Skip to main content

2021 | Book

The Semantic Web: ESWC 2021 Satellite Events

Virtual Event, June 6–10, 2021, Revised Selected Papers

Editors: Ruben Verborgh, Dr. Anastasia Dimou, Aidan Hogan, Claudia d'Amato, Dr. Ilaria Tiddi, Arne Bröring, Dr. med. Simon Mayer, Femke Ongenae, Assoc. Prof. Riccardo Tommasini, Dr. Mehwish Alam

Publisher: Springer International Publishing

Book Series : Lecture Notes in Computer Science


About this book

This book constitutes the proceedings of the satellite events held at the 18th Extended Semantic Web Conference, ESWC 2021, in June 2021. The conference was held online, due to the COVID-19 pandemic.

During ESWC 2021, the following six workshops took place:

1) the Second International Workshop on Deep Learning meets Ontologies and Natural Language Processing (DeepOntoNLP 2021)

2) the Second International Workshop on Semantic Digital Twins (SeDiT 2021)

3) the Second International Workshop on Knowledge Graph Construction (KGC 2021)

5) the 6th International Workshop on eXplainable SENTIment Mining and EmotioN deTection (X-SENTIMENT 2021)

6) the 4th International Workshop on Geospatial Linked Data (GeoLD 2021).

Table of Contents

Correction to: The Semantic Web: ESWC 2021 Satellite Events

The name of one of the volume editors, Simon Mayer, was erroneously misspelt in an earlier version of the cover and inside cover of this volume. This has now been corrected.

Ruben Verborgh, Anastasia Dimou, Aidan Hogan, Claudia d’Amato, Ilaria Tiddi, Arne Bröring, Simon Mayer, Femke Ongenae, Riccardo Tommasini, Mehwish Alam

Poster and Demo Track Papers

BiodivOnto: Towards a Core Ontology for Biodiversity

Biodiversity is the variety of life on earth which covers the evolutionary, ecological, and cultural processes that sustain life. Therefore, it is important to understand where biodiversity is, how it is changing over space and time, the driving factors of these changes and the resulting consequences on the diversity of life. To do so, it is necessary to describe and integrate the conditions and measures of biodiversity to fully capture the domain. In this paper, we present the design of a core ontology for biodiversity aiming to establish a link between the foundational and domain-specific ontologies. The proposed ontology is designed using the fusion/merge strategy by reusing existing ontologies and it is guided by data from several resources in the biodiversity domain.

Nora Abdelmageed, Alsayed Algergawy, Sheeba Samuel, Birgitta König-Ries
scikit-learn Pipelines Meet Knowledge Graphs
The Python kgextension Package

Python is currently the most used platform for data science and machine learning. At the same time, public knowledge graphs have been identified as a valuable source of background knowledge in many data science tasks. In this paper, we introduce the kgextension package for Python, which allows for using knowledge graph in data science pipelines built in Python. The demo shows how data from public knowledge graphs such as DBpedia and Wikidata can be used in data mining pipelines based on the popular Python package scikit-learn. We demonstrate the package’s utility by showing that the prediction accuracy on a popular Kaggle task can be significantly increased by using background knowledge from DBpedia.

Tabea-Clara Bucher, Xuehui Jiang, Ole Meyer, Stephan Waitz, Sven Hertling, Heiko Paulheim
SLURP: An Interactive SPARQL Query Planner

Triple Pattern Fragments (TPFs) allow for querying large RDF graphs with high availability by offering triple pattern-based access to the graphs. The limited expressivity of TPFs leads to higher client-side querying and communication costs with potentially many intermediate results that need to be transferred. Thus, the challenge of devising efficient query plans when evaluating SPARQL queries lies in minimizing these costs. Different heuristics and cost-based query planning approaches have been proposed to obtain such efficient query plans. However, we also require means to visualize, manually modify, and execute alternative query plans, to better understand the differences between existing planning approaches and their potential limitations. To this end, we propose Slurp ( ), an interactive SPARQL query planner that assists RDF data consumers to visualize, modify, and compare the performance of different query execution plans over TPFs.

