The Semantic Web – ISWC 2012

Classification is a fundamental reasoning task in ontology design, and there is currently a wide range of reasoners highly optimised for classification of OWL 2 ontologies. There are also several reasoners that are complete for restricted fragments of OWL 2 , such as the OWL 2 EL profile. These reasoners are much more efficient than fully-fledged OWL 2 reasoners, but they are not complete for ontologies containing (even if just a few) axioms outside the relevant fragment. In this paper, we propose a novel classification technique that combines an OWL 2 reasoner and an efficient reasoner for a given fragment in such a way that the bulk of the workload is assigned to the latter. Reasoners are combined in a black-box modular manner, and the specifics of their implementation (and even of their reasoning technique) are irrelevant to our approach.

Ontology mappings are often assigned a weight or confidence factor by matchers. Nonetheless, few semantic accounts have been given so far for such weights. This paper presents a formal semantics for weighted mappings between different ontologies. It is based on a classificational interpretation of mappings: if

O

1

and

O

2

are two ontologies used to classify a common set

X

, then mappings between

O

1

and

O

2

are interpreted to encode how elements of

X

classified in the concepts of

O

1

are re-classified in the concepts of

O

2

, and weights are interpreted to measure how precise and complete re-classifications are. This semantics is justifiable by extensional practice of ontology matching. It is a conservative extension of a semantics of crisp mappings. The paper also includes properties that relate mapping entailment with description logic constructors.

Modelling and understanding various contexts of users is important to enable personalised selection of Web APIs in directories such as Programmable Web. Currently, relationships between users and Web APIs are not clearly understood and utilized by existing selection approaches. In this paper, we present a semantic model of a Web API directory graph that captures relationships such as Web APIs, mashups, developers, and categories. We describe a novel configurable graph-based method for selection of Web APIs with personalised and temporal aspects. The method allows users to get more control over their preferences and recommended Web APIs while they can exploit information about their social links and preferences. We evaluate the method on a real-world dataset from ProgrammableWeb.com, and show that it provides more contextualised results than currently available popularity-based rankings.

In this paper, we describe a mechanism for ontology alignment using instance based matching of types (or classes). Instance-based matching is known to be a useful technique for matching ontologies that have different names and different structures. A key problem in instance matching of types, however, is scaling the matching algorithm to (a) handle types with a large number of instances, and (b) efficiently match a large number of type pairs. We propose the use of state-of-the art locality-sensitive hashing (LSH) techniques to vastly improve the scalability of instance matching across multiple types. We show the feasibility of our approach with DBpedia and Freebase, two different type systems with hundreds and thousands of types, respectively. We describe how these techniques can be used to estimate containment or equivalence relations between two type systems, and we compare two different LSH techniques for computing instance similarity.

We present

Tìpalo

, an algorithm and tool for automatically typing DBpedia entities. Tìpalo identifies the most appropriate types for an entity by interpreting its natural language definition, which is extracted from its corresponding Wikipedia page abstract. Types are identified by means of a set of heuristics based on graph patterns, disambiguated to WordNet, and aligned to two top-level ontologies: WordNet supersenses and a subset of DOLCE+DnS Ultra Lite classes. The algorithm has been tuned against a golden standard that has been built online by a group of selected users, and further evaluated in a user study.

Due to the high worst case complexity of the core reasoning problem for the expressive profiles of OWL 2, ontology engineers are often surprised and confused by the performance behaviour of reasoners on their ontologies. Even very experienced modellers with a sophisticated grasp of reasoning algorithms do not have a good mental model of reasoner performance behaviour. Seemingly innocuous changes to an OWL ontology can degrade classification time from instantaneous to too long to wait for. Similarly, switching reasoners (e.g., to take advantage of specific features) can result in wildly different classification times. In this paper we investigate performance variability phenomena in OWL ontologies, and present methods to identify subsets of an ontology which are performance-degrading for a given reasoner. When such (ideally small) subsets are removed from an ontology, and the remainder is much easier for the given reasoner to reason over, we designate them “hot spots”. The identification of these hot spots allows users to isolate difficult portions of the ontology in a principled and systematic way. Moreover, we devise and compare various methods for approximate reasoning and knowledge compilation based on hot spots. We verify our techniques with a select set of varyingly difficult ontologies from the NCBO BioPortal, and were able to, firstly, successfully identify performance hot spots against the major freely available DL reasoners, and, secondly, significantly improve classification time using approximate reasoning based on hot spots.

