| A | B | C | D | E | F | G | H | I | J | K | L | M | N | O | P | Q | R | S | T | U | V | W | X | Y | Z | AA | AB | AC | |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
1 | ran | SID | Submitted Time | Completed Time | Modified Time | Draft | IP Address | UID | Username | Talk-ID | Title of your presentation | Abstract | Description of your talk | Your Name | Your E-Mail | Gender | Your Organisation | Country | Academic degree or title | Speakers biography | In case you are presenting together with a second person | Real use case (yes, no, possible) | Field / Industry | Comment FB | Comment MB | Comment CD | Comment | Final list | 3Y |
2 | 302 | 19666 | ######## | ######## | 5/31/21 21:31 | 0 | 172.28.5.4 | 0 | IND21-0024 | DALICC AS A Service - A restful architecture for license clearance | Modern IT applications increasingly retrieve, store and process data from a variety of sources. This can raise questions about the compatibility of licenses and the application`s compliance with existing law. Issues of rights clearance are especially relevant in the creation of derivative works compiled from multiple software and data sources. Manual clearance of licenses can be complex and error prone, thus requiring a high degree of costly expert knowledge. To lower these costs and improve the quality of license clearance, we developen the DALICC framework, an open source and open data framework built on semantic web and linked data standards that supports the convenient and cost-effective clearance of licenses in the creation of derivative software and data works. DALICC is capable to process and reason over RDF representation of licenses and thus capable to identify conflicts between licenses. In essence, DALICC helps to determine which information can be shared with whom to what extent under which conditions, thus lowering the costs of rights clearance and stimulating the data economy. | Although the DALICC framework has been around as a proof of concept for three years, DALICC has recently been improved and redesigned towards a restful service architecture allowing for scalability and extensibility towards various application domains. In this talk we will introduce the audience to the idea and motivation behind DALICC and give hands on examples on how semantic web and linked data standards can be applied in the rea of regulatory and legal tech. We will demonstrate the applicability of DALICC for the purpose of license clearance in complex product setups and give an outlook over the extensibility of the framework towards other areas of contract management and legal compliance. Additionally, we will introduce teh audience to a recently launched developer and open education program administerred by the DALICC Association. | Tassilo Pellegrini | tassilo.pellegrini@fhstp.ac.at | male | University of Applied Sciences St. Pölten | Austria | Prof. Dr. | Tassilo Pellegrini is professor of economics at the University of Applied Sciences in St. Pölten, Austria. His research areas are data economics, policy-aware web and artificial intelligence. He is member of the International Network for Information Ethics (INIE), the African Network of Information Ethics (ANIE) and the Deutsche Gesellschaft für Publizistik und Kommunikationswissenschaft (DGPUK). He is co-founder of the Semantic Web Company in Vienna and Conference Chair of the annual I-SEMANTICS conference series founded in 2005. | Dr. Giray Havur, Siemens & Vienna University of Economics, giray.havur@wu.ac.at | Y | Y | Y | Y | TRUE | ||||
5 | 259 | 5546 | 3/31/21 16:51 | 3/31/21 16:51 | 3/31/21 16:51 | 0 | 172.28.5.3 | 0 | IND21-0015 | Graph based reasoning for scaling energy audits to many customers. | In a drive to reduce the carbon footprint of their customers’ Grundfos is revolutionizing how they engage on efficiency savings of their cooling systems. Initial communication with the customers about cooling systems needs to capture complex system information, yet be simple and relevant for each of the many customer profiles to maximize retention. The solution has to work with a wide variety of engineering systems and many stakeholders within the client. Adopting traditional forms-based methods would have resulted in a complex, costly and error prone sales process where the most likely result was customer drop out. We have developed a process built on semantic reasoning that guides the process, asks the next best questions and simplifies the whole process for the users. | We will begin by setting the context and impact that efficiencies in cooling systems needed to execute on CO2 reduction and meet the 13th Sustainable Development Goal 13. We will give a high-level description of a cooling system control service that can both make an impact and is lead to a profitable outcome for both the consumers and suppliers. The industry app (in production) has the goal of answering a simple business question: “will the energy savings method that the Grundfos team is deploying fit the cooling system of interest and yield sufficient energy savings to justify a partnership?”. Normally this question can only be answered by Grundfos experts performing interviews with system owners and facility managers – but the education of such Grundfos experts is a bottleneck for scaling the business. The complexity of understanding the cooling system of interest with the energy saving in mind is sketched out and modelled as a graph. This avoids burying business logic in the application logic, and captures the seed for a digital twin of the cooling system in a machine-readable format. This domain based knowledge graph combined with the answers iteratively given by users are reasoned over using RDFox. This allows the system to determine the next best question to ask. Within the talk we will demonstrate the system in action, show the architecture and user experience made possible through the application of reasoning and RDF supported questioning. Whilst what is presented here will be focused on an engineering system we believe the same guided questioning approach has the potential to be applied to many applications. | Mikkel Haggren Brynildsen | mbrynildsen@grundfos.com | male | Grundfos | Denmark | Chief Data Scientist (PhD) | The speaker has worked in the "Enterprise Data and AI" department of the pump manufacturer Grundfos for 4 years. Grundfos work tirelessly to improve water and energy efficiency for the greater good and believe in making a difference locally and globally. The “Enterprise Data and AI” team work with introducing and maturing AI solutions in Grundfos. Mikkel has a background as a mathematician, has taught at Aalborg university, participates in the “industrial ontologies foundry” (IOF) for several years and has spoken at several conferences on the topic of AI for IoT. | Christian Rasmussen, Head of Technology, Grundfos InnovationLab. Email: crasmussen@grundfos.com | Energy | Y | Y | Y | Y | TRUE | |||
10 | 265 | 12563 | 4/26/21 10:20 | 4/26/21 10:20 | 4/26/21 10:20 | 0 | 172.28.5.4 | 0 | IND21-0009 | Lynx: A FAIR Knowledge Graph Engine for Reference Data Integration & Look up Service | In this presentation, we will show and provide technical details of Lynx, a Knowledge Graph and FAIR-fueled system to enable seamless access across the Roche semantic ecosystem. On the one hand, Lynx exploits machine-readable, FAIR Knowledge Graphs to allow for accessing and combining multiple and disparate reference data systems. On the other hand, Lynx bridges the gap for non-experts with an intuitive and user-friendly way of finding and exploring FAIR data. | Roche, as a leading biopharmaceutical company, has a diverse and distributed ecosystem of platforms to manage reference data standards used at different parts of the organization. These diverse reference data standards include different forms of Knowledge Graphs (ontologies, vocabularies, taxonomies) to capture specifics of the research environment and also to describe how clinical trial data are collected, tabulated, analyzed and finally submitted to regulatory authorities. In the context of the EDIS program, these platforms and vocabularies have been brought together to improve data integration activities and also embraced FAIR to exploit machine-actionability and emphasise data-driven processes Lynx sets in this context and provides answers to two main questions: - How do we enable access to all connected standard vocabularies in a seamless way? - How do we enable the end user to interact with knowledge graphs using business language? Lynx provides access to these potentially connected vocabularies in a seamless way by offering a virtual, interconnected Knowledge Graph. It becomes a common entry for these reference data vocabularies represented as Knowledge Graphs. Lynx pursues to facilitate its use not only for machines via APIs (e.g. using SPARQL) but also in a human friendly way. Non semantic experts must be able to access and query the interconnected, underlying Knowledge Graphs using business language. To do so, Lynx provides a catalog of predefined, parameterizable and federated queries described using natural/business language. This presentation will focus on presenting the problem statement, Lynx concept and benefits, technical