Jannik Dresselhaus, Ilya Filippov, Johannes Gengenbach, Lars Heling, Tobias Käfer
Towards Easy Vocabulary Drafts with Neologism 2.0

Shared vocabularies and ontologies are essential for many applications. Although standards and recommendations already cover many areas, adaptations are usually necessary to represent concrete use-cases properly. Domain experts are unfamiliar with ontology engineering, which creates special requirements for needed tool support. Simple sketch applications are usually too imprecise, while comprehensive ontology editors are often too complicated for non-experts. We present Neologism 2.0 – an open-source tool for quick vocabulary creation through domain experts. Its guided vocabulary creation and its collaborative graph editor enable the quick creation of proper vocabularies, even for non-experts, and dramatically reduces the time and effort to draft vocabularies collaboratively. An RDF export allows quick bootstrapping of any other Semantic Web tool.

Johannes Lipp, Lars Gleim, Michael Cochez, Iraklis Dimitriadis, Hussain Ali, Daniel Hoppe Alvarez, Christoph Lange, Stefan Decker
Dataset Generation Patterns for Evaluating Knowledge Graph Construction

Confidentiality hinders the publication of authentic, labeled datasets of personal and enterprise data, although they could be useful for evaluating knowledge graph construction approaches in industrial scenarios. Therefore, our plan is to synthetically generate such data in a way that it appears as authentic as possible. Based on our assumption that knowledge workers have certain habits when they produce or manage data, generation patterns could be discovered which can be utilized by data generators to imitate real datasets. In this paper, we initially derived 11 distinct patterns found in real spreadsheets from industry and demonstrate a suitable generator called Data Sprout that is able to reproduce them. We describe how the generator produces spreadsheets in general and what altering effects the implemented patterns have.

Markus Schröder, Christian Jilek, Andreas Dengel
National Library of Latvia Subject Headings as Linked Open Data

This paper presents the linked data representation of the National Library of Latvia (NLL) authority data consisting of topical, geographic name and form/genre thesauri. It is an important step from a data silo (MARC 21 data) to linked open datasets, represented using SKOS, that can be reused both within and outside the library domain. The datasets are converted to SKOS using mc2skos and are published as linked data using the Skosmos application for publishing SKOS datasets. The paper describes the datasets, including information about their external links. The NLL topical terms dataset, in particular, contains many external links – 85% of its concepts have links to the Library of Congress subject headings dataset that is available as linked open data. The published datasets can be applied to interconnecting Latvia’s cultural heritage information by describing museum and archive items with URIs from these datasets.

Mārīte Apenīte, Uldis Bojārs
Automatic Skill Generation for Knowledge Graph Question Answering

Knowledge Graphs are a critical source for Question Answering, but their potential may be threatened due to the complexity of their query languages, such as SPARQL. On the opposite side, Virtual Assistants have witnessed an extraordinary interest as they enable users to pose questions in natural language. Many companies and researchers have combined Knowledge Graphs and Virtual Assistants, but no one has provided end-users with a generic methodology to generate extensions for automatically querying knowledge graphs. Thus, we propose a community shared software framework to create custom extensions to query knowledge graphs by virtual assistants, unlocking the potentialities of the Semantic Web technologies by bringing knowledge graphs in the “pocket” of everyone, accessible from smartphones or smart speakers.

Maria Angela Pellegrino, Mario Santoro, Vittorio Scarano, Carmine Spagnuolo
Converting UML-Based Ontology Conceptualizations to OWL with Chowlk

During the ontology conceptualization activity, developers usually generate preliminary models of the ontology in the form of diagrams. Such models drive the ontology implementation activity, where the models are encoded using an implementation language, typically by means of ontology editors. The goal of this demo is to take advantage of the developed ontology conceptualizations in order to accelerate the ontology implementation activity. For doing so we present Chowlk, a converter to transform digital UML-based ontology diagrams into OWL. This system aims at supporting users in the generation of the first versions of ontologies by reusing the ontology conceptualization output.

