The Semantic Web: Research and Applications

This article presents a technique for Stream Reasoning, consisting in incremental maintenance of materializations of ontological entailments in the presence of streaming information. Previous work, delivered in the context of deductive databases, describes the use of logic programming for the incremental maintenance of such entailments. Our contribution is a new technique that exploits the nature of streaming data in order to efficiently maintain materialized views of RDF triples, which can be used by a reasoner.

By adding expiration time information to each RDF triple, we show that it is possible to compute a new complete and correct materialization whenever a new window of streaming data arrives, by dropping explicit statements and entailments that are no longer valid, and then computing when the RDF triples inserted within the window will expire. We provide experimental evidence that our approach significantly reduces the time required to compute a new materialization at each window change, and opens up for several further optimizations.

Novel wireless handheld devices allow the adoption of revised and adapted discovery approaches originally devised for the Semantic Web in mobile ad-hoc networks. Nevertheless, capabilities of such devices require an accurate re-design of frameworks and algorithms to efficiently support mobile users. The paper focuses on an implementation of concept abduction and contraction algorithms in (fuzzy) \({\mathcal{ALN}}\)(D) DL settings to perform semantic matchmaking and provide logical explanation services. OWL-DL Knowledge Bases have been properly exploited to enable standard and non-standard inference services. The proposed framework has been implemented and tested in a fire hazards prevention case study: early experimental results are reported.

The sizes of datasets available as RDF (e.g., as part of the Linked Data cloud) are increasing continuously. For instance, the recent DBpedia version consists of nearly 500 millions triples. A common strategy to avoid problems that arise e.g., from limited network connectivity or lack of bandwidth is to replicate data locally, therefore making them accessible for applications without depending on a network connection. For mobile devices with limited capabilities, however, the replication and synchronization of billions of triples is not feasible. To overcome this problem, we propose an approach to replicate parts of an RDF graph to a client. Applications may then apply changes to this partial replica while being offline; these changes are written back to the original data source upon reconnection. Our approach does not require any kind of additional logic (e.g., change logging) or data structures on the client side, and hence is suitable to be applied on devices with limited computing power and storage capacity.

This paper describes a real-time routing system that implements a mobile museum tour guide for providing personalized tours tailored to the user position inside the museum and interests. The core of this tour guide originates from the CHIP (Cultural Heritage Information Personalization) Web-based tools set for personalized access to the Rijksmuseum Amsterdam collection. In a number of previous papers we presented these tools for interactive discovery of user’s interests, semantic recommendations of artworks and art-related topics, and the (semi-)automatic generation of personalized museum tours. Typically, a museum visitor could wander around the museum and get attracted by artworks outside of the current tour he is following. To support a dynamic adaptation of the tour to the current user position and changing interests, we have extended the existing CHIP mobile tour guide with a routing mechanism based on the SWI-Prolog Space package. The package uses (1) the CHIP user profile containing user’s preferences and current location; (2) the semantically enriched Rijksmuseum collection and (3) the coordinates of the artworks and rooms in the museum. This is a joint work between the Dutch nationally funded CHIP and Poseidon projects and the prototype demonstrator can be found at http://www.chip-project.org/spacechip .

Spatio-temporal data management has progressed significantly towards efficient storage and indexing of mobility data. Typically such mobility data analytics is assumed to follow the model of a stream of (x,y,t) points, usually coming from GPS-enabled mobile devices. With large-scale adoption of GPS-driven systems in several application sectors (shipment tracking to geo-social networks), there is a growing demand from applications to understand the spatio-semantic behavior of mobile entities. Spatio-semantic behavior essentially means a semantic (and preferably contextual) abstraction of raw spatio-temporal location feeds. The core contribution of this paper lies in presenting a Hybrid Model and a Computing Platform for developing a semantic overlay - analyzing and transforming raw mobility data (GPS) to meaningful semantic abstractions, starting from raw feeds to semantic trajectories. Secondly, we analyze large-scale GPS data using our computing platform and present results of extracted spatio-semantic trajectories. This impacts a large class of mobile applications requiring such semantic abstractions over streaming location feeds in real systems today.