implementation and initial user feedback. We will also touch upon upcoming improvements. | Javier D. Fernández | javier_d.fernandez@roche.com | male | F. Hoffmann-La Roche | Switzerland | Dr. Javier D. Fernández, Senior Information Architect at at F. Hoffmann-La Roche, | === Javier D. Fernandez === - Senior Information Architect at Roche in Basel, Switzerland - His work currently focuses on enabling data interoperability and, in general, data FAIRification, facilitating the creation, management and efficient consumption of Knowledge Graphs in the context of clinical data. - Javier holds a joint PhD in Computer Science at the University of Valladolid (Spain) and the University of Chile (Chile) and is author of more than 40 scientific contributions at international journals and conferences. - Before joining Roche, he was a researcher at the Institute of Information Business of Vienna University of Economics and Business (WU Wien), Sapienza Universita di Roma (Italy) and the Ontology Engineering Group (OEG) at Universidad Politécnica de Madrid (Spain). === Nelia Lasierra === - Principal Information Architect at Roche in Basel, Switzerland - In her current job, she is working towards improving meta (data) management activities by driving the implementation of FAIR principles through Knowledge Graphs and Linked Data based solutions. - Nelia has a doctorate from the University of Zaragoza (Spain) in the field of Medical Informatics. - Before joining Roche, she was a researcher at the University of Zaragoza (Spain), University of Innsbruck (Austria) and University for Health Sciences, Medical Informatics and Technology (UMIT) (Austria). | Dr. Nelia Lasierra, Principal Information Architect at F. Hoffmann-La Roche, nelia.lasierra@roche.com | yes | Healthcare | Y | Y | Y | Y | TRUE | ||
13 | 268 | 17848 | 5/19/21 10:34 | 5/19/21 10:34 | 5/19/21 10:34 | 0 | 172.28.5.2 | 0 | IND21-0004 | How a product development project becomes a data factory approach. | In this presentation we describe how we developed modules and pipelines for the industrial processing of data and content from a machine learning project in the legal domain. Lawyers’ work is far from being digitized. They need to deal with large amounts of printed documents of various types (e.g. letters, contracts, invoices, orders, offers, court documents). A single mandate could consist of several hundreds of paper folders. To work on specific aspects of a case, they need to review, annotate, reorder, even reprint and reattach their documents to a new folder. On top of that, legal disputes can take several years to settle. It´s not easy to keep track of all the relevant details. One of the main pain points for lawyers is to be able to grasp the connections and processes in a mandate in a reasonable amount of time. This setting became the starting point for our core use case and our user journey: With CaseWorx, we support lawyers to organize and to structure documents of a construction law mandate by applying machine learning and semantics combined with domain knowledge! | In this talk we will give a very brief overview on Wolters Kluwer and then start with describing the use case, which is forming the basis for our CaseWorx product. After introducing the problem space, the main goal of the presentation is to show how a machine learning project can be industrialized and an ML Ops pipeline can be built. We will focus on the following aspects: First, we will describe how we developed, measured and automated the manufacturing process itself. So what elements the pipeline consists of, what dependencies there are and how we made sure that we got this reliable and stable, taking into account that we talk about processing knowledge and not nails in an industrial fashion. Then we talk about our quality measures and how this is implemented in the process. Finally, we describe in more detail the aspects that make the ML Ops work – things like automation and modularization. At the end of the presentation we look at how our approach could be used in a more generic way with respect to scalability and adaptability - within our legal domain and beyond. A short demo of CaseWorx will show where machine learning and semantics are used and how this enables the application to significantly improve the complex and demanding daily business of a construction law professional. | Carsten Böhmert | carsten.boehmert@wolterskluwer.com | male | Wolters Kluwer Deutschland GmbH | Germany | Data Engineer | At the interface of language and technology, Carsten works agile at Wolters Kluwer with data and AI on the development and implementation of components for new digital products. As an expert in enhancing customer content, he analyzes data, carries out series of tests, takes business and customer requirements, and converts them into user stories. Carsten previously worked at LexisNexis and Wolters Kluwer in the editorial content area on the standardization, automation, and offshoring of data and processes and was Product Owner for EU content. | Christian Hartz (Legal Engineer at Wolters Kluwer Deutschland GmbH), christian.hartz@wolterskluwer.com | yes | Legal | Y | Y | Y | Y | TRUE | ||
20 | 275 | 18782 | 5/25/21 21:18 | 5/25/21 21:18 | 5/25/21 21:18 | 0 | 172.28.5.2 | 0 | IND21-0003 | Validating InterConnect’s interoperability layer in a Smart Home environment | The H2020 InterConnect project aims to improve the semantic interoperability in the smart home, building and grid domain. The project has a budget of 35 million euro’s and gathers 50 European entities from 11 countries. An vital component for success is its interoperability layer, which allows platforms, services and devices (knowledge bases) from different vendors to exchange data. The interoperability layer was designed to utilize semantic technologies like ontologies and reasoning. InterConnect’s interoperability layer will be realized by the Knowledge Engine. The Knowledge Engine is a working prototype that is built on Semantic Web technology to harness the power of domain knowledge. It uses a generic component called the Smart Connector to realize intelligent data exchange. Every platform, service or device runs an instance of this Smart Connector and configures it with its capabilities. This resulting network of Smart Connectors can exploit these capabilities to orchestrate data exchange with the help of a rule reasoner that uses the capability descriptions to determine where data should come from or go to. | The H2020 InterConnect project aims to improve the semantic interoperability in the smart home, building and grid domain. The project has a budget of 35 million euro’s and gathers 50 European entities from 11 countries. An vital component for success is its interoperability layer, which allows platforms, services and devices (knowledge bases) from different vendors to exchange data. The interoperability layer was designed to utilize semantic technologies like ontologies and reasoning. Furthermore, it should discover and connect to relevant knowledge bases automatically. Moreover, to address privacy and security concerns, it should be largely decentralized and function as a serving hatch and not as a data storage. InterConnect’s interoperability layer will be realized by the Knowledge Engine. The Knowledge Engine is a working prototype that is built on Semantic Web technology to harness the power of domain knowledge. By using this domain knowledge in the form of an ontology, it provides several ontology-based features, such as reasoning, explainability, fine grained security policies and smart data exchange. Within InterConnect we focus on the last one, which is realized by a generic component called the Smart Connector. Every platform, service or device that a home owner uses, runs an instance of this Smart Connector and configures it with its capabilities. This network of Smart Connectors can exploit these capabilities to orchestrate data exchange. To achieve this, the Smart Connector contains a rule reasoner that uses the capability descriptions to determine where data should come from or go to. Hyrde’s Ekco platform manages the appliances, devices and sensors that reside in the commercial space and apartments of the project’s Dutch pilot. Apart from monitoring and controlling these devices, the platform allows users to configure rules and triggers in automated workflows that optimize comfort and energy consumption. In our demonstration Ekco uses a Smart Connector to make a limited set of capabilities available to other knowledge bases via InterConnect’s interoperability layer. In the process of developing the Knowledge Engine and validating it with use cases from Hyrde and other partners, we learned the following lessons: the capability descriptions are not always expressive enough to intuitively capture the functionality that a particular knowledge base provides. Also, current reasoners seem unsuited for the distributed nature of the Knowledge Engine. Therefore, we consider to implement a more suitable reasoner within InterConnect. | Barry Nouwt | barry.nouwt@tno.nl | male | TNO | The Netherlands | MSc | Barry Nouwt (MSc) is a medior Scientist semantic technology at TNO within the Data Science department. He obtained a BSc degree in Computer Science from the Saxion University Of Applied Sciences and an MSc degree in Artificial Intelligence at Utrecht University in 2008. Until 2015, he worked with SemLab B.V. on commercial applications of Natural Language Processing (NLP) and Semantics, primarily in the Financial and Government domain. At TNO, Barry’s research activities centre around ontologies, model-driven development and semantic reasoning with a focus on increasing the value of formalized domain knowledge. He achieved a major improvement on reusability at the department by introducing a platform that offers ontology-based functionalities such as, e.g., orchestration and security that fully utilize ontological concepts, relations and constraints. He applies his research results and software engineering skills to diverse projects in the Health, Industry, Defence and Agricultural domain. For further details, see his linkedIn profile (https://www.linkedin.com/in/barry-nouwt-abab5710). Barry's publications can be found here: https://scholar.google.nl/scholar?q=author%3A%22Barry+Nouwt%22 | possible | Energy | Y | Y | Y | Y | TRUE | |||
