Knowledge Engineering and Knowledge Management

The process of extracting, structuring, and organizing knowledge from one or multiple data sources and preparing it for the Semantic Web requires a dedicated class of systems. They enable processing large and originally heterogeneous data sources and capturing new knowledge. Offering existing data as Linked Data increases its shareability, extensibility, and reusability. However, using Linking Data as a means to represent knowledge can be easier said than done. In this tutorial, we elaborate on the importance of semantically annotating data and how existing technologies facilitate their mapping to Linked Data. We introduce [R2]RML languages to generate Linked Data derived from different heterogeneous data formats –e.g., DBs, XML, or JSON– and from different interfaces –e.g., files or Web apis. Those who are not Semantic Web experts can annotate their data with the RMLEditor, whose user interface hides all underlying Semantic Web technologies to data owners. Last, we show how to easily publish Linked Data on the Web as Triple Pattern Fragments. As a result, participants, independently of their knowledge background, can model, annotate and publish data on their own.

This paper presents work in progress on an algorithm to track and identify changes in the vocabulary used to describe particular concepts over time, with emphasis on treating concepts as distinct from changes in word meaning. We apply the algorithm to word vectors generated from Google Books n-grams from 1800–1990 and evaluate the induced networks with respect to their flexibility (robustness to changes in vocabulary) and stability (they should not leap from topic to topic). We also describe work in progress using the British National Biography Linked Open Data Serials to construct a “ground truth” evaluation dataset for algorithms which aim to detect shifts in the vocabulary used to describe concepts. Finally, we discuss limitations of the proposed method, ways in which the method could be improved in the future, and other considerations.

Semantic drift is an active research field, which aims to identify and measure changes in ontologies across time and versions. Yet, only few practical methods have emerged that are directly applicable to Semantic Web constructs, while the lack of relevant applications and tools is even greater. This paper presents the findings, current limitations and lessons learned throughout the development and the application of a novel software tool, developed in the context of the PERICLES FP7 project, which integrates currently investigated methods, such as text and structural similarity, into the popular ontology authoring platform, Protégé. The graphical user interface provides knowledge engineers and domain experts with access to methods and results without prior programming knowledge. Its applicability and usefulness are validated through two proof-of-concept scenarios in the domains of Web Services and Digital Preservation; especially the latter is a field where such long-term insights are crucial.

Statistical Natural Language Processing (NLP) techniques allow to quantify lexical semantic change using large text corpora. Word-level results of these methods can be hard to analyse in the context of sets of semantically or linguistically related words. On the other hand, structured knowledge sources represent semantic relationships explicitly, but ignore the problem of semantic change. We aim to address these limitations by combining the statistical and symbolic approach: we enrich WordNet, a structured lexical database, with quantitative lexical change scores provided by HistWords, a dataset produced by distributional NLP methods. We publish the result as Linked Open Data and demonstrate how queries on the combined dataset can provide new insights.

This paper presents the Learning Scorecard (LS), a platform that enables students to monitor their learning progress in a Higher Education course during the semester, generating the data that will also support the ongoing supervision of the class performance by the course coordinator. The LS uses gamification techniques to increase student engagement with the course. Business Intelligence best practices are also applied to provide an analytical environment for student and faculty to monitor course performance. This paper describes the initial design of the LS, based on a Balanced Scorecard approach, and the prototype version of the platform, currently in use by graduate and undergraduate students in the fall semester of 2016–2017.

The heterogeneity and the high volume of data on the Web are the main features that make it a promoter field of knowledge engineering for researcher. However, the user is getting lost towards the diversity of information on the Web. In this paper, we propose an approach, to assist user, based on Semantic Web technologies. Our scenario is focused on the field of education and in particular higher education. This choice is motivated by the diversity of information sources where the student is dispersed.

In the education field, in order to achieve learning goals, it is necessary to define learning paths that foresee a gradual and incremental acquisition of certain knowledge and skills that students should acquire. In this paper we analyze the educational progressions of the French educational system, we show how to formalize them through a web ontology and how to perform knowledge extraction from the official texts describing them to automate the population of such an ontology.