Serge Chávez-Feria, Raúl García-Castro, María Poveda-Villalón
Monetising Resources on a SoLiD Pod Using Blockchain Transactions

Our demo showcases a system that allows users to provide access to web resources in exchange for payments via the blockchain. The system enables users to create offers for their resources or buy access rights for resources belonging to other users. Access rights can be granted only for a limited amount of time. We built our system as SoLiD Pods and Apps: We developed two server modules for SoLiD Pods that automatically (1) grant access for valid payments via the blockchain and (2) remove expired access rights. On top, we developed a SoLiD App that allows to offer resources, browse and request offered resources, and make payments via the blockchain.

Hendrik Becker, Hung Vu, Anett Katzenbach, Christoph H.-J. Braun, Tobias Käfer
Towards Scientific Data Synthesis Using Deep Learning and Semantic Web

One of the added values of long running and large scale collaborative projects is the ability to answer complex research questions based on the comprehensive set of data provided by their central repositories. In practice, however, finding data in such a repository to answer a specific question often proves to be a demanding task even for project scientists. In this paper, we aim to ease this task, thereby enabling cross-cutting analyses. To achieve that we introduce a new data analysis and summarization approach combining semantic web and machine learning approaches. In particular, the proposed approach makes use of the capability of machine learning to categorize a given dataset into a domain topic and to extract hidden links between its data attributes and data attributes from other datasets. The proposed approach has been developed in the frame of CRC AquaDiva ( ) and has been applied to its datasets.

Alsayed Algergawy, Hamdi Hamed, Birgitta König-Ries
RaiseWikibase: Fast Inserts into the BERD Instance

We create a knowledge graph of German companies in order to facilitate research with Business, Economic and Related Data (BERD), both modern and historical. For the implementation we chose Wikibase, but the wrappers of the Wikibase API turned out to be slow for filling it with millions of entities. This work presents the open source tool RaiseWikibase for speeding up data filling and knowledge graph construction by inserting data directly into the database. We test its performance for creating the items and wikitexts and share a reusable example for knowledge graph construction.

Renat Shigapov, Jörg Mechnich, Irene Schumm
Do Judge an Entity by Its Name! Entity Typing Using Language Models

The entity type information in a Knowledge Graph (KG) plays an important role in a wide range of applications in Natural Language Processing such as entity linking, question answering, relation extraction, etc. However, the available entity types are often noisy and incomplete. Entity Typing is a non-trivial task if enough information is not available for the entities in a KG. In this work, neural language models and a character embedding model are exploited to predict the type of an entity from only the name of the entity without any other information from the KG. The model has been successfully evaluated on a benchmark dataset.

Russa Biswas, Radina Sofronova, Mehwish Alam, Nicolas Heist, Heiko Paulheim, Harald Sack
Towards a Domain-Agnostic Computable Policy Tool

Policies are often crucial for decision-making in a wide range of domains. Typically they are written in natural language, which leaves room for different individual interpretations. In contrast, computable policies offer standardization for the structures that encode information, which can help decrease ambiguity and variability of interpretations. Sadly, the majority of computable policy frameworks are domain-specific or require tailored customization, limiting potential applications of this technology. For this reason, we propose ADAPT, a domain-agnostic policy tool that leverages domain knowledge, expressed in knowledge graphs, and employs W3C standards in semantics and provenance to enable the construction, visualization, and management of computable policies that include domain knowledge to reduce terminology inconsistencies, and augment the policy evaluation process.

Mitchell Falkow, Henrique Santos, Deborah L. McGuinness
Towards an Evaluation Framework for Expressive Stream Reasoning

Stream Reasoning, and more particularly RDF Stream Processing (RSP), has focused on processing data streams in a timely manner, while expressive reasoning techniques, such as OWL2 DL, allow to fully model and interpret their domain knowledge. However, expressive reasoning techniques have thus far mostly focused on static data, as it tends to become slow with growing datasets. Expressive Stream Reasoning aims to combine these fields and evaluate expressive reasoning techniques in a timely manner over volatile data streams through various reasoning optimizations. Both expressive reasoning and RSP have benchmarks and frameworks for evaluating and comparing proposed solutions. However, no benchmarks or evaluation frameworks for Expressive Stream Reasoning are currently available.In this paper, we propose OWL2Streams, a resource framework for the evaluation of Expressive Stream Reasoning solutions. We identified challenges and opportunities for optimizations when dealing with expressive reasoning over the combination of streams and static data. OWL2Streams proposes three streaming scenarios, each tackling different challenges.