Query answering on a wide and heterogeneous environment such as the Web can return a large number of results that can be hardly manageable by users/agents. The adoption of grouping criteria of the results could be of great help. Up to date, most of the proposed methods for aggregating results on the (Semantic) Web are mainly grounded on syntactic approaches. However, they could not be of significant help, when the values instantiating a grouping criterion are all equal (thus creating a unique group) or are almost all different (thus creating one group for each answer). We propose a novel approach that is able to overcome such drawbacks: given a query in the form of a conjunctive query, grouping is grounded on the exploitation of the semantics of background ontologies during the aggregation of query results. Specifically, we propose a solution where answers are deductively grouped taking into account the subsumption hierarchy of the underlying knowledge base. In this way, the results can be shown and navigated similarly to a faceted search. An experimental evaluation of the proposed method is also reported.

With large datasets such as Linked Open Data available, there is a need for more user-friendly interfaces which will bring the advantages of these data closer to the casual users. Several recent studies have shown user preference to Natural Language Interfaces (NLIs) in comparison to others. Although many NLIs to ontologies have been developed, those that have reasonable performance are domain-specific and tend to require customisation for each new domain which, from a developer’s perspective, makes them expensive to maintain. We present our system FREyA, which combines syntactic parsing with the knowledge encoded in ontologies in order to reduce the customisation effort. If the system fails to automatically derive an answer, it will generate clarification dialogs for the user. The user’s selections are saved and used for training the system in order to improve its performance over time. FREyA is evaluated using Mooney Geoquery dataset with very high precision and recall.

Geo-spatial ontologies provide knowledge about places in the world and spatial relations between them. They are fundamental in order to build semantic information retrieval systems and to achieve semantic interoperability in geo-spatial applications. In this paper we present GeoWordNet, a semantic resource we created from the full integration of GeoNames, other high quality resources and WordNet. The methodology we followed was largely automatic, with manual checks when needed. This allowed us accomplishing at the same time a never reached before accuracy level and a very satisfactory quantitative result, both in terms of concepts and geographical entities.

The use of natural language identifiers as reference for ontology elements—in addition to the URIs required by the Semantic Web standards—is of utmost importance because of their predominance in the human everyday life, i.e.speech or print media. Depending on the context, different names can be chosen for one and the same element, and the same element can be referenced by different names. Here homonymy and synonymy are the main cause of ambiguity in perceiving which concrete unique ontology element ought to be referenced by a specific natural language identifier describing an entity. We propose a novel method to resolve entity references under the aspect of ambiguity which explores only formal background knowledge represented in RDF graph structures. The key idea of our domain independent approach is to build an entity network with the most likely referenced ontology elements by constructing steiner graphs based on spreading activation. In addition to exploiting complex graph structures, we devise a new ranking technique that characterises the likelihood of entities in this network, i.e. interpretation contexts. Experiments in a highly polysemic domain show the ability of the algorithm to retrieve the correct ontology elements in almost all cases.

Recent research has shown that annotations are useful for representing access restrictions to the axioms of an ontology and their implicit consequences. Previous work focused on assigning a label, representing its access level, to each consequence from a given ontology. However, a security administrator might not be satisfied with the access level obtained through these methods. In this case, one is interested in finding which axioms would need to get their access restrictions modified in order to get the desired label for the consequence. In this paper we look at this problem and present algorithms for solving it with a variety of optimizations. We also present first experimental results on large scale ontologies, which show that our methods perform well in practice.

Description Logic Programs (DL-programs) have been introduced to combine ontological and rule-based reasoning in the context of the Semantic Web. A DL-program loosely combines a Description Logic (DL) ontology with a non-monotonic logic program (LP) such that dedicated atoms in the LP, called DL-atoms, allow for a bidirectional flow of knowledge between the two components. Unfortunately, the information sent from the LP-part to the DL-part might cause an inconsistency in the latter, leading to the trivial satisfaction of every query. As a consequence, in such a case, the answer sets that define the semantics of the DL-program may contain spoiled information influencing the overall deduction. For avoiding unintuitive answer sets, we introduce a refined semantics for DL-programs that is sensitive for inconsistency caused by the combination of DL and LP, and dynamically deactivates rules whenever such an inconsistency would arise. We analyze the complexity of the new semantics, discuss implementational issues and introduce a notion of stratification that guarantees uniqueness of answer sets.