We propose an approach to semantically enrich metadata records of satellite imagery with external data. As a result we are able the identify relevant images using a larger set of matching criteria. Conventional methods for annotating data sets are usually based on metadata records (with attributes such as title, provider, access mode, and spatio-temporal characteristics), which offer a narrow view of the world. Enriching metadata with contextual information (e.g., the region depicted in the image has been recently affected by extreme weather) requires formalizing spatio-temporal relationships between metadata records and external data sources. Semantic technologies play a key role in such scenarios by providing an infrastructure based on RDF and ontologies.

The need of handling semantic heterogeneity of resources is a key problem of the Semantic Web. State of the art techniques for ontology matching are the key technology for addressing this issue. However, they only partially exploit the natural language descriptions of ontology entities and they are mostly unable to find correspondences between entities having different logical types (e.g. mapping properties to classes). We introduce a novel approach aimed at finding correspondences between ontology entities according to the intensional meaning of their models, hence abstracting from their logical types. Lexical linked open data and frame semantics play a crucial role in this proposal. We argue that this approach may lead to a step ahead in the state of the art of ontology matching, and positively affect related applications such as question answering and knowledge reconciliation.

IKEYS is an interactive and cooperative system aimed to query corporate linked data that allows users define explicit and unambiguous queries. Users first express their information needs through coarse keyword queries (‘track J. Morrison 1971’) that may then be refined with explicit projection and selection statements involving comparison operators and aggregation functions (‘title of track composed by J. Morrison before 1971’). This demonstration shows how intuitive and efficient IKEYS is to find the exact answer to enhanced keyword queries.

The verification of the soundness and consistency of data acquisition plans is an important requirement in the loading of data generated by physical and social devices. In this paper we discuss these properties in the context of the StreamLoader system.

We present Joyce, a scalable tool for identifying and assembling relevant (pieces of) ontologies from a repository of source ontologies, thus enabling the effective and efficient reuse of formalized domain knowledge. Joyce includes a conceptual filter to identify relevant classes, minimizes unintended redundancies, i.e. concept duplicates, and excludes knowledge considered irrelevant for the specific conceptual design task.

Annotating semantic data with metadata is becoming more and more important to provide information about the statements. While there are solutions to represent temporal information about a statement, a general annotation framework which allows representing more contextual information is needed. In this paper, we extend the 4dFluents ontology by Welty and Fikes to any dimension of context.

We face the issue of formally modeling roles in the semantic representation of events: we propose a distinction between thematic roles and social roles, and we show that they have important ontological differences, suggesting distinct formal representations. We apply our approach in the context of the Harlock’900 project, including the definition of thematic roles as binary properties in the HERO ontology.

This paper introduces StreamJess, a Stream Reasoning system that layers on top of a state of the art query processing system such as C-SPARQL to enable closed-world, non-monotonic and time-aware reasoning with Jess rules. The system is validated in the water quality monitoring domain by demonstrating water bodies’ classification and pollution sources investigation.

RDF aims at being the universal abstract data model for structured data on the Web. While there is effort to convert data in RDF, the vast majority of data available on the Web does not conform to RDF. Indeed, exposing data in RDF, either natively or through wrappers, can be very costly. In this context, transformation or mapping languages that define generation of RDF from non-RDF data represent an efficient solution. Furthermore, the declarative aspect of these solutions makes them easy to adapt to any change in the input data model, or in the output knowledge model. This paper introduces a novel such transformation language (SPARQL-Generate), an extension of SPARQL for querying not only RDF datasets but also documents in arbitrary formats. Its implementation on top of Apache Jena currently covers use cases from related work and more, and enables to query and transform web documents in XML, JSON, CSV, HTML, CBOR, and plain text with regular expressions.

Librarians and music professionals often us complex models and ontologies such as FRBRoo to represent music metadata. As a consequence, this metadata is not easily consumable by general search engines or external web applications. This paper presents a methodology, composed of a set of recipes, for mapping a complex ontology to a simpler model, namely Schema.org.

Knowledge acquisition is a central issue of the Semantic Web. Knowledge cannot always be automatically extracted from existing data, thus contributors have to make efforts to create new data. In this paper, we propose FORMULIS, a dynamic form-based interface designed to make RDF data authoring easier. FORMULIS guides contributors through the creation of RDF data by suggesting fields and values according to the previously filled fields and the previously created resources.