Pieter Bonte, Filip De Turck, Femke Ongenae
Schema-Backed Visual Queries over Europeana and Other Linked Data Resources

We describe and demonstrate the process of extracting a data-driven schema of the Europeana cultural heritage Linked data resource (with actual data classes, properties and their connections, and cardinalities) and application of the extracted schema to create a visual query environment over Europeana. The extracted schema information allows generating SHACL data shapes describing the actual data endpoint structure. The schema extraction process can be applied also to other data endpoints with a moderate data schema size and a potentially large data triple count, as e.g., British National Bibliography Linked data resource.

Kārlis Čerāns, Jūlija Ovčiņņikova, Uldis Bojārs, Mikus Grasmanis, Lelde Lāce, Aiga Romāne
CLiT: Combining Linking Techniques for Everyone

While the path in the field of Entity Linking (EL) has been long and brought forth a plethora of approaches over the years, many of these are exceedingly difficult to execute for purposes of detailed analysis. In many cases, implementations are available, but far from being a plug-and-play experience. We present Combining Linking Techniques (CLiT), a framework with the purpose of executing singular linking techniques and complex combinations thereof, with a higher degree of reusability, reproducibility and comparability of existing systems in mind. Furthermore, we introduce protocols for the exchange of sub-pipeline-level information with existing and novel systems for heightened out-of-the-box compatibility. Among others, our framework may be used to consolidate multiple systems in combination with meta learning approaches and increase support for backwards compatibility of existing benchmark annotation systems.

Kristian Noullet, Samuel Printz, Michael Färber
SANTé: A Light-Weight End-to-End Semantic Search Framework for RDF Data

Natural language interfaces are one of the most powerful technologies to enable content access. It is a diverse and thriving topic that tackles a multitude of challenges ranging from designing better ranking models to user interfaces. Developing or adapting search engines is a very time-demanding and resource-consuming task. We present SANTé, a semantic search framework that facilitates publishing, querying, and browsing RDF data sets. We show the different interfaces implemented by SANTé through guided steps from raw RDF data to the search result using keyword queries. We demonstrate how SANTé can be used to publish and consume RDF data.Repository: License: FOAF demo: Pokémon demo:

Edgard Marx, André Valdestilhas, Hannah Beck, Tommaso Soru
Coverage-Based Summaries for RDF KBs

As more and more data become available as linked data, the need for efficient and effective methods for their exploration becomes apparent. Semantic summaries try to extract meaning from data, while reducing its size. State of the art structural semantic summaries, focus primarily on the graph structure of the data, trying to maximize the summary’s utility for query answering, i.e. the query coverage. In this poster paper, we present an algorithm, trying to maximize the aforementioned query coverage, using ideas borrowed from result diversification. The key idea of our algorithm is that, instead of focusing only to the “central” nodes, to push node selection also to the perimeter of the graph. Our experiments show the potential of our algorithm and demonstrate the considerable advantages gained for answering larger fragments of user queries.

Giannis Vassiliou, Georgia Troullinou, Nikos Papadakis, Kostas Stefanidis, Evangelia Pitoura, Haridimos Kondylakis
Named Entity Recognition as Graph Classification

Injecting real-world information (typically contained in Knowledge Graphs) and human expertise into an end-to-end training pipeline for Natural Language Processing models is an open challenge. In this preliminary work, we propose to approach the task of Named Entity Recognition, which is traditionally viewed as a Sequence Labeling problem, as a Graph Classification problem, where every word is represented as a node in a graph. This allows to embed contextual information as well as other external knowledge relevant to each token, such as gazetteer mentions, morphological form, and linguistic tags. We experiment with a variety of graph modeling techniques to represent words, their contexts, and external knowledge, and we evaluate our approach on the standard CoNLL-2003 dataset. We obtained promising results when integrating external knowledge through the use of graph representation in comparison to the dominant end-to-end training paradigm.

