Semantic Web 0 (2019) 1 1 IOS Press 1 1 2 2 3 3 4 A More Decentralized Vision for Linked Data 4 5 5 Axel Polleres a,b,*, Maulik Rajendra Kamdar c, Javier David Fernández a,b, Tania Tudorache c, 6 c 6 7 Mark Alan Musen 7 a 8 Institute for Information Business, Vienna University of Economics and Business, Vienna, Austria 8 b 9 Complexity Science Hub Vienna, Vienna, Austria 9 10 E-mails: [email protected], [email protected] 10 c 11 Center for Biomedical Informatics Research, Stanford University, CA, USA 11 12 E-mails: [email protected], [email protected], [email protected] 12 13 13 14 Editors: Pascal Hitzler, Kansas State University, Manhattan, KS, USA; Krzysztof Janowicz, University of California, Santa Barbara, USA 14 15 Solicited reviews: Anna Lisa Gentile, IBM Research, CA, USA; Jens Lehmann, University of Bonn, Germany 15 16 16 17 17 18 18 19 19 20 Abstract. In this deliberately provocative position paper, we claim that more than ten years into Linked Data there are still 20 21 (too?) many unresolved challenges towards arriving at a truly machine-readable and decentralized Web of data. We take a deeper 21 1 22 look at key challenges in usage and adoption of Linked Data from the ever-present “LOD cloud” diagram. Herein, we try to 22 23 highlight and exemplify both key technical and non-technical challenges to the success of LOD, and we outline potential solution 23 strategies. We hope that this paper will serve as a discussion basis for a fresh start towards more actionable, truly decentralized 24 24 Linked Data, and as a call to the community to join forces. 25 25 26 Keywords: Linked Data, Decentralization, Semantic Web 26 27 27 28 28 29 29 30 1. Decentralization Myths on the Semantic Web appear not to have found major adoption at a level 30 31 comparable with these siloed sites.3 31 32 Let us start with a rant, arguing that the Semantic – We envisioned a decentralized network of on- 32 33 33 Web may well be considered a story of failed promises tologies on the Web that would enable smart 34 34 with regards to decentralization: agents to seamlessly talk to each other, and that 35 35 – We had hopes (as a community) to revolutionize would enable easy integration of data by follow- 36 ing the guiding principles of ontology engineer- 36 37 Social Networks in a way that every data subject 37 owns and controls their social network data in de- ing and Gruber’s often cited vision of ontolo- 38 gies as shared conceptualizations [25].4 While 38 centralized FOAF [13] files published in their 39 there are indeed certain areas in which ontolo- 39 40 personal Web space – we got siloed, centralized 40 41 social networks (Facebook, LinkedIn). Attempts 41 42 to re-decentralize the Social Web, for instance, 3While ActivePub has been picked up by several imple- 42 43 through the work of the W3C Social Web WG2 mentations, cf. https://en.wikipedia.org/wiki/ActivityPub# 43 Implementations, reversing the network effects that have drawn a 44 44 critical mass of users to these siloed sites seems still far away. 45 4Or, as Dan Brickley, one of the inventors of FOAF stated slightly 45 46 *Corresponding author. E-mail: [email protected]. sarcastically in personal communication: “we took one useful fea- 46 47 1The Linking Open Data cloud diagram, available at http:// ture of RDF/RDFS (fine grained vocabulary composition) and ele- 47 48 lod-cloud.net/, which has been regularly updated since 2007 by An- vated it to a cult-like holy law, to the extent that anyone who created 48 dreas Abele, John P. McCrae, Paul Buitelaar, Anja Jentzsch and a useful RDF vocabulary and wanted to keep improving it, found 49 49 Richard Cyganiak, with its latest version having been created in themselves pushed instead into combining it with dozens of other 50 March 2019 [1]. half-finished, poorly documented efforts that weren’t really designed 50 51 2https://www.w3.org/wiki/Socialwg to fit together nicely.” 51 1570-0844/19/$35.00 c 2019 – IOS Press and the authors. All rights reserved 2 A. Polleres et al. / A More Decentralized Vision for Linked Data 1 gies are used to share conceptualizations of a do- known examples of knowledge graphs (Google’s, 1 2 main, mostly these are insular efforts that do the Bing’s, and Yahoo’s knowledge graphs as well as 2 3 job well for a particular community. However, their open pendants DBpedia and Wikidata) are 3 4 on Web scale, ontology and vocabulary reuse is NOT decentralized. 