26 | 281 | 18901 | 5/26/21 16:14 | 5/26/21 16:14 | 5/26/21 16:14 | 0 | 172.28.5.2 | 0 | IND21-0006 | The EU Knowledge Graph | The European Commission is collecting structured data from many sources, in particular from its member states. There are scenarios where this data is distributed, heterogeneous and published with different schemas. We present how such data is aggregated and integrated at the European Commission using a Knowledge Graph which is called "EU Knowledge Graph". We present the knowledge graph architecture we use, how the data in the knowledge graph is accessed by non-expert users, how it can be accessed using natural language and which services are constructed on top. The approach that we follow is novel with respect to two aspects: first we reuse Wikibase (the software that runs Wikidata) as an infrastructure for the Knowledge Graph and second we make it accessible via natural language using QAnswer. This allows to perform analytic queries by non-expert users. | Problem The European Commission, like many other big institutions, encounters the problem of aggregating and integrating data from different data sources. Knowledge Graphs are ideal for such scenarios. In this context we present how we integrated data related to projects financed by the European Commission in its 27 member states. The objective is to make this data easily accessible to citizens and queryable for decision makers. The data is inserted in the "EU Knowledge Graph" which is accessible at https://linkedopendata.eu/. Description In this talk we are going to describe how we set up and maintain the knowledge graph infrastructure of the "EU Knowledge Graph". In particular we will cover the following points: - we will describe Wikibase, the infrastructure that we use to host and maintain the EU Knowledge Graph - we describe how we modelled the knowledge around the European Commission using publicly available data - we show how data internal to the commission is ingested and linked to the initial knowledge - we describe how the knowledge is maintained both by humans and bots - we describe how we enrich the data using public web-services and public data sources - we show which services are relying on the EU Knowledge Graph - finally we show how Question Answering Technology is used to perform analytics queries on top of the dataset Innovation 1) We reuse Wikibase as an infrastructure for the Eu Knowledge Graph. 2) We present how the data can be accessed by non-expert users using question answering technologies. Lessons Learned We will point out which are the benefits of using Wikibase and Question Answering Technologies as well as the challenges that we encounter. In particular we will focus on multilingualism, access of the knowledge by non-expert users, data evolution and synchronisation. We believe that this talk can give interesting insights in the construction of knowledge graphs for enterprises and institutions. Additional Presenters: - Max De Wilde, Information Architect, DG CNECT, European Commission, max.de-wilde@ext.ec.europa.eu - Anne Thollard, Team Leader Knowledge Management, DG REGIO, European Commission, anne.thollard@ec.europa.eu | Dennis Diefenbach | dennis.diefenbach@the-qa-company.com | male | The QA Company | Germany | Dennis Diefenbach is a PhD researcher and entrepreneur in the area of Question Answering over Knowledge Graphs. He did his PhD under a Marie Skłodowska-Curie action, Europe's most competitive and prestigious research and innovation fellowship. He worked for IBM and SAP. Dennis Diefenbach published over 20 publications in the area of QA over Knowledge Graphs in renowned Conferences and Journals like the International Semantic Web Conference and the Semantic Web Journal. He is the main contributor of QAnswer, the first AI driven platform to query Knowledge Graphs in natural language. Currently his main goal is to bring these exciting technologies to industry! | Max De Wilde and Anne Thollard (see description for details) | yes | Government | Y | Y | Y | Y | TRUE | |||
29 | 284 | 18908 | 5/26/21 17:00 | 5/26/21 17:00 | 5/26/21 17:00 | 0 | 172.28.5.3 | 0 | IND21-0016 | Knowledge Graphs in engineering of capital process plants | This presentation shows the challenges a large engineering contractor has with the diversity of the data handled in constructing process plants. Fluor discusses how they have joined a cooperation of 80 companies to design a data model, and to help author an ISO standard based on semantic web technologies. It clearly shows the advantages of semantic web over relational databases, the innovation involved, and the lessons learned. Possibilities are listed that offers opportunity for open source and commercial software companies. | Fluor is an engineering contractor, constructing process plants and other capital facilities. There is a large amount of data involved in engineering. Especially challenging is its data diversity, within a project, and between different projects. In cooperation with more than 80 companies a data model was developed, as well as an ISO standard for data integration of process plants. The ISO standard is based on semantic web technology. This standard is innovative in its way to enable data modeling of any kind of data structure, without having to create thousands of tables as would be required in a relational database. This was the lesson learned in the early start of making the standard. This method clearly shows the big advantage of using semantic web technology. In this presentation the challenges in engineering are discussed, that can be addressed by using Knowledge Graphs for data integration and analytics. Not only structured data can be integrated but also unstructured data and the knowledge that Fluor has gathered in her Knowledge Management databases. Fluor can guide other companies into making similar data structures and how to join the cooperation. Analysis of the data flows of an engineering project still shows how much manual work is involved in making the requirements for plant items, then purchasing them, and then do the quality check on the delivered equipment. These manual steps are a big risk to the construction phase. If errors have not been detected in the quality checks and then discovered when the equipment arrives at the job site, it can result in a schedule slip. Knowledge Graphs can help to automate much of the quality checks. The ISO standard provides for a single data model that can be used to deal with the data variety. The semantic technology provides the means for federating data between companies in the supply chain. The uniformity of the data model enables new kinds of user interfaces to be used, that are simpler, more intuitive, and less work to build; this gives an opportunity for open source as well as commercial software companies. | Paap, Onno | onno.paap@fluor.com | male | Fluor Corporation | The Netherlands | Onno Paap is an experienced engineer and software developer, who was involved in innovative engineering software projects for his entire career. Within the Fluor company he is the Fellow in Semantic Technologies and Data Interoperability. He has led software teams for in-house development of all engineering disciplines. Currently he is part of the Knowledge Enablement team in Fluor, who is developing Knowledge Graphs and new methods of data integration that can be used on Fluor’s hundreds of construction projects. | Manufacturing | Y | Y | 0 | Y | TRUE | |||||