Ismail Harrando, Raphaël Troncy
Exploiting Transitivity for Entity Matching

The goal of entity matching in knowledge graphs is to identify sets of entities that refer to the same real-world object. Methods for entity matching in knowledge graphs, however, produce a collection of pairs of entities claimed to be duplicates. This collection that represents the sameAs relation may fail to satisfy some of its structural properties such as transitivity. We show that an ad-hoc enforcement of transitivity on the set of identified entity pairs may decrease precision. We therefore propose a methodology that starts with a given similarity measure, generates a set of entity pairs, and applies cluster editing to enforce transitivity, leading to overall improved performance.

Jurian Baas, Mehdi M. Dastani, Ad J. Feelders
The Nuremberg Address Knowledge Graph

The research of European history across various time layers gives insights about the development of the European cultural identity. Nuremberg as one of the great European metropolises during the Middle Ages experienced a number of transformations throughout the centuries. Within the TRANSRAZ research project, Nuremberg and the development of its architecture and culture is recreated from the 17th to the 21st century. It will be available for researchers and the public by means of an interactive 3D environment. Goal of this poster paper is to discuss the ongoing work of connecting heterogeneous historical data from sources previously hidden in archives to the 3D model using knowledge graphs for a scientifically accurate exploration of Nuremberg. The contribution of this paper is the Nuremberg Address Knowledge Graph (NA-KG) which contains information of people and organizations in Nuremberg from unstructured data of Nuremberg address books.

Oleksandra Bruns, Tabea Tietz, Mehdi Ben Chaabane, Manuel Portz, Felix Xiong, Harald Sack
SaGe-Path: Pay-as-you-go SPARQL Property Path Queries Processing Using Web Preemption

SPARQL property path queries allow to write sophisticated navigational queries on knowledge graphs (KGs). However, the evaluation of these queries on online KGs are often interrupted by fair use policies, returning only partial results. SaGe-Path addresses this issue by relying on the Web preemption and the concept of Partial Transitive Closure (PTC). Under PTC, the graph exploration for SPARQL property path queries is limited to a predefined depth. When the depth limit is reached, frontier nodes are returned to the client. A PTC-client is then able to reuse frontier nodes to continue the exploration of the graph. In this way, SaGe-Path follows a pay-as-you-go approach to evaluate SPARQL property path queries. This demonstration shows how queries that do not complete on the public Wikidata SPARQL endpoint can complete using SaGe-Path. An extended user-interface provides real-time visualization of all SaGe-Path internals, allowing to understand the approach overheads and the effects of different parameters on performance. SaGe-Path demonstrates how complex SPARQL property path queries can be efficiently evaluated online with guaranteed complete results.

Julien Aimonier-Davat, Hala Skaf-Molli, Pascal Molli
Ontology for Informatics Research Artifacts

The IRAO ontology, as a new contribution to the network of ontologies for the scholarly domain, aims to model the most tangible aspect of research in computing disciplines – the research artifacts. It consists of parts focusing on the concepts of researcher, research artifact classification, research artifact meta information, relationships between artifacts, and research artifact quality evaluation benchmarks that are used to express the quality and maturity of each research artifact. We describe the ontology design requirements using competency questions and the evaluation of the ontology by the same questions that helped in defining the concept domain coverage.

Viet Bach Nguyen, Vojtěch Svátek
Non-named Entities – The Silent Majority

Knowledge Bases (KBs) usually contain named entities. However, the majority of entities in natural language text are not named. In this position paper, we first study the nature of these entities. Then we explain how they could be represented in KBs. Finally, we discuss open challenges for adding non-named entities systematically to KBs.