4 5 still limited or tied to inherent challenges, see So, here we are after more than ten years. However, 5 6 also [26, 27]. Rather, we see a rise of main cen- there is one lighthouse project that clearly has imple- 6 7 tral schema (schema.org), and fast-growing com- mented the vision of a decentralized Semantic Web. 7 8 munity projects like Wikidata [50] refusing to buy This single project that we, as a community, hinge 8 9 into the need for re-using terms from other on- upon and tend to accept as a clear success to wipe away 9 10 tologies.5 all the failed promises mentioned above is: Linked 10 11 – We put a lot of effort into formal semantics and Data [9]. The promise to be able to publish structured 11 12 clean axiomatization of those ontologies – we data in a truly decentralized fashion, with a couple of 12 13 got logical inconsistency.6 Whereas, serious at- simple principles to enable the automatic retrieval and 13 14 tempts to apply such reasoning about Web Data integration by just “following your nose” (i.e., derefer- 14 15 in the wild have either had to restrict themselves encing HTTP links). This principle is the most power- 15 16 to lightweight ontologies or have not been fur- ful promise that filled the community with new enthu- 16 17 ther developed in the past five years, with (a) siasm through the so-called “LOD cloud”, cf. Fig. 1. If 17 18 the semantics of OWL [45] and even parts of we measure the number of datasets published accord- 18 19 RDF(S) [12] turning out to be too hard to grasp ing to the four Linked Data principles [6] and that link 19 20 for normal Web users and developers to survive to each other, we find evidence of growth and prosper- 20 21 in the World Wild Web [24, 36]; and (b) the DL ity (cf. Fig. 2), and hope to finally make the vision of a 21 22 community mostly having turned their back to se- decentralized Web of data come true. Meanwhile, in- 22 23 riously taking the challenge of decentralized ap- deed this “cloud” contains over 1,184 datasets, which 23 24 plications at Web scale. should be considered good news. 24 25 – Berners-Lee et al. in their original Semantic Web However, as we will discuss in the present paper, 25 26 article [7] promised Web-scale automation: au- there are still serious barriers to consume and use 26 27 tomated calendar synchronisation, personalised Linked Data from the “cloud”, wherein we have to 27 28 health care assistance, home automation – some question the usefulness of the current LOD cloud. 28 29 of these applications are a reality now (Amazon Thus, we would like to take a step back and assess 29 30 Alexa’s home control, or Google’s and Apple’s the situation. We call for a more principled and, in our 30 31 widely used services), but rather than relying on opinion, more useful restart and for more collaboration 31 32 a decentralized Semantic Web, use single compa- and decentralization in the community itself. 32 33 nies’ curated knowledge bases – also now called Along these lines, in the remainder of this paper, we 33 34 “Knowledge Graphs” – that enhance these com- start with some background on the genesis of the cur- 34 35 panies’ services’ backend systems. rent LOD cloud in Section 2. We will then highlight 35 36 – More specifically, we see knowledge graphs five perceived main challenges we deem important to 36 37 evolve and embrace them as a success story of be addressed to make Linked Data more usable and, 37 38 the Semantic Web. Yet a good definition of what therefore, useful. These challenges will be presented – 38 39 a Knowledge Graph is and what differentiates it by examples and discussing their implications – in the 39 40 from an “ontology” is still to be provided – apart course of Section 3. Finally, we conclude with a call to 40 41 from the single distinguishing feature that all collaboratively and openly address these challenges as 41 42 a community in order to (re-)decentralize the Semantic 42 43 Web again in Section 4. 43 5The main reason for Wikidata not to prescribe existing vocab- 44 44 ularies was to leave the community freedom to link and use what 45 they deem useful within one consistent scheme/namespace: one of 45 46 the reasons was to avoid the needed buy-in to existing ontologies the 2. Background: The genesis of the LOD Cloud 46 47 popularity of which or agreement about could shift over time. There- 47 48 fore, they “left it to the community whether to choose stronger se- The creation of a complete Web index is a never- 48 mantics (e.g., OWL) or weaker semantics (e.g., SKOS [37]) or not” 49 49 (personal communication Denny Vrandeci˘ c).´ ending story.