31 | 286 | 18925 | 5/26/21 18:21 | 5/26/21 18:21 | 5/26/21 18:21 | 0 | 172.28.5.3 | 0 | IND21-0022 | Smart police reports: creating semantically rich police reports | To mitigate the problem of costly and time-consuming creation of document generation software, we present a solution used by the Dutch National Police that uses semantic models (SHACL/RDF and SHACL Rules) and XSL stylesheets to generate PDF documents that can be used as legal documents in court. The required time and effort could be reduced by using such a model-based approach while maintain the legal quality of the documents. An opportunity was investigated successfully to make these documents machine-readable as well as readable by humans by embedding RDFa data within the XHTML documents. | The creation of police reports is an important part of crime-fighting, as these reports are used as legal proof in court. Due to this nature, these documents have to be in print and conform certain legal requirements. Digital processing is possible, but only under strict conditions. As most information that is required in these documents is registered as structured data in police IT systems, specific IT software is written to generate these police (PDF) reports. Due to its nature, the realization of this software is not easy and takes a lot of time and effort. As the documents have to be in print, it is not easy for other parties (judges, public prosecutors, lawyers) to process these documents automatically. In most cases manual labor is needed to extract the structured information from these documents. To mitigate these problems, we present a solution that has not only decreased the time and effort needed to realize the software, but also has the promise to embed the original structured data within the generated documents, making them both readable for humans and machines. The Dutch National Police is using a large-scale triple store for the registration of law enforcement data. A SHACL/OWL model of the registration is already available. We created a separate SHACL/OWL model for the information needed in a specific police report and used SHACL Rules to describe the translation between these two models. We created an XSL template for the police report and used this template to transform an RDF/XML export of the police report triples to an XHTML document. This XHTML document is further processed into a PDF document, using freely available open source libraries. By using a semantic model instead of the traditional pure software-based solution, we learned that we could substantially decrease the amount of time and effort that was needed to create the software for a particular police report. To get even more business value, we used RDFa to embed the original triples within the XHTML document. Although this embedded data is lost during the transformation to PDF, it created the possibility for machine-readable police reports whenever these XHTML document may be used in court. At this moment, only PDF documents are allowed, due to legal and practical operational issues. We did, however, investigate whether it would be possible from a legal and operational point of view to use XHTML documents in court instead of PDF documents. | Thurlings, Ingrid | ingrid.thurlings@ordina.nl | female | Ordina | The Netherlands | MSc | Ingrid started her data driven work during her biomedical studies. Since then, she has been using Linked Data and semantic modeling techniques for several projects in the Netherlands as a consultant for Ordina. She has been working as a semantic consultant for the Dutch National Police for over two years, helping them to make the steps to a data driven organisation. | Government | Y | Y | Y | Y | Y | TRUE | |||
35 | 290 | 18949 | 5/26/21 21:41 | 5/26/21 21:41 | 5/26/21 21:41 | 0 | 172.28.5.2 | 0 | IND21-0019 | Data driven investigation and registration at the Dutch National Police | The Dutch National Police is one of the largest governmental organisations in the Netherlands, with over 65.000 employees. Currently, we are undertaking the effort to redesign and realize the new ad-ministrative information system that should support the day-to-day processes of most of the common police agents ‘on the street’. As we are using RDF technology, this will arguable create one of the larg-est online transaction processing software solutions using RDF. The current software production cycle is performed using SAFE en Agile methodologies by more than a dozen high-performance teams. As one might expect, the modeling of this database is a huge effort, which needs to be managed accordingly. The department of Data Use and Data Management of the Dutch National Police is using the MIM 1.1 standard to generate company wide semantic models to ensure semantically correct data integration in the business. Application designers use the conceptual models to generate application specific logical and technical models. Speeding up the police process requires semantically clear data on subjects and the highest data quality possible. The Semantic Model of the Police enables a registration process that in-tegrates applications, departments and even strategic goals of national security. | Initial Situation The challenges the National Police faced were the integration of data between applications and the use of shared data by users of different instances of the main registration application of the police. Approach and IT-Solution The force radically changed its approach to an innovative data driven strategy to strengthen the information position of officers in the field and in the office. The establishment of a data layer, approachable by all apps became the focal point of IT development. The solution is composed of an Oracle database in which the triples are stored as the foundation of a shared semantic model. The toughest challenge has been to provide the organization with sufficient human resources to further design and realize the data infrastructure. The semantic components support the process from data to knowledge, using an integral set of models. We are adhering to the MIM 1.1 standard. The MIM Meta model for Information Models provides the modeling standard for a four-level framework of models: 1. Semantic model of concepts / vocabulary 2. Conceptual information model / ontology 3. Logical information model / data model 4. Technical information model / physical or implementation model. Lesson learned Success Criteria for and Benefit of the Semantic Solution Crucial factors to the success of our project have been: • Support and understanding of linked data by police officers; • Participation of police officers in the software realization process; • The establishment of an Ontology Design Authority; • The software production cycle is performed using SAFE en Agile methodologies by more than a dozen high-performance teams. The biggest obstacles were the lack of sufficient software architects able to transfer their knowledge to developers and knowledge gaps among developers that led to sub-optimal implementations with architectural debt as a result. The benefit of our solution we can measure by time saved by users executing police processes in a fraction of the time needed in the past. The organization benefits from our solution being able to register frequently occurring crimes like shop lifting on the spot in the street, saving the organizations tens of thousands hours a year. Prospects and Recommendations The next steps planned in our project are the extension of crimes and offences that can be registered on our linked data platform. We recommend other organizations with the desire to work with linked data, to adhere to the available standards and to exchange ideas with governmental organizations. | Saskia van der Elst | saskia.van.der.elst@politie.nl | female | Nationale Politie | The Netherlands | Drs | Sr. Data Architect, Nationale Politie. Book and Information Science, University of Amsterdam. https://www.linkedin.com/in/svdelst/ | yes | Government / Legal | Y | Y | 0 | Y | TRUE | |||
36 | 291 | 18959 | 5/26/21 22:43 | 5/26/21 22:43 | 5/26/21 22:43 | 0 | 172.28.5.2 | 0 | IND21-0002 | Datafication of juridical texts in real estate | In the datafication of juridical texts in real estate we combine a complete round-trip set of semantic web technogies. We built a skos-based taxonomy to organize the knowledge we derive from the text models and translate that knowledge into an ontology. The result is an rdf-a document that contains all the relevant data that conforms to the ontology and is self-explaining by the taxonomy. | The Dutch program ‘Carefree real estate’ is a partnership between notaries, brokers, financial suppliers and the Dutch Cadastre. The program aims to give buyers and sellers of real estate more certainty at the start of the chain of buying a house. We want to achieve this goal by providing as much as possible verified data about a house as early as possible in the process. This data must be ready to re-use in later steps. The initial situation was (paper) document-based with only partly verified information. This brought a lot of uncertainty to buyers and sellers, where buying or selling a house is a serious live event. Also in every step in the process each partner had to derive the data from the document again for its own. We started with datafication of the purchase agreement. This agreement is made up by a real estate broker. The data is relevant for banks that supply mortgages, for notaries who take care of the actual transfer of the house and in the end for the Cadastre that registers the new juridical situation. We realised the data model of this agreement using several standards and toolkits we developed the last 7 years. In the first step we described the ‘universe of discourse’ by building a skos-based thesaurus on the knowledge we derived from the textual model of