LOINC® and SNOMED CT® are two of the most used biomedical terminology standards to conjointly describe medical laboratory data into patient Electronic Health Records. The institutions owning them entered in a collaboration 4 years ago. The intention was to provide alignments between LOINC® and SNOMED CT® in order to improve query and aggregation of patient data. This work brings input on the LOINC—SNOMED CT alignment effort: (i) we developed algorithms aiding to align LOINC® and SNOMED CT® efficiently and (ii) we demonstrated the benefits of the SNOMED CT® conceptual hierarchy and tests model to query data initially coded in LOINC®.

During the last decade, the Linked Open Data cloud has grown with much enthusiasm and a lot organizations are publishing their data as Linked Data. However, it is not evident whether enough efforts have been invested in maintaining those data or ensuring their quality. Data quality, defined as “fitness for use”, is an important aspect for Linked Data to be useful. Data consumers use quality indicators to decide whether or not to use a dataset in a given use case, which makes quality assessment of Linked Data an important activity. Accessibility, which is defined as the degree to which the data can be accessed, is a highly relevant quality characteristic to achieve the benefits of Linked Data. In this demo paper presents LD Sniffer, a web-based open source tool for performing quality assessment on the accessibility of Linked Data. It generates unambiguous and comparable assessment results with explicit semantics by defining both quality metrics as well as assessment results in RDF using the W3C Data Quality vocabulary. LD-Sniffer is also distributed as a Docker image improving ease of use with zero configurations.

Stream Processing/Reasoning, an active research topic [5], has been picked up by different communities which developed a diversity of stream processors/reasoners. This however makes empirical evaluation and comparison of these engines a non-trivial task [4]. Different classes of those engines work on different formats of input data, use different languages to formulate queries, evaluate these queries using different semantics and produce different formats of output. To be able to compare such engines, a benchmarking framework that can cope with this wide diversity is needed.

Ontologies are modeled using specific concepts of the knowledge domain as well as using generic concepts. Life science ontologies like MeSH, Agrovoc, and DrugBank are helpful for searching through large corpora. The distinct linkage to either the agricultural domain or the medical domain cannot be resolved for generic concepts that were created when modeling the domain. In information retrieval, it is required to filter knowledge resources for domain specific concepts in order to avoid noise in search results caused by generic concepts. Here, we present an exploratory step towards evaluating concept frequencies amongst different knowledge domains when employing ontologies in the retrieval on a large corpus.

In this demo paper we introduce SPACIN, one of the main tools used in the OpenCitations Project for producing RDF-based citation data from information available in trusty sources, such as Europe PubMed Central, Crossref, and ORCID.

In this paper we present a relation validation method for KBP slot filling task by exploring some graph features to classify the candidate slot fillers as correct or incorrect. The proposed features with voting feature collectively performs better than the baseline voting feature.

Probabilistic Inductive Logic Programming (PILP) is gaining attention for its capability of modeling complex domains containing uncertain relationships among entities. Among PILP systems, cplint provides inference and learning algorithms competitive with the state of the art. Besides parameter learning, cplint provides one of the few structure learning algorithms for PLP, SLIPCOVER. Moreover, an online version was recently developed, cplint on SWISH, that allows users to experiment with the system using just a web browser. In this demo we illustrate cplint on SWISH concentrating on structure learning with SLIPCOVER. cplint on SWISH also includes many examples and a step-by-step tutorial.

This paper presents an ongoing work on sentiment analysis in the financial domain and explores an approach to identifying sentiment orientations of words for a given financial index. The proposed approach takes advantage of the movement of the given financial index and employs an information theoretic measure for estimating sentiment orientation of word combinations in an efficient way. Results on preliminary experiments are reported.

For a given set of URIs, finding their common graph patterns may provide useful knowledge. We present an algorithm searching for the best patterns while trying to extend the set of relevant URIs. It involves interaction with the user in order to supervise extension of the set.

Swift Linked Data Miner (SLDM) is a data mining algorithm capable to infer new knowledge and thus extend an ontology by mining a Linked Data dataset. We present an extension to WebProtégé providing SLDM capabilities in a web browser. The extension is open source and readily available to use.

Linked Datasets typically change over time, and knowledge of this historical information can be useful. This makes the storage and querying of Dynamic Linked Open Data an important area of research. With the current versioning solutions, publishing Dynamic Linked Open Data at Web-Scale is possible, but too expensive. We investigate the possibility of using the low-cost Triple Pattern Fragments (tpf) interface to publish versioned Linked Open Data. In this paper, we discuss requirements for supporting versioning in the tpf framework, on the level of the interface, storage and client, and investigate which trade-offs exist. These requirements lay the foundations for further research in the area of low-cost, Web-Scale dynamic Linked Open Data publication and querying.