Pierre-Henri Paris, Fabian Suchanek
Unsupervised Relation Extraction Using Sentence Encoding

Relation extraction between two named entities from unstructured text is an important natural language processing task. In the absence of labelled data, semi-supervised and unsupervised approaches are used to extract relations. We present a novel approach that uses sentence encoding for unsupervised relation extraction. We use a pre-trained, SBERT based model for sentence encoding. Our approach classifies identical sentences using a clustering algorithm. These sentences are used to extract relations between two named entities in a given text. The system calculates a confidence value above a certain threshold to avoid semantic drift. The experimental results show that without any explicit feature selection and independent of the size of the corpus, our proposed approach achieves a better F-score than state-of-the-art unsupervised models.

Manzoor Ali, Muhammad Saleem, Axel-Cyrille Ngonga Ngomo
evoKGsim+: A Framework for Tailoring Knowledge Graph-Based Similarity for Supervised Learning

Knowledge graphs represent an unparalleled opportunity for machine learning, given their ability to provide meaningful context to the data through semantic representations. However, general-purpose knowledge graphs may describe entities from multiple perspectives, with some being irrelevant to the learning task. Despite the recent advances in semantic representations such as knowledge graph embeddings, existing methods are unsuited to tailoring semantic representations to a specific learning target that is not encoded in the knowledge graph.We present evoKGsim+, a framework that can evolve similarity-based semantic representations for learning relations between knowledge graph entity pairs, which are not encoded in the graph. It employs genetic programming, where the evolutionary process is guided by a fitness function that measures the quality of relation prediction. The framework combines several taxonomic and embedding similarity measures and provides several baseline evaluation approaches that emulate domain expert feature selection and optimal parameter setting.

Rita Torres Sousa, Sara Silva, Catia Pesquita
Extraction of Union and Intersection Axioms from Biomedical Text

Many ontology, especially the ones created automatically by the ontology learning systems, have only shallow relationships between the concepts, i.e., simple subclass relations. Expressive axioms such as the class union and intersection are not part of the ontology. These expressive axioms make the ontology rich and play an essential role in the performance of downstream applications. However, such relations can generally be found in the text documents. We propose a mechanism and discuss our initial results in extracting union and intersection axioms from biomedical text using entity linking and taxonomic tree search.

Nikhil Sachdeva, Monika Jain, Raghava Mutharaju

PhD Symposium Track Papers

Implementing Informed Consent with Knowledge Graphs

The GDPR legislation has brought to light one’s rights and has highlighted the importance of consent, which has caused a major shift in how data processing and sharing are handled. Data sharing has been a popular research topic for many years, however, a unified solution for the transparent implementation of consent, in compliance with GDPR that could be used as a standard, has not been presented yet. This research proposes a solution for implementing informed consent for sensor data sharing in compliance with GDPR with semantic technology, namely knowledge graphs. The main objectives are to model the life cycle of informed consent (i.e. the request, comprehension, decision and use of consent) with knowledge graphs so that it is easily interpretable by machines, and to graphically visualise it to individuals in order to raise legal awareness of what it means to consent and the implications that follow.

Anelia Kurteva
Improving Decision Making Using Semantic Web Technologies

With the rapid advance of technology, we are moving towards replacing humans in decision making–the employment of robotics and computerised systems for production and delivery and autonomous cars in the travel sector. The focus is placed on the use of techniques, such as machine learning and deep learning. However, despite advances in machine learning and deep learning, they are incapable of modelling the relationships that are present in the real world, which are necessary for making a decision. For example, automating sociotechnical systems requires an understanding of both human and technological aspects and how they influence one another. Using machine learning, we can not model the relationships of a sociotechnical systems. Semantic Web technologies, which is based on the concept of linked-data technology, can represent relationships in a more realistic way like in the real world, and be useful to make better decisions. The study looks at the use of Semantic Web technologies, namely ontologies and knowledge graphs to improve decision making process.