the standard purchase agreement. To enrich the knowledge model we made an extension on skos to annotate legal acts, actors and objects in the text. The next step was to translate this knowledge model to a data model. Therefore we developed a more or less standard pattern, where the skos extensions were of good help. We also used the Dutch Metamodel for Information Models (MIM), to guarantee the interoperability whit other data models that are in use. From this abstract MIM data model we used a toolkit to automatically generate a serialisation for a Linked data ontology. Provenance data was added to ‘prove’ the authenticity of all the data. The final specification step was to create an rdf-a specification that contained on the one side all the juridical relevant text and on the other side all the data conform the ontology. The resulting package can be digitally signed by a qualified signing service. So a ‘proven’ set of text and data can be re-used in every next step in the process. Using Linked data techniques for linking to explanations we last but not least created a self-explaining ‘document’. Next steps are datafication of deeds of transfer, mortgage deeds and sustainability scenario’s. In the end the complete process from orientation on the real estate market to registration at the Cadastre will be based on (from te beginning verified) data instead of on documents. | Arjen Santema | arjen.santema@kadaster.nl | male | Kadaster | The Netherlands | drs.ir. | Arjen Santema studied psychology and informatics. At the moment he works as datamanagement consultant at the Dutch Cadastre. In the past he worked as an IT architect at the Department of Internal Affairs, at the Department of Justice and in a forensic clinic. Personal drivers are connecting people and share knowledge. | possible | Legal | Y | Y | Y | Y | TRUE | |||
37 | 292 | 18960 | 5/26/21 22:43 | 5/26/21 22:43 | 5/26/21 22:43 | 0 | 172.28.5.4 | 0 | IND21-0010 | Salus: A Solution for Labor Agreement Processing in Dutch. | The Ministry of Social Affairs & Employment in the Netherlands conducts around 30 research studies on labor agreements per annum, with the topics varying from flexible labor market, diversity in the workplace, and the impact of the Covid-19 crisis on the collective labor agreement. The aim of these studies is twofold: on the one hand the collective labor agreements are a gauge for policy, and at the same time they can form an inspiration. The current study method is labor-intensive, takes around 4-5 months to complete and does not cover all labor agreements in the Netherlands (100 largest collective labor agreements, with a coverage ratio of 85%). To automate and enhance the process, DEUS, together with the experts from the ministry, has built Salus. Capitalising on the recent advances in natural language processing, all labour agreements have been made searchable and question answering (QA) models have been built to process research questionnaires in Dutch. This has led to a significant increase in completeness of labor agreement processing. Furthermore, agreements can be processed at once, allowing for faster and more complete research and freeing up time for the experts to focus on the in-depth data analysis. | During our presentation we would like to discuss a practical use case in which we briefly explain the current working method of the researchers at the Ministry of Social Affairs and Employment in the Netherlands and the solution that has been implemented to optimise the working process by DEUS. The problem the researchers encountered is the infeasibility of conducting complete, up-to-date and precise research into all collective labor agreements in the Netherlands, for all subjects in which the ministry is interested. The current solution is to use a sample of collective labor agreements, which unfortunately entails limitations in terms of representativeness, among other things. The newly proposed solution provides a partially automated working method, whereby the innovative character of this case lies in both the working method and the substantive method. The method chosen is a short and flexible period in which DEUS and the Ministry of Social Affairs and Employment have closely collaborated to build Salus, a solution that capitalises on the recent advances in the Natural Language Processing (NLP) field. Salus includes indexing and search functionality, which allows end-users to create their own indices and retrieve relevant paragraphs per labor agreement, and question answering (QA) module for factoid and boolean questions. We have evaluated available pre-trained models for Dutch, such as multilingual BERT and BERTje, on the QA task. Because of the lack of labeled data for question answering in Dutch, the QA models have been built in the iterative fashion with several evaluation rounds, which allowed us to gather QA examples from labor agreements to fine-tune models. Our experiments have shown that model fine-tuning was crucial to attain higher performance and more reliable probability estimates. As a result, only answers with high probability are displayed to the end-user. In addition to a single query functionality, users can create research questionnaires and compare results across indices, which provides insights into potential revisions of labor agreements over time. The application has been dockerized, deployed and tested in the cloud environment. We will conclude the presentation with a short demo. | Katrenko, Sophia | sophia.katrenko@deus.ai | female | DEUS | The Netherlands | PhD | Sophia Katrenko, Head of Data Science at DEUS, has worked in advertising, research, finance, publishing, and continues exploring Data Science & Analytics in other sectors, preferably with a measurable impact on society. Sophia is particularly interested in bridging the gap between research and industry, and promoting knowledge valorisation. She has organised events on machine learning, built high-performing teams across multiple geographies, supervised data scientists in industrial and academic environments, and served on boards of program committees of multiple international conferences. Sophia holds a PhD in Computer Science from the University of Amsterdam. | Daniella van de Langenberg, the Ministry of Social Affairs & Employment, dvdlangenberg@minszw.nl | yes | Legal | Y | Y | Y | Y | TRUE | ||
39 | 294 | 18972 | 5/27/21 0:38 | 5/27/21 0:38 | 5/27/21 0:38 | 0 | 172.28.5.4 | 0 | IND21-0007 | Cognitive Services in LegalTech: Combining automated Text Processing Methods to enrich Law Documents | Models for recognizing categories, entities, roles, and patterns in legal text documents have been developed mainly for forms: They have in common that they require some fixed structure to reliably obtain the relevant information and only predefined subjects are covered by existing algorithms. Neither polystructured nor free text data are interpreted in a meaningful way. Cognitive Services in LegalTech is considered as a platform to handle incoming documents regardless from its format and according to the following aspects: Subdivision and Separation, Categorization, Entity Recognition, and the Anonymization of all personal data, where the assignment of recognized entities to specific roles is the specific complexity that will determine the success of the project in terms of readability and usability. A particular prerequisite for the success of the new approach is also the development of a live quality measure for the quality of the data analysis. | We introduce our solution consisting of different legal text processing use cases that were implemented and brought into production for the Austrian Federal Ministry of Justice. The initial situation had no sufficient categorization of text documents (resp. mainly assigned as „miscellaneous“). Structured metadata is available, but showed inaccuracies given by manual entries (typos, acronyms, entries in wrong fields etc.). As for court decisions and their provision to the public and also other judges, only manually anonymized data could be provided (due to regulations). Therefore, we created a Machine Learning & AI platform where text documents could be automatically split and assigned to a category, based on its content only. Moreover, an automated anonymization was established that is supposed to work for court decisions, but could be used not only for them. The strength of anonymization is configurable. The innovative character is shown by the variety of approaches used in the system as well as their combination in order to balance out weaknesses of single models. Therefore, a couple of methods may be used such as normalization, weighting and grouping. The probabilistic approach of Deep Learning Models (NLP, NER) is an advantage for defining a robust model in the first place, but needs to be strengthened by heuristics, dictionaries and rules – dependent on the customer’s needs and requirements. The number of different recognizers are a result of that. Specifications of the Austrian Law are to be considered as well as the fact that general rules for anonymization or de-anonymization are not easy to be defined. We also tried to set up a Live Quality Validation to ease the handling of processed files and minimizing possible errors. Hence, we are glad to share our insights, best practices, but also possible pitfalls and lessons learnt from our project that we have been working on for more than three years now. | Frederick Bednar | frederick.bednar@ebcont.com | male | EBCONT proconsult GmbH | Austria | Mag. | Not only because of his statistical background (Master degree in Management Science at WU Wien), Frederick Bednar has been gathered experience as a "data aficionado" for about two decades now. At the Austrian IT full-stack service company EBCONT, he works as a Data Analytics and Data Science Consultant and is responsible for projects especially in Machine Learning & Deep Learning, NLP & NER, but also BI consulting and IoT projects. | N.N. (a second colleague of my team at EBCONT could join, either Stefan Milchram or Thorsten Jojart) | yes | Legal | Y | Y | Y | Y | TRUE | ||