Detecting, much less understanding, the difference between two description logic based ontologies is challenging for ontology engineers due, in part, to the possibility of complex, non-local logic effects of axiom changes. First, it is often quite difficult to even determine which concepts have had their meaning altered by a change. Second, once a concept change is pinpointed, the problem of distinguishing whether the concept is directly or indirectly affected by a change has yet to be tackled. To address the first issue, various principled notions of “semantic diff” (based on deductive inseparability) have been proposed in the literature and shown to be computationally practical for the expressively restricted case of

${\mathcal ELH}^r$

-terminologies. However, problems arise even for such limited logics as

${\mathcal ALC}$

: First, computation gets more difficult, becoming undecidable for logics such as

${\mathcal SROIQ}$

which underly the Web Ontology Language (OWL). Second, the presence of negation and disjunction make the standard semantic difference too sensitive to change: essentially, any logically effectual change always affects all terms in the ontology. In order to tackle these issues, we formulate the central notion of finding the

minimal change set

based on model inseparability, and present a method to differentiate changes which are specific to (thus directly affect) particular concept names. Subsequently we devise a series of computable approximations, and compare the variously approximated change sets over a series of versions of the NCI Thesaurus (NCIt).

The distributed and heterogeneous nature of Linked Open Data requires flexible and federated techniques for query evaluation. In order to evaluate current federation querying approaches a general methodology for conducting benchmarks is mandatory. In this paper, we present a classification methodology for federated

SPARQL

queries. This methodology can be used by developers of federated querying approaches to compose a set of test benchmarks that cover diverse characteristics of different queries and allows for comparability. We further develop a heuristic called

SPLODGE

for automatic generation of benchmark queries that is based on this methodology and takes into account the number of sources to be queried and several complexity parameters. We evaluate the adequacy of our methodology and the query generation strategy by applying them on the 2011 billion triple challenge data set.

Recent developments in hardware have shown an increase in parallelism as opposed to clock rates. In order to fully exploit these new avenues of performance improvement, computationally expensive workloads have to be expressed in a way that allows for fine-grained parallelism. In this paper, we address the problem of describing RDFS entailment in such a way. Different from previous work on parallel RDFS reasoning, we assume a shared memory architecture. We analyze the problem of duplicates that naturally occur in RDFS reasoning and develop strategies towards its mitigation, exploiting all levels of our architecture. We implement and evaluate our approach on two real-world datasets and study its performance characteristics on different levels of parallelization. We conclude that RDFS entailment lends itself well to parallelization but can benefit even more from careful optimizations that take into account intricacies of modern parallel hardware.

The primary challenge of machine perception is to define efficient computational methods to derive high-level knowledge from low-level sensor observation data. Emerging solutions are using ontologies for expressive representation of concepts in the domain of sensing and perception, which enable advanced integration and interpretation of heterogeneous sensor data. The computational complexity of OWL, however, seriously limits its applicability and use within resource-constrained environments, such as mobile devices. To overcome this issue, we employ OWL to formally define the inference tasks needed for machine perception – explanation and discrimination – and then provide efficient algorithms for these tasks, using bit-vector encodings and operations. The applicability of our approach to machine perception is evaluated on a smart-phone mobile device, demonstrating dramatic improvements in both efficiency and scale.

Most of the semantic content available has been generated automatically by using annotation services for existing content. Automatic annotation is not of sufficient quality to enable focused search and retrieval: either too many or too few terms are semantically annotated. User-defined semantic enrichment allows for a more targeted approach. We developed a tool for semantic annotation of digital documents and conducted an end-user study to evaluate its acceptance by and usability for non-expert users. This paper presents the results of this user study and discusses the lessons learned about both the semantic enrichment process and our methodology of exposing non-experts to semantic enrichment.