In this paper we present the DKA-robo framework, where a mobile robot is used to update the statements of a knowledge base that have lost validity in time. Managing the dynamic information of knowledge bases constitutes a key issue in many real-world scenarios, because constantly reevaluating data requires efforts in terms of knowledge acquisition and representation. Our solution to such a problem is to use RDF and SPARQL to represent and manage the time validity of information, combined with a robot acting as a mobile sensor which updates the outdated statements in the knowledge base, therefore always guaranteeing time-valid results against user queries. This demo shows the implementation of our approach in the working environment of our research lab, where a robot is used to sense temperature, humidity, wifi-signal and number of people on demand, updating the lab knowledge base with time-valid information.

We present a first version of a system for selecting chemical publications for inclusion in a chemistry information database. This database, Reaxys (https://www.elsevier.com/solutions/reaxys), is a portal for the retrieval of structured chemistry information from published journals and patents. There are three challenges in this task: (i) Training and input data are highly imbalanced; (ii) High recall ($${\ge }95\%$$) is desired; and (iii) Data offered for selection is numerically massive but at the same time, incomplete. Our system successfully handles the imbalance with the undersampling technique and achieves relatively high recall using chemical named entities as features. Experiments on a real-world data set consisting of 15,822 documents show that the features of chemical named entities boost recall by $$8\%$$ over the usual n-gram features being widely used in general document classification applications. For fostering research on this challenging topic, a part of the data set compiled in this paper can be requested.

Given a knowledge-based system running virtually forever able to acquire and automatically store new open-domain knowledge, one of the challenges is to evolve by continuously integrating new knowledge. This needs to be done while handling conflicts, redundancies and linking existing knowledge to the incoming one. We refer to this task with the name Knowledge integration. In this paper we define the problem by discussing its challenges, we propose an approach for tackling the problem, and, we suggest a methodology for the evaluation of results.

Biomedical Knowledge Organization Systems (KOS) play a key role in enriching information in order to make them machine understandable. This is done through semantic annotation which consists in the association of concept labels taken from KOS with pieces of digital information taken from the source to annotate. However, the dynamic nature of these KOS directly impacts on the annotations, creating a mismatch between the enriched data and the concept labels. This PhD study addresses the evolution of semantic annotations due to the evolution of KOS and aims at proposing an approach to automatize the maintenance of semantic annotations.

Art historical photo archives that want to expose their data in Linked Open Data need to rely on shareable models. Merging possibly contradictory information may affect data reliability. In this paper are introduced two ontologies, i.e. F Entry Ontology and OA Entry Ontology, which provide a complete description of items related to Photography and Arts domains and address the description of questionable information provided by different institutions. A preliminary analysis of the Zeri’s photo archive was performed for guiding the creation of the ontologies and the mapping of all the partners’ metadata schemas.

ETL (Extraction-Transform-Load) tools, traditionally developed to operate offline, need to be enhanced to deal with various, fast, big and fresh data and be executed on the edge of the network during the acquisition process. In this dissertation we wish to develop facilities that from one side make easy, scalable and controllable the development of data acquisition plans that can be executed on the edge of the network during loading and transmission. From the other side, we wish to deal with the variety of the data and verify when the developed data acquisition plans adhere to the common semantics adopted in the Domain Ontology. These facilities are included in StreamLoader, a web application tailored for the specification and monitoring of sensor data acquisition plans.

Industry 4.0 initiatives have fostered the definition of different standards, e.g., AutomationML or OPC UA, allowing for the specification of industrial objects and for machine-to-machine communication in Smart Factories. Albeit facilitating interoperability at different steps of the production life-cycle, the information models generated from these standards are not semantically defined, making the semantic data integration a challenging problem. We tackle the problems of integrating data from documents specified either using the same or different Industry 4.0 standards, and propose a rule-based framework that combines deductive databases and Semantic Web technologies to effectively solve these problems. As a proof-of-concept, we have developed a Datalog-based representation for AutomationML documents, and a set of rules for identifying semantic heterogeneity problems among these documents. We have empirically evaluated our proposed framework against several benchmarks and the initial results suggest that exploiting deductive and Semantic Web techniques allows for increasing scalability, efficiency, and coherence of models for Industry 4.0 manufacturing environments.