41 | 296 | 18997 | 5/27/21 9:15 | 5/27/21 9:15 | 5/27/21 9:15 | 0 | 172.28.5.4 | 0 | IND21-0012 | Maps and Cartography resources linked in a Knowledge Graph for big audiences: the semantic web of National Geographic Institute | The National Geographic Institute of Spain (NGI) knowledge graph and semantic web is a project built with symbolic artificial intelligence based on a semantically interpreted knowledge graph, powered with GNOSS technology. The knowledge graph of the National Geographic Institute of Spain integrates more than 2 million geographical resources, coming from 3 different sources of NGI, described in a very expressive way and offers a set of interfaces for people that allow their discovery, exploration, navigation, search and visualization. We will present the data sources integrated; the digital semantic model defined for the cartographic resources and maps of NGI; the processes of semantic annotation, grouping and recognition of entities to build and enrich the NGI knowledge graph; and the main applications for the end-users, as well as the main challenge address during the development of the project. The framework for developers used to build the knowledge graph and the end-user applications is GNOSS Knowledge Graph Builder. In the talk we will present a demonstration of the semantic web for the public. Now, the access to the web is restricted to a group of test users. | We will present the case of the National Geographic Institute of Spain (NGI) knowledge graph and its semantic web for the public, explaining the process and the challenges we have address to develop it. It is a project built with symbolic artificial intelligence based on a semantically interpreted knowledge graph, powered with GNOSS technology. Among other functions, the NGI develops programming of the National Cartographic Plan and the production, updating, and exploitation of topographic and cartographic databases at the national level for their integration into geographic information systems. This institution falls within the domain of the Directorate General of the Ministry of Development of the Government of Spain. The main problem addressed was to integrate the information coming from different sources of the NGI and offer to the public a new web user experience to interrogate and retrieve the information in a more intuitive, useful, significant, and simple way. The final aim is to achieve a situation in which any citizen can efficiently obtain and download or purchase maps created and edited by the NGI in a straightforward, practical manner. The innovative approach consisted of building a unified and interrogable Knowledge Graph and the end-user applications that allow to put all the data to work. We will present the main works and results: - Define a Digital Semantic Model to semantically represent the cartographic resources and maps. A summary of the model will be presented. - Integrate the data coming from three not-connected structured data sources of NGI: Download Center (geographic files in digital format grouped into series), Map library (resources corresponding to national and international cartographic funds, ancient and modern, and all the productions of the NGI of Spain) and CNIG Virtual Store (products for sale). - Consolidate the data into a knowledge graph, link them and enrich the graph, developing different processes of semantic annotation, grouping and recognition of entities are carried out. - Build the Semantic Web of NGI for the public, as an exploitation of the Knowledge Graph. It includes: interrogation systems based on human reasoning (search engines), geographic Search Systems, enriched web pages for every geographical product, contextual information and recommendation. We will also present the main challenges addressed, summarized as follows: sort algorithms to resolve the most relevant results on the first page, improving search engine performance for complex queries, building grouped geographical resources; converting textual searches into geographical searches and into entity searches. | LÓPEZ SOLA, SUSANA | susanalopez@gnoss.com | female | GNOSS | Spain | PhD in Chemistry, University of La Rioja | Dr. Susana López Sola (PhD in Chemistry, University of La Rioja) is a consultant and manager of the commercial department of GNOSS. She joined the company in January 2008, where she has participated in various technological development projects both internal (R&D projects related to the evolution of the platform) and external associated to Semantic Web, Ontological Engineering, Data Representation and construction and exploitation of Knowledge Graphs. She has participated in some of the main digital projects carried out by GNOSS. She had the responsibility of Project Manager in the following projects, among others: European Project AFEL "Analytics for Everyday Learning" (financed by the H2020 program, http://cordis.europa.eu/project/rcn/199117_en.html) whose main objective for RIAM is the development of tools to improve the teaching/ learning processes on the basis of the data (data about the patterns of use and behaviour) that users deposit in the semantic platform; Triodos Bank Project 'Management system by knowledge components of the operations management system of Triodos Bank'; BBVA Project, Metasearch of bbva.com and BBVA Research; 'Deusto Knowledge Hub' project for the University of Deusto and 'Comillas Knowledge Hub' for the University of Comillas; the search engine of educational centers of Madrid. As a consultant, she has also taken part in several strategic projects of regional training and innovation systems carried out by RIAM I+L Lab. Previously she worked for six years as a research scientist in Chemistry at the University of Heidelberg-Germany (June-December 2002), the University of La Rioja (2003-2007), and the University of Maryland-USA (Oct-Nov 2005). | Government | Y | Y | 0 | MB: Working solution with links between geo and semantics | Y | TRUE | |||
49 | IND21-0030 | Invited talk Florian: Enterprise Knowledge - Employee 360 (Semantic technologies and knowledge graphs in Human Resource usecases) | Florian Bauer | florian.bauer@semantic-web.com | Florian Bauer | florian.bauer@semantic-web.com | Y | Y | Y | Y | TRUE | ||||||||||||||||||
50 | |||||||||||||||||||||||||||||
51 | 30 | ||||||||||||||||||||||||||||
52 | 6 | yes'S | 18 | 21 | 12 | 30 | 13 | ||||||||||||||||||||||
53 | female | 9 | thereof female | 4 | 5 | 2 | 4 | ||||||||||||||||||||||
54 | male | 32 | |||||||||||||||||||||||||||
55 | |||||||||||||||||||||||||||||
56 | Missed in List (now added by FB) | ||||||||||||||||||||||||||||
57 | 298 | 19001 | ####### | ####### | 27/05/2021 09:43 | 0 | 172.28.5.4 | 0 | How the Prado Museum is once again at the forefront of online museums by applying Artificial Intelligence to its collections. | On the occasion of the bicentennial of the Museo del Prado, two new functionalities were presented and made available to the world, which promote and deepen the digital strategy begun by the Museum years ago with the launch of the Prado on the web project, built with symbolic artificial intelligence based on a semantically interpreted knowledge graph, powered with GNOSS technology. These two new functionalities highlighting the resolute Museum´s digital transformation are Augmented Reading and The Prado Timeline GNOSS will not only share in depth those new functionalities but what the value they bring to users and visitors showing how deeply transform the web visitor experience. First of them, the augmented reading of the collection, makes the Museum's unstructured content intellectually closer and more accessible to its web visitors by providing them with contextual information about the works they are viewing, recognizing and extracting the entities contained in their descriptions, and providing additional information, so that anyone reading them can understand the works in depth Second one, The Prado Timeline, brings a contextual, historical and interdisciplinary view of the Prado collection through the connection of its works and authors with related data providing the frame of reference that explains them. | We will explain how the project was developed, what the rationale behind those new functionalities and many additional relevant information about them: what the technology they are based