Tek Raj Chhetri
Ontological Formalisation of Mathematical Equations for Phenomic Data Exploitation

In recent years, plant phenomics community has adopted Semantic Web technologies in order to harmonise heterogeneous, multi-scale and multi-source datasets. Semantic Web provides inference services for representing logic relationships in an unambiguous, homogeneous and clean manner, which enhances data harmonisation. However, mathematical relationships involving numerical attributes are poorly formalised, despite the fact that they are supported for a theoretical and well-defined structure. For instance, whilst unit ontologies (e.g. UO, OM, QUDT) provide relationships and annotations to perform unit conversion, they are not effectively used for automating the integration of heterogeneous measurements. Here we propose an ontological framework for representing mathematical equations supporting the automatised use of inference services, metadata, domain ontologies, and the internal structure of mathematical equations. This approach is evaluated using two plant phenomics case studies involving the calculation of unit conversions and thermal time.

Felipe Vargas-Rojas
Identifying Events from Streams of RDF-Graphs Representing News and Social Media Messages

Identifying news events and relating current news to past events or already identified ones is an open challenge for news agencies. In this paper, I propose a study to identify events from semantic RDF graph representations of real-time and big data streams of news and pre-news. The proposed solution must provide acceptable accuracy over time and consider the requirements of incremental clustering, big data and real-time streams. To design a solution for identifying events, I want to study which clustering approaches are best for this purpose including methods for clustering RDF graphs using machine learning and “classical” algorithmic approaches. I also present three different evaluation approaches.

Marc Gallofré Ocaña
Towards Visually Intelligent Agents (VIA): A Hybrid Approach

Service robots can undertake tasks that are impractical or even dangerous for us - e.g., industrial welding, space exploration, and others. To carry out these tasks reliably, however, they need Visual Intelligence capabilities at least comparable to those of humans. Despite the technological advances enabled by Deep Learning (DL) methods, Machine Visual Intelligence is still vastly inferior to Human Visual Intelligence. Methods which augment DL with Semantic Web technologies, on the other hand, have shown promising results. In the lack of concrete guidelines on which knowledge properties and reasoning capabilities to leverage within this new class of hybrid methods, this PhD work provides a reference framework of epistemic requirements for the development of Visually Intelligent Agents (VIA). Moreover, the proposed framework is used to derive a novel hybrid reasoning architecture, to address real-world robotic scenarios which require Visual Intelligence.

Agnese Chiatti
Using Knowledge Graphs for Machine Learning in Smart Home Forecasters

Internet of Things (IoT) brings together heterogeneous data from smart devices in smart homes. Smart devices operate within different platforms, but ontologies can be used to create a common middle ground that allows communications between these smart devices outside of those platforms. The data communicated by the smart devices can be used to train the prediction algorithms used in forecasters. This research will first focus on the creation of a mapping to transform IoT data into a knowledge graph than can be used in the common middle ground and investigate the effect of using that IoT knowledge graph data as input for prediction algorithms. Experiments to determine the impact of incorporating other related information in the training of the prediction algorithms will be performed by using external datasources that can be linked to the knowledge graph and by using federated learning over IoT data from other smart homes. Initial results on the transformation mapping of IoT data to an ontology is presented.

Roderick van der Weerdt
Stigmergic Multi-Agent Systems in the Semantic Web of Things

Intelligent, autonomous agents are still not available in the Semantic Web at large scale today. Also the fields of Semantic Web and Multi-Agent Systems are not working together very closely although they could profit much from each other. Existing approaches merely use the Web as a transport layer and are not properly aligned to the architectural style of the Web. The Internet of Things, which would be very useful for agents to act upon Things in the real world, on the other side is very fragmented and not easily inter-operable. The Web of Things has emerged as an approach to use the (Semantic) Web as an application layer for Things. It is still unclear however, how agents on the WoT should look like. We propose that the Semantic Web is a suitable integration layer, both for agents and Things. We investigate how a Multi-Agent System in the Semantic Web of Things can be build by utilizing simple reflex agents and the communication paradigm of stigmergy. We map Things and artifacts to Web resources that are managed by a Web server and provide affordances to agents through hypermedia.