We propose a framework for querying probabilistic instance data in the presence of an

OWL2 QL

ontology, arguing that the interplay of probabilities and ontologies is fruitful in many applications such as managing data that was extracted from the web. The prime inference problem is computing answer probabilities, and it can be implemented using standard probabilistic database systems. We establish a

PTime

vs. #P dichotomy for the data complexity of this problem by lifting a corresponding result from probabilistic databases. We also demonstrate that query rewriting (backwards chaining) is an important tool for our framework, show that non-existence of a rewriting into first-order logic implies #P-hardness, and briefly discuss approximation of answer probabilities.

A key issue in semantic reasoning is the computational complexity of inference tasks on expressive ontology languages such as OWL DL and OWL 2 DL. Theoretical works have established worst-case complexity results for reasoning tasks for these languages. However, hardness of reasoning about individual ontologies has not been adequately characterised. In this paper, we conduct a systematic study to tackle this problem using machine learning techniques, covering over 350 real-world ontologies and four state-of-the-art, widely-used OWL 2 reasoners. Our main contributions are two-fold. Firstly, we learn various classifiers that accurately predict classification time for an ontology based on its metric values. Secondly, we identify a number of metrics that can be used to effectively predict reasoning performance. Our prediction models have been shown to be highly effective, achieving an accuracy of over 80%.

Data provenance is the history of derivation of a data artifact from its original sources. As the real-life provenance records can likely cover thousands of data items and derivation steps, one of the pressing challenges becomes development of formal frameworks for their automated verification.

In this paper, we consider data expressed in standard Semantic Web ontology languages, such as OWL, and define a novel verification formalism called

provenance specification logic

, building on dynamic logic. We validate our proposal by modeling the test queries presented in The First Provenance Challenge, and conclude that the logic core of such queries can be successfully captured in our formalism.

The paper presents an approach for cost-based query planning for SPARQL queries issued over an OWL ontology using the OWL Direct Semantics entailment regime of SPARQL 1.1. The costs are based on information about the instances of classes and properties that are extracted from a model abstraction built by an OWL reasoner. A static and a dynamic algorithm are presented which use these costs to find optimal or near optimal execution orders for the atoms of a query. For the dynamic case, we improve the performance by exploiting an individual clustering approach that allows for computing the cost functions based on one individual sample from a cluster. Our experimental study shows that the static ordering usually outperforms the dynamic one when accurate statistics are available. This changes, however, when the statistics are less accurate, e.g., due to non-deterministic reasoning decisions.

We describe a system that incrementally translates SPARQL queries to Pig Latin and executes them on a Hadoop cluster. This system is designed to work efficiently on complex queries with many self-joins over huge datasets, avoiding job failures even in the case of joins with unexpected high-value skew. To be robust against cost estimation errors, our system

interleaves

query optimization with query execution, determining the next steps to take based on data samples and statistics gathered during the previous step. Furthermore, we have developed a novel skew-resistant join algorithm that replicates tuples corresponding to popular keys. We evaluate the effectiveness of our approach both on a synthetic benchmark known to generate complex queries (BSBM-BI) as well as on a Yahoo! case of data analysis using RDF data crawled from the web. Our results indicate that our system is indeed capable of processing huge datasets without pre-computed statistics while exhibiting good load-balancing properties.

We present a large-scale relation extraction (RE) system which learns grammar-based RE rules from the Web by utilizing large numbers of relation instances as seed. Our goal is to obtain rule sets large enough to cover the actual range of linguistic variation, thus tackling the long-tail problem of real-world applications. A variant of distant supervision learns several relations in parallel, enabling a new method of rule filtering. The system detects both binary and

n

-ary relations. We target 39 relations from Freebase, for which 3M sentences extracted from 20M web pages serve as the basis for learning an average of 40K distinctive rules per relation. Employing an efficient dependency parser, the average run time for each relation is only 19 hours. We compare these rules with ones learned from local corpora of different sizes and demonstrate that the Web is indeed needed for a good coverage of linguistic variation.