on, etc. In order to situate the attendees, we will begin with a brief review of the Prado on the web project, followed immediately by a detailed analysis of the new functionalities. We will also enter in those different dimensions and impact of those new features for the Prado Museum as they not only bring new ways for the end users of enjoying and discovering Prado´s works but also new capabilities to their internal users helping the Museum to enhance them, what is not only consistent with the Museum objectives but perfectly aligned with a desired impact on user experience, changes in operational processes and new business model opportunities. For example, Augmented Reading is from the user perspective specially relevant because understanding a non-contemporary work of art often requires access to and knowledge of a context plenty of mythological, religious, literary, or philosophical references of antiquity that may be unfamiliar to contemporary people approaching Prado´s collection. But from the museum perspective Augmented Reading has an additional and very important dimension to be considered as it brings more technical level of exploitation, aimed at improving the processes of documentation and representation of the works in the collection. Related with the Prado Timeline functionality its aim was to provide the Prado's collection and authors with a broad, historical and interdisciplinary context that frames, explains and universalises them; a context that at the same time allows the Prado and its collection to function as a gateway to knowledge as a whole, We will explaing how we got it applying the principles of the Linked Open Data Web., what the challenges we g¡faced and how we ssurpassed them through a set of hybridised ontologies and vocabularies built in accordance with Semantic Web standards. | ALONSO MATURANA, RICARDO | riam@gnoss.com | male | GNOSS | SPAIN | PhD in Sociology, B.S. in Philosophy | Dr. Ricardo Alonso Maturana (PhD in Sociology, B.S. in Philosophy) is the founder, promoter and CEO of RIAM I+L Lab, the enterprise that is owner of GNOSS and Didactalia. GNOSS is a technology company dedicated to the construction and exploitation of Knowledge Graphs in an Artificial Intelligence environment. Ricardo Alonso Maturana has been responsible for the direction, design and conceptualization of the GNOSS platform project, and the main use cases/projects developed by the company RIAM in the field of culture, tourism, education or banking to name a few. He has conceptualized, imagined and directed the most important projects of digital transformation of the organizations carried out by GNOSS in the last 10 years for clients such as Museo del Prado, Patrimonio Nacional de España, Instituto de Salud Carlos III, Museo Lázaro Galdiano, Mis Museos (My Museums), BBVA, Triodos Bank, Grupo Santillana, University of Deusto, University of Comillas, La Rioja Turismo, Previsión Sanitaria Nacional, Instituto Geográfico Nacional, Fundación COTEC, ADVEO Group, National Library of Spain, Air Liquide-Athelia, among others, as well as the direction in projects for AAPP as Ministry of Education, Culture and Sport, Ministry of Finance, Regional AAPP as La Rioja, Aragón, Vasque Country, Madrid or Castilla la Mancha, and Hércules project by University of Murcia (semantic architecture and ontological infrastructure of the research management system of Spanish universities);. He was Co-director of the European Project AFEL "Analytics for Everyday Learning" (financed by the H2020 program). AFEL main goal for RIAM is the development of tools to improve the teaching/ learning processes on the basis of the massive semantic data analysis that users deposit in the semantic platform, in order to personalize and recommend learning paths. | N (already have Gnoss) | N | N | ||||||||
58 | 299 | 19013 | ####### | ####### | 27/05/2021 11:49 | 0 | 172.28.5.3 | 0 | Knowledge Graphs for Solution Sizing in Factory Automation | In this presentation we describe the data processing workflow and querying approach of an online, customer-facing application for the interactive sizing of automation solutions from Festo, a global leader in factory automation. The semantic application uses knowledge graph (KG) technologies to present customers a ranked selection of proper solutions out of millions of electric drive trains from given parameters such as working range, moving mass, travel time, etc. The talk will provide insights to the KG relevant parts of the productive application such as: - transformation of product data into an RDF knowledge graph by using a R2RML driven ETL process - reasoning based on OWL RL and SWRL to infer product compatibility from technical interface specifications - enrichment of the KG with business related constraints and sales information via SPARQL-Update - high-performance querying using a cascade of concurrent SPARQL queries that involve complex technical filter and ranking conditions | Festo SE & Co. KG is a global manufacturer of products for factory automation. When designing a linear electric movement application, a customer has the choice from millions of different solutions, build from some hundred actuators, mounting kits, gears, motors and controller. A reasonable sizing tool has to find the best technical solutions ordered by price for a particular user provided configuration within seconds. This involves a vast amount of component compatibility checks each of which with complex arithmetic filter calculations and a ranking of the complete result set. The presentation describes the data workflow as well as processing concept of the sizing application: During the data preparation workflow, product data are transformed from a central ERP System into an OWL RL knowledge model and is enriched with expert knowledge on product interoperability. SWRL rules are applied to infer compatibility between basic components. Based on this a valid drive train consists of a chain of compatible components from an actuator via other components such as mounting kit, motor, etc. through to a controller. The resulting solution space of drive trains includes millions of possible combinations. In the Festo sizing tool the user can specify his technical requirements via a web interface. The sizing query service retrieves feasible drive trains, ranked by their fitness regarding the customer requirement. A major challenge is the realization of performant query generation and answering in order to enable almost instant responsiveness of the application. Ranking all solutions in a single query is not feasible. Therefore, we divide a user request in a cascade of queries based on elaborate groupings by a mix of product group and technical criteria. The discussed sizing tool is deployed world-wide on the AWS cloud and accessible via Web to customers as well as the internal support team at Festo. | Jens Wissmann | jens.wissmann@festo.com | male | Festo SE & Co. KG | Germany | Dr. | Jens Wissmann is a knowledge engineering and smart data expert at Festo. He is responsible for the development of the Festo Semantics Platform that provides services for solution composition and search. Prior to joining Festo, he worked at 1&1 (United Internet) applying rule systems, business process models, and big data technologies in web analytics and fulfillment processes. He holds a PhD in Computer Science from City University London and worked as researcher at FZI Research Center Karlsruhe. | ? | 0 | Y | ||||||||
59 | 300 | 19014 | ####### | ####### | 27/05/2021 11:50 | 0 | 172.28.5.2 | 0 | Evaluating NLP software for automatic knowledge graph construction - What to look for | Knowledge graphs are increasingly becoming important in the AI world as an enabling technology for data integration and analytics, semantic search and question answering, and other cognitive applications. However, developing and maintaining large knowledge graphs in a manual way is too expensive and time consuming. To accelerate and scale the process, methods and techniques from the areas of information extraction and natural language processing (NLP) can be very helpful. In this talk we'll see the main NLP tasks that knowledge graph mining involves, the factors that affect how easy or difficult the execution of these tasks can be, and some common pitfalls that we need to avoid in order to mine high quality knowledge graphs. We will also describe some important questions that can help us evaluate available knowledge extraction tools and decide if and to what extent we should use them. The talk is of interest to practitioners who develop knowledge graphs and look for ways to automate the process, as well as to practitioners and vendors who develop NLP solutions for automatic knowledge graph construction. | With knowledge graphs gaining increasing popularity and importance within organizations, the need for effective and scalable knowledge extraction tools that can help accelerate the construction of such graphs is also getting more pressing. For organizations that have no such tools of their own, reusing existing ones (academic or commercial) looks like an attractive option. Nevertheless, not all such tools (that utilize NLP, machine learning and other information extraction