Daniel Schraudner
Towards an Ontology for Propaganda Detection in News Articles

The proliferation of mis/disinformation in the media has had a profound impact on social discourse and politics in the United States. Some argue that democracy itself is threatened by the lies, chicanery, and flimflam - in short, propaganda - emanating from the highest pulpits, podiums, and soapboxes in the land. Propaganda differs from mis/disinformation in that it need not be false, but instead, it relies on rhetorical devices which aim to manipulate the audience into a particular belief or behavior. While falsehoods can be debunked, albeit with disputable efficacy, beliefs are harder to cut through. The detection of “Fake News” has received a lot of attention recently with some impressive results, however, propaganda detection remains challenging. This proposal aims to further the research into propaganda detection by constructing an ontology with this specific goal in mind, while drawing from multiple disciplines within Computer Science and the Social Sciences.

Kyle Hamilton

Industry Track Papers

A Virtual Knowledge Graph for Enabling Defect Traceability and Customer Service Analytics

In this paper, we showcase the implementation of a semantic information model and a virtual knowledge graph at ZF Friedrichshafen AG company, with two main goals in mind: 1) integration of heterogeneous data sources following a pay-as-you-go approach; and the 2) combination core domain concepts from ZF’s production line with meta-data of its internal data sources. We employ the developed semantic information model in two use cases, defect traceability and customer service, demonstrating and discussing the benefits and opportunities provided by following an agile semantic virtual integration approach.

Nico Wilhelm, Diego Collarana, Jens Lehmann
Constructing Micro Knowledge Graphs from Technical Support Documents

Short technical support pages such as IBM Technotes are quite common in technical support domain. These pages can be very useful as the knowledge sources for technical support applications such as chatbots, search engines and question-answering (QA) systems. Information extracted from documents to drive technical support applications is often stored in the form of Knowledge Graph (KG). Building KGs from a large corpus of documents poses a challenge of granularity because a large number of entities and actions are present in each page. The KG becomes virtually unusable if all entities and actions from these pages are stored in the KG. Therefore, only key entities and actions from each page are extracted and stored in the KG. This approach however leads to loss of knowledge represented by entities and actions left out of the KG as they are no longer available to graph search and reasoning functions.We propose a set of techniques to create micro knowledge graph (micrograph) for each of such web pages. The micrograph stores all the entities and actions in a page and also takes advantage of the structure of the page to represent exactly in which part of that page these entities and actions appeared, and also how they relate to each other. These micrographs can be used as additional knowledge sources by technical support applications. We define schemas for representing semi-structured and plain text knowledge present in the technical support web pages. Solutions in technical support domain include procedures made of steps. We also propose a technique to extract procedures from these webpages and the schemas to represent them in the micrographs. We also discuss how technical support applications can take advantage of the micrographs.

Atul Kumar, Nisha Gupta, Saswati Dana
Use Case: Ontologies and RDF-Star for Knowledge Management

Our client in this case study is a software company which develops, publishes, and distributes video games for consoles, PCs, smartphones, and tablets in both physical and digital formats. They also create educational and cultural software, cartoons, and literary, cinematographic, and television works. It owns several brands and a diversified portfolio of franchises.The client required a centralized vocabulary management software platform to provide standardized concepts across a decentralized, global organization to find, browse, and discover enterprise content. They needed the ability to push vocabularies out to consuming systems and users while also allowing users to suggest new concepts without requiring them to log in to the taxonomy and ontology management software.In addition to the out-of-the-box Graphite ontology management software functionality, the client required bespoke work in the system and dedicated API connectors which became part of the common code base for all versions going forward. Their requirements presented the opportunity for Synaptica to explore uses for the new specification, RDF-star (Arndt et al. 2021), in our implementation. As a new and developing specification in RDF graph databases, the use of RDF-star is groundbreaking work for commercial enterprise ontology management systems.

Bob Kasenchak, Ahren Lehnert, Gene Loh
The Semantic Web: ESWC 2021 Satellite Events
Ruben Verborgh
Dr. Anastasia Dimou
Aidan Hogan
Claudia d'Amato
Dr. Ilaria Tiddi
Arne Bröring
Dr. med. Simon Mayer
Femke Ongenae
Assoc. Prof. Riccardo Tommasini
Dr. Mehwish Alam
Copyright Year
Electronic ISBN
Print ISBN