InPhilip Howard – ResearchBrief Director, Information Management www..com Daniel Howard – Senior Researcher 116 W 23rd Street, Suite 500 New York, NY 10011, USA Tel: +1 929 239 0659 Ontotext GraphDB Email: [email protected]

The company CREATIVITY SCALE Ontotext was one of the first vendors into this space, having been originally founded in 2000 (in Bulgaria) to investigate semantic technologies. Head office remains in Sofia but the company also offices in London and New York. It is part of the larger Sirma Group. In addition to its commercial clients – the company’s main focus – it is also worth noting that, over the years, Ontotext has won funding from the EU for upwards of 30 different projects, mostly around the semantic web, and Open and

Linked Data. Most recently (May 2018) it was EXECUTION TECHNOLOGY awarded funding for the Intelligent Matching and Linking of Company Data project (CIMA), which The image in this Mutable Quadrant is derived from 13 high level metrics, the more the image covers a section the better. will run until May 2020. CIMA will use machine Execution metrics relate to the company, Technology to the learning, semantic modeling and data integration product, Creativity to both technical and business innovation and Scale covers the potential business and market impact. as well as logical inference and validation to make a company and related data (persons, locations, industry taxonomies, technology fields) What is it? better harmonised, integrated, interlinked and GraphDB (previously known as OWLIM) is a native easier to use. RDF with dynamic indexing that integrates with various search technologies, as well as text mining. Unlike most other vendors in this space the company has developed specific solutions for various industry sectors, including publishing, market intelligence, financial services, life sciences and healthcare. There is a developer edition (GraphDB Free Edition) that is free but limited by the number of concurrent users. Otherwise the options are the Standard Edition, the Enterprise Edition or GraphDB Cloud, which is a database as a service offering running on AWS. The Ontotext Platform extends GraphDB with text mining capabilities. Target environments include reference and master data management, metadata-based content management, the development of knowledge graphs Figure 1 – The parsing of data to create triples via semantic data integration, information and relationship discovery, and content management Ontotext partners with various leading IT solutions that involve text analytics on top of big services providers, a number of which embed knowledge graphs. GraphDB in their AI platforms. Such partners include InfoSys, Atos Origin, NTT Docomo (via What does it do? Everis) and Fujitsu. As far as features go, GraphDB includes various capabilities that extend beyond the database, notably ontology visualisation and connectors to a

© Bloor 2019 Analytics Language Ease of Use Operations Features Performance Integration Scalability

“ We are happy to have GraphDB as a central metadata store in SciGraph. This semantic variety of third party environments Why should you care? (mostly search engines such as by Ontotext is a Graph technology is all about getting an reliable product that offers fast Solr, Lucene and ). understanding of relationships – typically multi- data loading, efficient data Internally, GraphDB’s inference updates and it’s easy hop relationships – that span large datasets. This engine employs forward chaining to use. is difficult for relational technology to achieve and the company has patented SciGraph” because of the number of self-joins that are often functions with respect to materialised required, and which cannot be optimised within inferences, which retract inferred a relational environment. Further, at least vis a statements, which are no longer supported after vis Ontotext, relational technology is not good at deletion. In recent releases (the current version understanding relationships that are based on is 8.7) the company has focused on cluster semantic equivalence, or management and performance (including load relatedness, since that performance), enhanced security (including LDAP concept does not support), visualisation capabilities and increased Everything exist within the “ support for workflow processes to enable the relational space. with the migration has loading of structured data into GraphDB. As an Although been going well as far as I example of this, Figure 1 shows a screenshot Ontotext has can see. We have had very little illustrating the parsing of, in this case, airport data issues with any of the servers, a very definite so that that can be converted into triples. which is great and I’m hoping it focus on text, stays that way. You guys have content and been very helpful and I’m related areas appreciative of that. Elsevier ” – and most of its clients are in some way involved in publishing or media – this does not preclude the use of GraphDB for more general-purpose operational and hybrid operational/analytic use cases. Nevertheless, the company offers a one-stop shop for both the database and text mining, and the strength of this offering – the way that it works with enterprise knowledge graphs – is a significant differentiator for the company.

The Bottom Line Figure 2 – Visualisation of a query result Most RDF graph vendors have opted to either Once parsed, you can cleanse the resulting ape their property graph competitors or focus triples using OntoRefine (which in turn leverages specifically on complex analytics. Ontotext has GREL [general refine expression language]) never done this: it has significant expertise in and then, if necessary, modify the automatically semantics, search and text mining, and is a market generated schema. Once this process is complete leader in this context. It is noteworthy, from a the results can be reconciled, queried in SPARQL and performance perspective, that GraphDB is a imported. Figure 2 shows a related visualisation. native graph database rather than a multi-model Ontotext provides the sort of database database running on some other platform. administration tools and similar features that you would normally expect. Also notable are support for geo-spatial constraints, as well as semantic similarity based on graph embedding. FOR FURTHER INFORMATION AND RESEARCH CLICK HERE

© Bloor 2019