methods) are equally capable of extracting good quality knowledge, nor do they have all the features and abilities that industry practitioners really care about when they contemplate their usage. In this talk I will describe the main NLP tasks that automatic knowledge graph extraction involves, the factors that affect how easy or difficult the execution of these tasks can be, and some common pitfalls that may hamper the quality of the results. I will demonstrate through real-world tools and cases how certain tasks are yet too challenging to fully automate, and I will describe some important questions that can help us evaluate available knowledge extraction tools and decide if and to what extent we should use them. The talk is of interest to practitioners who develop knowledge graphs and look for ways to automate the process, as well as to practitioners and vendors who develop NLP solutions for automatic knowledge graph construction. These practitioners will learn: - What are the basic tasks that knowledge graph mining involves, their key dimensions, and the main methods and tools that can be used to tackle them. - How good are current methods and tools in tackling these tasks and what factors affect their effectiveness. - What pitfalls to avoid when developing and/or using such tools - What questions should be asked when evaluating such systems for real-world use, and what kind of answers should be expected to be given. | Panos Alexopoulos | p.alexopoulos@gmail.com | male | Textkernel | Panos Alexopoulos works since 2006 at the intersection of data, semantics, and software, contributing to building intelligent systems that deliver value to business and society. He is currently Head of Ontology at Textkernel BV, leading a team of data professionals in developing and delivering a large cross-lingual Knowledge Graph in the HR and Recruitment domain. He is also the author of the book "Semantic Modeling for Data - Avoiding Pitfalls and Breaking Dilemmas" (O'Reilly, 2020), and a regular speaker and trainer in both academic and industry venues, striving to bridge the gap between academia and industry. | Y | Y | Y | ||||||||||
60 | 301 | 19015 | ####### | ####### | 27/05/2021 11:50 | 0 | 172.28.5.2 | 0 | Knowledge retrieving Coronabot for the German Government driven by Semantic Technologies | The COVID-19 pandemic is a showcase for a data-driven society. Hence, the German government was aiming to provide free access to COVID-19 data to all citizens. However, making the corresponding data accessible by non-experts is not easy due to local characteristics and time-dependent metrics (e.g., the data is only collected on district level). We present the Coronabot facilitating the access to German COVID-19 data capable of answering German and English questions. The component-based system is capable of understanding questions relating time & (even small) places in Germany (s.t., a citizen might ask for the infection numbers in a self defined range of past time). A core demand was high and verifiable quality of the semantic search functionality as citizens need to trust in the provided data. As our system is driven by RDF (that means all internal components interact with each other using RDF and SPARQL), we are enabled to provide a controllable interface which is providing solid and traceable results. Hence, this significantly raises the level of quality assurance compared to traditional implementation approaches, while allowing microbenchmarking of each component using SPARQL on the collected trace information that is represented by RDF. | In this talk the joint work of three partners is presented that were in charge to implement a solution for retrieving COVID-19 related data: Federal Ministry of the Interior, Building and Community (German: Bundesministerium des Innern, für Bau und Heimat) of Germany, as German government-owned technology partner the Informationstechnikzentrum Bund (ITZBund), and as research partner the Anhalt University of Applied Sciences (Germany) was responsible for the knowledge-driven/semantic components. First, we will give a brief introduction into the motivation and requirements. The problem will be described, i.e., providing a high-quality, multilingual and flexible interface that should be enabled to deal with any kind of COVID-19 case-related question that includes a reference to a place located in Germany. Secondly, we will describe the components (and their purpose) of our system and how they interact with each other while computing one or more queries that can be used to retrieve data from the official German Web service providing access to COVID-19 data. As some of the implemented components using public knowledge bases from the Linked Open Data cloud they are highlighted. In the next step, we will put the focus here on the traceability of the system which is only possible due to the core of the used (open source) Qanary framework which interacts with a process-centralizing triplestore. Each component is storing and receiving the data from it. As this triplestore is used as a process memory, each intermediate result can be analyzed (using SPARQL). This is used on the last part of our talk where we use an SPARQL-driven quality assurance process to validate the quality of our system, proving the high-quality despite many requirements. In the conclusion we will provide our findings while implementing such an innovative solution. While using RDF in the core of our system we were enabled to address the challenge of a high-quality solution without having a large volume of training data. The same is true for integrating LOD data and semantic-driven components, s.t., it was possible to implement our solution in a very short time which is crucial in times of a pandemic. | Andreas Both | andreas.both@hs-anhalt.de | male | Anhalt University of Applied Sciences | Germany | Prof. Dr. | Prof. Both studied computer science and obtained his Ph.D. at the Martin Luther University Halle (until 2010) in Germany. Subsequently, he held several leading R&D positions in medium and large IT companies in the segments Web, e-commerce as well as business software and digitalization. Since October 2018, he held the professorship for Web Engineering at the Anhalt University of Applied Sciences and is also Head of Research of the large business software development company DATEV (located in Germany). He is a frequent author of research publications in the field of Question Answering and Linked Data. | Ann Kristin Falkenhain, AnnKristin.Falkenhain@bmi.bund.de, Lead of Chatbot Activities at Federal Ministry BMI of Germany | ? | 0 | Y | |||||||
61 | |||||||||||||||||||||||||||||
62 | |||||||||||||||||||||||||||||
63 | |||||||||||||||||||||||||||||
64 | |||||||||||||||||||||||||||||
65 | |||||||||||||||||||||||||||||
66 | |||||||||||||||||||||||||||||
67 | |||||||||||||||||||||||||||||
68 | |||||||||||||||||||||||||||||
69 | |||||||||||||||||||||||||||||
70 | |||||||||||||||||||||||||||||
71 | |||||||||||||||||||||||||||||
72 | |||||||||||||||||||||||||||||
73 | |||||||||||||||||||||||||||||
74 | |||||||||||||||||||||||||||||
75 | |||||||||||||||||||||||||||||
76 | |||||||||||||||||||||||||||||
77 | |||||||||||||||||||||||||||||
78 | |||||||||||||||||||||||||||||
79 | |||||||||||||||||||||||||||||
80 | |||||||||||||||||||||||||||||
81 | |||||||||||||||||||||||||||||
82 | |||||||||||||||||||||||||||||
83 | |||||||||||||||||||||||||||||
84 | |||||||||||||||||||||||||||||
85 | |||||||||||||||||||||||||||||
86 | |||||||||||||||||||||||||||||
87 | |||||||||||||||||||||||||||||
88 | |||||||||||||||||||||||||||||
89 | |||||||||||||||||||||||||||||
90 | |||||||||||||||||||||||||||||
91 | |||||||||||||||||||||||||||||
92 | |||||||||||||||||||||||||||||
93 | |||||||||||||||||||||||||||||
94 | |||||||||||||||||||||||||||||
95 | |||||||||||||||||||||||||||||
96 | |||||||||||||||||||||||||||||
97 | |||||||||||||||||||||||||||||
98 | |||||||||||||||||||||||||||||
99 | |||||||||||||||||||||||||||||
100 | |||||||||||||||||||||||||||||
101 | |||||||||||||||||||||||||||||
102 | |||||||||||||||||||||||||||||
103 | |||||||||||||||||||||||||||||
104 | |||||||||||||||||||||||||||||
105 | |||||||||||||||||||||||||||||
106 | |||||||||||||||||||||||||||||
107 | |||||||||||||||||||||||||||||
108 | |||||||||||||||||||||||||||||
109 | |||||||||||||||||||||||||||||
110 | |||||||||||||||||||||||||||||
111 | |||||||||||||||||||||||||||||
112 | |||||||||||||||||||||||||||||
113 | |||||||||||||||||||||||||||||
114 | |||||||||||||||||||||||||||||
115 | |||||||||||||||||||||||||||||
116 | |||||||||||||||||||||||||||||
117 | |||||||||||||||||||||||||||||
118 | |||||||||||||||||||||||||||||
119 | |||||||||||||||||||||||||||||
120 | |||||||||||||||||||||||||||||
121 | |||||||||||||||||||||||||||||
122 | |||||||||||||||||||||||||||||
123 | |||||||||||||||||||||||||||||
124 | |||||||||||||||||||||||||||||
125 | |||||||||||||||||||||||||||||
126 | |||||||||||||||||||||||||||||
127 | |||||||||||||||||||||||||||||
128 | |||||||||||||||||||||||||||||
129 | |||||||||||||||||||||||||||||
130 | |||||||||||||||||||||||||||||
131 | |||||||||||||||||||||||||||||
132 | |||||||||||||||||||||||||||||
133 | |||||||||||||||||||||||||||||
134 | |||||||||||||||||||||||||||||