In this paper we build on our methodology for combining and selecting alignment techniques for vocabularies, with two alignment case studies of large vocabularies in two languages. Firstly, we analyze the vocabularies and based on that analysis choose our alignment techniques. Secondly, we test our hypothesis based on earlier work that first generating alignments using simple lexical alignment techniques, followed by a separate disambiguation of alignments performs best in terms of precision and recall. The experimental results show, for example, that this combination of techniques provides an estimated precision of 0.7 for a sample of the 12,725 concepts for which alignments were generated (of the total 27,992 concepts). Thirdly, we explain our results in light of the characteristics of the vocabularies and discuss their impact on the alignments techniques.

In previous work we have shown that the MapReduce framework for distributed computation can be deployed for highly scalable inference over RDF graphs under the RDF Schema semantics. Unfortunately, several key optimizations that enabled the scalable RDFS inference do not generalize to the richer OWL semantics. In this paper we analyze these problems, and we propose solutions to overcome them. Our solutions allow distributed computation of the closure of an RDF graph under the OWL Horst semantics.

We demonstrate the WebPIE inference engine, built on top of the Hadoop platform and deployed on a compute cluster of 64 machines. We have evaluated our approach using some real-world datasets (UniProt and LDSR, about 0.9-1.5 billion triples) and a synthetic benchmark (LUBM, up to 100 billion triples). Results show that our implementation is scalable and vastly outperforms current systems when comparing supported language expressivity, maximum data size and inference speed.

In SPARQL, conjunctive queries are expressed by using shared variables across sets of triple patterns, also called basic graph patterns. Based on this characterization, basic graph patterns in a SPARQL query can be partitioned into groups of acyclic patterns that share exactly one variable, or star-shaped groups. We observe that the number of triples in a group is proportional to the number of individuals that play the role of the subject or the object; however, depending on the degree of participation of the subject individuals in the properties, a group could be not much larger than a class or type to which the subject or object belongs. Thus, it may be significantly more efficient to independently evaluate each of the groups, and then merge the resulting sets, than linearly joining all triples in a basic graph pattern. Based on this observation, we have developed query optimization and evaluation techniques on star-shaped groups. We have conducted an empirical analysis on the benefits of the optimization and evaluation techniques in several SPARQL query engines. We observe that our proposed techniques are able to speed up query evaluation time for join queries with star-shaped patterns by at least one order of magnitude.

Medical imaging plays an important role in today’s clinical daily tasks, such as patient screening, diagnosis, treatment planning and follow up. But still a generic and flexible image understanding is missing. Although, there exist several approaches for semantic image annotation, those approaches do not make use of practical clinical knowledge, such as best practice solutions or clinical guidelines. We introduce a knowledge engineering approach aiming for reasoning-based enhancement of medical images annotation by integrating practical clinical knowledge. We will exemplify the reasoning steps of the methodology along a use case for automatic lymphoma patient staging.

In this paper we describe the application of various Semantic Web technologies and their combination with emerging Web 2.0 use patterns in the eParticipation domain and show how they are used in an operational system for the Regional Government of the Prefecture of Samos, Greece. We present parts of the system that are based on Semantic Web technology and how they are merged with a Web 2.0 philosophy and explain the benefits of this approach, as showcased by applications for annotating, searching, browsing and cross-referencing content in eParticipation communities.

Within the archaeology domain, datasets frequently refer to time periods using a variety of textual or numeric formats. Traditionally controlled vocabularies of time periods have used classification notation and the collocation of terms in the printed form to represent and convey tacit information about the relative order of concepts. The emergence of the semantic web entails encoding this knowledge into machine readable forms, and so the meaning of this informal ordering arrangement can be lost. Conversion of controlled vocabularies to Simple Knowledge Organisation System (SKOS) format provides a formal basis for semantic web indexing but does not facilitate chronological inference - as thesaurus relationship types are an inappropriate mechanism to fully describe temporal relationships. This becomes an issue in archaeological data where periods are often described in terms of (e.g.) named monarchs or emperors, without additional information concerning relative chronological context.