The lightweight ontology language OWL RL is used for reasoning with large amounts of data. To this end, the W3C standard provides a simple system of deduction rules, which operate directly on the RDF syntax of OWL. Several similar systems have been studied. However, these approaches are usually complete for instance retrieval only. This paper asks if and how such methods could also be used for computing entailed subclass relationships. Checking entailment for arbitrary OWL RL class subsumptions is co-NP-hard, but tractable rule-based reasoning is possible when restricting to subsumptions between atomic classes. Surprisingly, however, this cannot be achieved in any RDF-based rule system, i.e., the W3C calculus cannot be extended to compute all atomic class subsumptions. We identify syntactic restrictions to mitigate this problem, and propose a rule system that is sound and complete for many OWL RL ontologies.

We present Strabon, a new RDF store that supports the state of the art semantic geospatial query languages stSPARQL and GeoSPARQL. To illustrate the expressive power offered by these query languages and their implementation in Strabon, we concentrate on the new version of the data model stRDF and the query language stSPARQL that we have developed ourselves. Like GeoSPARQL, these new versions use OGC standards to represent geometries where the original versions used linear constraints. We study the performance of Strabon experimentally and show that it scales to very large data volumes and performs, most of the times, better than all other geospatial RDF stores it has been compared with.

One of the main tasks when creating and maintaining knowledge bases is to validate facts and provide sources for them in order to ensure correctness and traceability of the provided knowledge. So far, this task is often addressed by human curators in a three-step process: issuing appropriate keyword queries for the statement to check using standard search engines, retrieving potentially relevant documents and screening those documents for relevant content. The drawbacks of this process are manifold. Most importantly, it is very time-consuming as the experts have to carry out several search processes and must often read several documents. In this article, we present DeFacto (Deep Fact Validation) – an algorithm for validating facts by finding trustworthy sources for it on the Web. DeFacto aims to provide an effective way of validating facts by supplying the user with relevant excerpts of webpages as well as useful additional information including a score for the confidence DeFacto has in the correctness of the input fact.

Web APIs have gained increasing popularity in recent Web service technology development owing to its simplicity of technology stack and the proliferation of mashups. However, efficiently discovering Web APIs and the relevant documentations on the Web is still a challenging task even with the best resources available on the Web. In this paper we cast the problem of detecting the Web API documentations as a text classification problem of classifying a given Web page as Web API associated or not. We propose a supervised generative topic model called feature latent Dirichlet allocation (feaLDA) which offers a generic probabilistic framework for automatic detection of Web APIs. feaLDA not only captures the correspondence between data and the associated class labels, but also provides a mechanism for incorporating side information such as labelled features automatically learned from data that can effectively help improving classification performance. Extensive experiments on our Web APIs documentation dataset shows that the feaLDA model outperforms three strong supervised baselines including naive Bayes, support vector machines, and the maximum entropy model, by over 3% in classification accuracy. In addition, feaLDA also gives superior performance when compared against other existing supervised topic models.

Top-k queries, i.e. queries returning the top

k

results ordered by a user-defined scoring function, are an important category of queries. Order is an important property of data that can be exploited to speed up query processing. State-of-the-art SPARQL engines underuse order, and top-k queries are mostly managed with a

materialize-then-sort

processing scheme that computes all the matching solutions (e.g. thousands) even if only a limited number

k

(e.g. ten) are requested. The

$\mathcal{S}$

PARQL-

$\mathcal{R}$

ANK algebra is an extended SPARQL algebra that treats order as a first class citizen, enabling efficient

split-and-interleave

processing schemes that can be adopted to improve the performance of top-k SPARQL queries. In this paper we propose an incremental execution model for

$\mathcal{S}$

PARQL-

$\mathcal{R}$

ANK queries, we compare the performance of alternative physical operators, and we propose a rank-aware join algorithm optimized for native RDF stores. Experiments conducted with an open source implementation of a

$\mathcal{S}$

PARQL-

$\mathcal{R}$

ANK query engine based on ARQ show that the evaluation of top-k queries can be sped up by orders of magnitude.

