesMap ; rml:logicalSource <source> ; 1 rr:subjectMap [ 2 rr:template "http://gdelt.stream/ist/{GLOBALEVENTID}"; 3 rr:class gdelt:Event; 4 rr:graphMap [ 5 rr:template "http://gdelt.stream/time/{DATEADDED}"]]; 6 rr:predicateObjectMap [ rr:predicate rdf:type; 7 rr:objectMap [ 8 rr:template "http://gdelt.stream/cameo/{EventCode}" 9

;]]; rr:predicateObjectMap [ rr:predicate gdelt:actor; rr:objectMap [ rr:parentTriplesMap <Actor1TM> ]];...

6 https://www.w3.org/TR/vocab-dcat/ 7 https://www.w3.org/2001/sw/rdb2rdf/r2rml/

Tommasini et al. such R2RML mapping. Notably, we used rr:class to assign the gdelt:Event type to the data at line 4. Moreover, we assign the CAMEO type using rdf:type.

The Step (4) Publish The RDF Stream aims for serving the data to the audience of interest. Datasets { opportunely described with contextual vocabularies [ 1 ] { are either added to the Linked Open Data (LOD) Cloud, shared using REST APIs, or exposed via SPARQL endpoints. Critical aspects of this step are licensing, audit, and access control [ 7 ].

The unbounded nature of Streaming Linked Data makes it impossible to directly add a data stream to the LOD Cloud [ 10, 9 ]. However, using an RSP Engine that focuses on RDF Stream provisioning like TripleWave [ 9 ], one can publish a stream describing using VoCaLS. TripleWave relies on Barbieri et al's vision [ 2 ] of serving a "static" named graph called S-Graph via REST API, to identify and describe the streams. Moreover, they envisioned that the stream elements of an RDF Graph are called I-Graphs. We shared the VoCALS description les like the one in Listing 1.1 as S-GRAPH via REST APIs. Since the concept of event is rst-class, we opted for a graph-based stream data model.

Conclusion This paper presents a rst attempt to publish streaming Linked Data following W3C's best practices [ 7 ]. Some real-world Web streams from the GDELT project are available as RDF Stream, described using VoCaLS, and accessible at http: //gdelt.stream. Future work include publishing other Web streams towards a catalog of Linked Streams.

References

1. Alexander , K. , Cyganiak , R. , Hausenblas , M. , Zhao , J. : Describing linked datasets . In: Proceedings of the WWW2009 Workshop on Linked Data on the Web, LDOW 2009 , Madrid, Spain, April 20 , 2009 . ( 2009 )

2. Barbieri , D.F. , Della Valle , E.: A proposal for publishing data streams as linked data - A position paper . In: Proceedings of the WWW2010 Workshop on Linked Data on the Web, LDOW 2010 , Raleigh , USA, April 27 , 2010 ( 2010 )

3. Della Valle , E. , Dell'Aglio , D. , Margara , A. : Taming velocity and variety simultaneously in big data with stream reasoning: tutorial . In: DEBS ( 2016 )

4. DellAglio , D. , Della Valle , E., van Harmelen , F. , Bernstein , A. : Stream reasoning: A survey and outlook . Data Science 1 ( 1-2 ), 59 { 83 ( 2017 )

5. Dimou , A. , Sande , M.V. , Slepicka , J. , Szekely , P.A. , Mannens , E. , Knoblock , C.A. , de Walle , R.V.: Mapping hierarchical sources into RDF using the RML mapping language . In: International Conference on Semantic Computing ( 2014 )

6. Gerner , D.J. , Schrodt , P.A. , Yilmaz , O. , Abu-Jabr , R. : Con ict and mediation event observations (cameo) . International Studies Association , New Orleans ( 2002 )

7. Hyland , B. , Wood , D. : The joy of data-a cookbook for publishing linked government data on the web . In: Linking government data , pp. 3 { 26 . Springer ( 2011 )

8. Margara , A. , Urbani , J., van Harmelen , F. , Bal , H.E. : Streaming the web: Reasoning over dynamic data . J. Web Sem . 25 , 24 { 44 ( 2014 )

9. Mauri , A. , Calbimonte , J. , Dell'Aglio , D. , Balduini , M. , Brambilla , M. , Della Valle , E. , Aberer , K. : Triplewave: Spreading RDF streams on the web . In: ISWC ( 2016 )

10. Tommasini , R. , Sedira , Y.A. , Dell'Aglio , D. , Balduini , M. , Ali , M.I. , Phuoc , D.L. , Della Valle , E. , Calbimonte , J.: Vocals: Vocabulary and catalog of linked streams