An exercise in supplementing existing controlled vocabularies of time period concepts with dates and temporal relationships was undertaken as part of the Semantic Technologies for Archaeological Resources (STAR) project. The general aim of the STAR project is to demonstrate the potential benefits in cross searching archaeological data conforming to a common overarching conceptual data structure schema - the CIDOC Conceptual Reference Model (CRM). This paper gives an overview of STAR applications and services and goes on to particularly focus on issues concerning the extraction and representation of time period information.

While the Semantic Web is rapidly filling up, appropriate tools for searching it are still at infancy. In this paper we describe an approach that allows humans to access information contained in the Semantic Web according to its semantics and thus to leverage the specific characteristic of this Web. To avoid the ambiguity of natural language queries, users only select already defined attributes organized in facets to build their search queries. The facets are represented as nodes in a graph visualization and can be interactively added and removed by the users in order to produce individual search interfaces. This provides the possibility to generate interfaces in arbitrary complexities and access arbitrary domains. Even multiple and distantly connected facets can be integrated in the graph facilitating the access of information from different user-defined perspectives. Challenges include massive amounts of data, massive semantic relations within the data, highly complex search queries and users’ unfamiliarity with the Semantic Web.

This paper presents an approach for the interactive discovery of relationships between selected elements via the Semantic Web. It emphasizes the human aspect of relationship discovery by offering sophisticated interaction support. Selected elements are first semi-automatically mapped to unique objects of Semantic Web datasets. These datasets are then crawled for relationships which are presented in detail and overview. Interactive features and visual clues allow for a sophisticated exploration of the found relationships. The general process is described and the RelFinder tool as a concrete implementation and proof-of-concept is presented and evaluated in a user study. The application potentials are illustrated by a scenario that uses the RelFinder and DBpedia to assist a business analyst in decision-making. Main contributions compared to previous and related work are data aggregations on several dimensions, a graph visualization that displays and connects relationships also between more than two given objects, and an advanced implementation that is highly configurable and applicable to arbitrary RDF datasets.

A single datum or a set of a categorical data has little value on its own. Combinations of disparate sets of data increase the value of those data sets and helps to discover interesting patterns or relationships, facilitating the construction of new applications and services. In this paper, we describe an implementation of using open geographical data as a core set of “join point”(s) to mesh different public datasets. We describe the challenges faced during the implementation, which include, sourcing the datasets, publishing them as linked data, and normalising these linked data in terms of finding the appropriate “join points” from the individual datasets, as well as developing the client application used for data consumption. We describe the design decisions and our solutions to these challenges. We conclude by drawing some general principles from this work.

The HermesWiki project is a semantic wiki application on Ancient Greek History. As an e-learning platform, it aims at providing students effective access to concise and reliable domain knowledge, that is especially important for exam preparation. In this paper, we show how semantic technologies introduce new methods of learning by supporting teachers in the creation of contents and students in the personalized identification of required knowledge. Therefore, we give an overview of the project and characterize the semi-formalized content. Additionally, we present several use cases and describe the semantic web techniques that are used to support the application. Furthermore, we report on the user experiences regarding the usefulness and applicability of semantic technologies in this context.

In this paper we present a concrete case study in which semantic technology has been used to enable a territorial innovation. Firstly, we describe a scenario of the ICT regional demand in Trentino, Italy; where the main idea of territorial innovation is based on the so-called innovation tripole. Specifically, we believe that innovation arises as a result of the synergic coordination and technology transfer among three main innovation stakeholders: (i) final users, bringing domain knowledge, (ii) enterprises and SMEs, bringing knowledge of the market, and (iii) research centers, bringing the latest research results. The tripole is instantiated/generated for innovation projects, and, technically, can be viewed as a competence search (based on metadata) among the key innovation stakeholders for those projects. Secondly, we discuss the implementation of the tripole generation within the TasLab portal, including the use of domain ontologies and thesauri (e.g., Eurovoc), indexing and semantic search techniques we have employed. Finally, we provide a discussion on empirical and strategic evaluation of our solution, the results of which are encouraging.