Tracking user interests over time is important for making accurate recommendations. However, the widely-used time-decay-based approach worsens the sparsity problem because it deemphasizes old item transactions. We introduce two ideas to solve the sparsity problem. First, we divide the users’ transactions into epochs i.e. time periods, and identify epochs that are dominated by interests similar to the current interests of the active user. Thus, it can eliminate dissimilar transactions while making use of similar transactions that exist in prior epochs. Second, we use a taxonomy of items to model user item transactions in each epoch. This well captures the interests of users in each epoch even if there are few transactions. It suits the situations in which the items transacted by users dynamically change over time; the semantics behind classes do not change so often while individual items often appear and disappear. Fortunately, many taxonomies are now available on the web because of the spread of the Linked Open Data vision. We can now use those to understand dynamic user interests semantically. We evaluate our method using a dataset, a music listening history, extracted from users’ tweets and one containing a restaurant visit history gathered from a gourmet guide site. The results show that our method predicts user interests much more accurately than the previous time-decay-based method.

Time-efficient algorithms are essential to address the complex linking tasks that arise when trying to discover links on the Web of Data. Although several lossless approaches have been developed for this exact purpose, they do not offer theoretical guarantees with respect to their performance. In this paper, we address this drawback by presenting the first Link Discovery approach with theoretical quality guarantees. In particular, we prove that given an achievable reduction ratio

r

, our Link Discovery approach

$\mathcal{HR}^3$

can achieve a reduction ratio

r

′ ≤ 

r

in a metric space where distances are measured by the means of a Minkowski metric of any order

p

 ≥ 2. We compare

$\mathcal{HR}^3$

and the HYPPO algorithm implemented in LIMES 0.5 with respect to the number of comparisons they carry out. In addition, we compare our approach with the algorithms implemented in the state-of-the-art frameworks LIMES 0.5 and SILK 2.5 with respect to runtime. We show that

$\mathcal{HR}^3$

outperforms these previous approaches with respect to runtime in each of our four experimental setups.

Three conflicting requirements arise in the context of knowledge base (KB) extraction: the size of the extracted KB, the size of the corresponding signature and the syntactic similarity of the extracted KB with the original one. Minimal module extraction and uniform interpolation assign an absolute priority to one of these requirements, thereby limiting the possibilities to influence the other two. We propose a novel technique for

${\mathcal EL}$

that does not require such an extreme prioritization. We propose a tractable rewriting approach and empirically compare the technique with existing approaches with encouraging results.

For a number of years now we have seen the emergence of repositories of research data specified using OWL/RDF as representation languages, and conceptualized according to a variety of ontologies. This class of solutions promises both to facilitate the integration of research data with other relevant sources of information and also to support more intelligent forms of querying and exploration. However, an issue which has only been partially addressed is that of generating and characterizing semantically the relations that exist between research areas. This problem has been traditionally addressed by manually creating taxonomies, such as the ACM classification of research topics. However, this manual approach is inadequate for a number of reasons: these taxonomies are very coarse-grained and they do not cater for the fine-grained research topics, which define the level at which typically researchers (and even more so, PhD students) operate. Moreover, they evolve slowly, and therefore they tend not to cover the most recent research trends. In addition, as we move towards a semantic characterization of these relations, there is arguably a need for a more sophisticated characterization than a homogeneous taxonomy, to reflect the different ways in which research areas can be related. In this paper we propose Klink, a new approach to i) automatically generating relations between research areas and ii) populating a bibliographic ontology, which combines both machine learning methods and external knowledge, which is drawn from a number of resources, including Google Scholar and Wikipedia. We have tested a number of alternative algorithms and our evaluation shows that a method relying on both external knowledge and the ability to detect temporal relations between research areas performs best with respect to a manually constructed standard.

Despite the increase in the number of linked instances in the Linked Data Cloud in recent times, the absence of links at the concept level has resulted in heterogenous schemas, challenging the interoperability goal of the Semantic Web. In this paper, we address this problem by finding alignments between concepts from multiple Linked Data sources. Instead of only considering the existing concepts present in each ontology, we hypothesize new composite concepts defined as disjunctions of conjunctions of (RDF) types and value restrictions, which we call

restriction classes

, and generate alignments between these composite concepts. This extended concept language enables us to find more complete definitions and to even align sources that have rudimentary ontologies, such as those that are simple renderings of relational databases. Our concept alignment approach is based on analyzing the extensions of these concepts and their linked instances. Having explored the alignment of conjunctive concepts in our previous work, in this paper, we focus on concept coverings (disjunctions of

restriction classes

). We present an evaluation of this new algorithm to Geospatial, Biological Classification, and Genetics domains. The resulting alignments are useful for refining existing ontologies and determining the alignments between concepts in the ontologies, thus increasing the interoperability in the Linked Open Data Cloud.