Systematic review, a form of evidence synthesis that critically appraises existing studies on the same topic and synthesizes study results, helps reduce the evidence gap. However, keeping the systematic review up-to-date is a great challenge partly due to the difficulty in interpreting the conclusion of a systematic review. A promising approach to this challenge is to make semantic representation of the claims made in both the systematic review and the included studies it synthesizes so that it’s possible to automatically predict whether the conclusion of a systematic review changes given a new study. In this dissertation work, we developed a taxonomy to represent knowledge claims both in systematic review and its included studies with the goal of automatically updating a systematic review. We then developed machine learning models to automatically predict a synthesized claim from claims in individual studies.

Leveraging external RDF data for OLAP analysis opens a wide variety of possibilities that enable analysts to gain interesting insights related to their businesses. While variations of statistical linked data are easily accessible to OLAP systems, exploiting non-statistical linked data, such as DBpedia, for OLAP analysis is not trivial. An OLAP system for these data should, on the one hand, take into account the big volume, heterogeneity, graph nature, and semantics of the RDF data. On the other hand, dealing with external RDF data requires a degree of self-sufficiency of the analyst, which can be met via self-service OLAP, without assistance of specialists. In this paper, we argue the need for self-service OLAP endpoints for linked RDF datasets. We review the related literature and sketch an approach. In particular, we propose the use of multidimensional schemas and analysis graphs over linked RDF datasets, which will empower users to perform self-service OLAP analysis on the linked RDF datasets.

Learning ontologies from software requirements specifications with individuals and relations between individuals to represent detailed information, such as input, condition and expected result of a requirement, is a difficult task. System specification ontologies (SSOs) can be developed from software requirement specifications to represent requirements and can be used to automate some time-consuming activities in software development processes. However, manually developing SSOs to represent requirements and domain knowledge of a software system is a time-consuming and a challenging task. The focus of this PhD is how to create ontologies semi-automatically from SRS. We will develop a framework that can be a possible solution to create semi-automatically ontologies from SRS. The developed framework will mainly be evaluated by using the constructed ontologies in the software testing process and automating a part of it. i.e. test case generation.

Dealing with ontology changes is of high importance for ontology engineers in different application domains. In some applications domain ontologies cover overlapping topics. Ontology mappings represent this overlap and allow specifying a connection between domain ontologies. Adapting the mappings when the domain ontologies are changed is a problem still lacking a thoroughly evaluated, systematic approach. This doctoral research aims at providing a new approach to this problem by comparing the inferences computed from the relevant regions of the domain ontologies and the mappings before and after changes occur. A change model can be used to categorize differences in the inferences as intentional or unintentional and propose solutions to adapt the mappings.

Knowledge models that are currently in-use for describing music metadata are insufficient to express the wealth of complex information about creative works, performances, publications, authors and performers. In this thesis, we aim to propose a method for structuring the music information coming from heterogeneous librarian repositories. In particular, we research and design an appropriate music ontology based on existing models and controlled vocabularies and we implement tools for converting and visualizing the metadata. Moreover, we research how this data can be consumed by end-users, through the development of a web application for exploring the data. We ultimately aim to develop a recommendation system that takes advantage of the richness of the data.

This multidisciplinary research aims to allow urban planners and decision makers to better understand urban and/or peri-urban migrations in order to adjust their decisions. The more specific question concerns the residential trajectories of inhabitants (i.e. the succession of their residential choices in time and space). To better understand the circumstances that lead people to move, we propose a model that takes into account the different aspects of the individuals lives (family professional, spare-time activities, etc.) and allows to elicit explanatory factors between life events (e.g. a move due to a birth to come). Our approach also leads us to design an innovative participatory software to collect and analyze descriptive and localized data on the residential trajectories between different territories of a metropolitan area.

Digitizing scholarly communication is a major challenge of our era. In this thesis, we focus particularly on facilitating the digital handling of scholarly communication metadata, i.e. bibliographic data, metadata about scientific events, courseware, projects, organizations etc. We describe these metadata domains and develop representation schemes for semantically representing respective information. We develop a conceptual lifecycle model for facilitating the management of scholarly communication data. Furthermore, we present some concrete strategies and applications for semantically representing and linking bibliographic data, for crowdsourcing and analysing events metadata and for quality assessment of opencourseware and scientific events based on this metadata.