Semantic enrichment of textual data is the operation of linking mentions with the entities they refer to, and the subsequent enrichment of such entities with the background knowledge about them available in one or more knowledge bases (or in the entire web). Information about the context in which a mention occurs, (e.g., information about the time, the topic, and the space, which the text is relative to) constitutes a critical resource for a correct semantic enrichment for two reasons. First, without context, mentions are “too little text” to unambiguously refer to a single entity. Second, knowledge about entities is also context dependent (e.g., speaking about political life of Illinois during 1996, Obama is a Senator, while since 2009, Obama is the US president). In this paper, we describe a concrete approach to context-driven semantic enrichment, built upon four core sub-tasks: detection of mentions in text (i.e., finding references to people, locations and organizations); determination of the context of discourses of the text, identification of the referred entities in the knowledge base, and enrichment of the entity with the knowledge relevant to the context. In such approach, context-driven semantic enrichment needs also to have contextualized background knowledge. To cope with this aspect, we propose a customization of Sesame, one of state-of-the-art knowledge repositories, to support representation and reasoning with contextualized knowledge. The approach has been fully implemented in a system, which has been practically deployed and applied to the textual archive of the local Italian newspaper “L’Adige”, covering the decade of years from 1999 to 2009.

The aim of this paper is to benchmark various semantic repositories in order to evaluate their deployment in a commercial image retrieval and browsing application. We adopt a two-phase approach for evaluating the target semantic repositories: analytical parameters such as query language and reasoning support are used to select the pool of the target repositories, and practical parameters such as load and query response times are used to select the best match to application requirements. In addition to utilising a widely accepted benchmark for OWL repositories (UOBM), we also use a real-life dataset from the target application, which provides us with the opportunity of consolidating our findings. A distinctive advantage of this benchmarking study is that the essential requirements for the target system such as the semantic expressivity and data scalability are clearly defined, which allows us to claim contribution to the benchmarking methodology for this class of applications.

Controlled vocabularies of various kinds (e.g., thesauri, classification schemes) play an integral part in making Cultural Heritage collections accessible. The various institutions participating in the Dutch CATCH programme maintain and make use of a rich and diverse set of vocabularies. This makes it hard to provide a uniform point of access to all collections at once. Our SKOS-based vocabulary and alignment repository aims at providing technology for managing the various vocabularies, and for exploiting semantic alignments across any two of them. The repository system exposes web services that effectively support the construction of tools for searching and browsing across vocabularies and collections or for collection curation (indexing), as we demonstrate.

This paper presents an ontology management system and ontology processing techniques used to support a distributed event-triggered knowledge network (ETKnet), which has been developed for deployment in a national network for rapid detection and reporting of crop disease and pest outbreaks. The ontology management system, called Lyra, is improved to address issues of terminology mapping, rule discovery, and large ABox inference. A domain ontology that covers the concepts related to events, rules, roles and collaborating organizations for this application in ETKnet was developed. Terms used by different organizations can be located in the ontology by terminology searching. Services that implement knowledge rules and rule structures can be discovered through semantic matching using the concepts defined in the ontology. A tableau algorithm was extended to lazy-load only the needed instances and their relationships into main memory. With this extension, Lyra is capable of processing a large ontology database stored in secondary storage even when the ABox cannot be entirely loaded into memory.

RDF will often be the metadata model of choice in the Semantic Sensor Web. However, RDF can only represent thematic metadata and needs to be extended if we want to model spatial and temporal information. For this purpose, we develop the data model stRDF and the query language stSPARQL. stRDF is a constraint data model that extends RDF with the ability to represent spatial and temporal data. stSPARQL extends SPARQL for querying stRDF data. In our extension to RDF, we follow the main ideas of constraint databases and represent spatial and temporal objects as quantifier-free formulas in a first-order logic of linear constraints. Thus an important contribution of stRDF is to bring to the RDF world the benefits of constraint databases and constraint-based reasoning so that spatial and temporal data can be represented in RDF using constraints.