Ontology and other logical languages are built around the idea that axioms enable the inference of new facts about the available data. In some circumstances, however, the data is meant to be complete in certain ways, and deducing new facts may be undesirable. Previous approaches to this issue have relied on syntactically specifying certain axioms as constraints or adding in new constructs for constraints, and providing a different or extended meaning for constraints that reduces or eliminates their ability to infer new facts without requiring the data to be complete. We propose to instead directly state that the extension of certain concepts and roles are complete by making them DBox predicates, which eliminates the distinction between regular axioms and constraints for these concepts and roles. This proposal eliminates the need for special semantics and avoids problems of previous proposals.

The Linking Open Data (LOD) project is an ongoing effort to construct a global data space, i.e. the Web of Data. One important part of this project is to establish

owl:sameAs

links among structured data sources. Such links indicate equivalent instances that refer to the same real-world object. The problem of discovering

owl:sameAs

links between pairwise data sources is called

instance matching

. Most of the existing approaches addressing this problem rely on the quality of prior schema matching, which is not always good enough in the LOD scenario. In this paper, we propose a schema-independent instance-pair similarity metric based on several general descriptive features. We transform the instance matching problem to the binary classification problem and solve it by machine learning algorithms. Furthermore, we employ some transfer learning methods to utilize the existing

owl:sameAs

links in LOD to reduce the demand for labeled data. We carry out experiments on some datasets of OAEI2010. The results show that our method performs well on real-world LOD data and outperforms the participants of OAEI2010.

Existing approaches for link prediction, in the domain of network science, exploit a network’s topology to predict future connections by assessing existing edges and connections, and inducing links given the presence of mutual nodes. Despite the rise in popularity of Attention-Information Networks (i.e. microblogging platforms) and the production of content within such platforms, no existing work has attempted to exploit the semantics of published content when predicting network links. In this paper we present an approach that fills this gap by a) predicting

follower

edges within a directed social network by exploiting concept graphs and thereby significantly outperforming a random baseline and models that rely solely on network topology information, and b) assessing the different behaviour that users exhibit when making followee-addition decisions. This latter contribution exposes latent factors within social networks and the existence of a clear need for topical affinity between users for a follow link to be created.

An increasing amount of data is published and consumed on the Web according to the Linked Data paradigm. In consideration of both publishers and consumers, the temporal dimension of data is important. In this paper we investigate the characterisation and availability of temporal information in Linked Data at large scale. Based on an abstract definition of temporal information we conduct experiments to evaluate the availability of such information using the data from the 2011 Billion Triple Challenge (BTC) dataset. Focusing in particular on the representation of temporal meta-information, i.e., temporal information associated with RDF statements and graphs, we investigate the approaches proposed in the literature, performing both a quantitative and a qualitative analysis and proposing guidelines for data consumers and publishers. Our experiments show that the amount of temporal information available in the LOD cloud is still very small; several different models have been used on different datasets, with a prevalence of approaches based on the annotation of RDF documents.

Sentiment analysis over Twitter offer organisations a fast and effective way to monitor the publics’ feelings towards their brand, business, directors, etc. A wide range of features and methods for training sentiment classifiers for Twitter datasets have been researched in recent years with varying results. In this paper, we introduce a novel approach of adding semantics as additional features into the training set for sentiment analysis. For each extracted entity (e.g. iPhone) from tweets, we add its semantic concept (e.g. “Apple product”) as an additional feature, and measure the correlation of the representative concept with negative/positive sentiment. We apply this approach to predict sentiment for three different Twitter datasets. Our results show an average increase of F harmonic accuracy score for identifying both negative and positive sentiment of around 6.5% and 4.8% over the baselines of unigrams and part-of-speech features respectively. We also compare against an approach based on sentiment-bearing topic analysis, and find that semantic features produce better Recall and F score when classifying negative sentiment, and better Precision with lower Recall and F score in positive sentiment classification.

The last decade of research in ontology alignment has brought a variety of computational techniques to discover correspondences between ontologies. While the accuracy of automatic approaches has continuously improved, human contributions remain a key ingredient of the process: this input serves as a valuable source of domain knowledge that is used to train the algorithms and to validate and augment automatically computed alignments. In this paper, we introduce

CrowdMap

, a model to acquire such human contributions via microtask crowdsourcing. For a given pair of ontologies,

CrowdMap

translates the alignment problem into microtasks that address individual alignment questions, publishes the microtasks on an online labor market, and evaluates the quality of the results obtained from the crowd. We evaluated the current implementation of

CrowdMap

in a series of experiments using ontologies and reference alignments from the Ontology Alignment Evaluation Initiative and the crowdsourcing platform CrowdFlower. The experiments clearly demonstrated that the overall approach is feasible, and can improve the accuracy of existing ontology alignment solutions in a fast, scalable, and cost-effective manner.

The inherent heterogeneity of datasets on the Semantic Web has created a need to interlink them, and several tools have emerged that automate this task. In this paper we are interested in what happens if we enrich these matching tools with knowledge of the domain of the ontologies. We explore how to express the notion of a domain in terms usable for an ontology matching tool, and we examine various methods to decide what constitutes the domain of a given dataset. We show how we can use this in a matching tool, and study the effect of domain knowledge on the quality of the alignment.

We perform evaluations for two scenarios: Last.fm artists and UMLS medical terms. To quantify the added value of domain knowledge, we compare our domain-aware matching approach to a standard approach based on a manually created reference alignment. The results indicate that the proposed domain-aware approach indeed outperforms the standard approach, with a large effect on ambiguous concepts but a much smaller effect on unambiguous concepts.

The amount of data available in the Linked Data cloud continues to grow. Yet, few services consume and produce linked data. There is recent work that allows a user to define a linked service from an online service, which includes the specifications for consuming and producing linked data, but building such models is time consuming and requires specialized knowledge of RDF and SPARQL. This paper presents a new approach that allows domain experts to rapidly create semantic models of services by demonstration in an interactive web-based interface. First, the user provides examples of the service request URLs. Then, the system automatically proposes a service model the user can refine interactively. Finally, the system saves a service specification using a new expressive vocabulary that includes lowering and lifting rules. This approach empowers end users to rapidly model existing services and immediately use them to consume and produce linked data.

This paper presents an approach to automatically extract entities and relationships from textual documents. The main goal is to populate a knowledge base that hosts this structured information about domain entities. The extracted entities and their expected relationships are verified using two evidence based techniques: classification and linking. This last process also enables the linking of our knowledge base to other sources which are part of the Linked Open Data cloud. We demonstrate the benefit of our approach through series of experiments with real-world datasets.

In RDF, a

blank node

(or anonymous resource or bnode) is a node in an RDF graph which is not identified by a URI and is not a literal. Several RDF/S Knowledge Bases (KBs) rely heavily on blank nodes as they are convenient for representing complex attributes or resources whose identity is unknown but their attributes (either literals or associations with other resources) are known. In this paper we show how we can exploit blank nodes anonymity in order to reduce the delta (diff) size when comparing such KBs. The main idea of the proposed method is to build a mapping between the bnodes of the compared KBs for reducing the delta size. We prove that finding the optimal mapping is NP-Hard in the general case, and polynomial in case there are not directly connected bnodes. Subsequently we present various polynomial algorithms returning approximate solutions for the general case.

For making the application of our method feasible also to large KBs we present a signature-based mapping algorithm with

n

log

n

complexity. Finally, we report experimental results over real and synthetic datasets that demonstrate significant reductions in the sizes of the computed deltas. For the proposed algorithms we also provide comparative results regarding delta reduction, equivalence detection and time efficiency.

For Linked Data query engines, there are inherent trade-offs between centralised approaches that can efficiently answer queries over data cached from parts of the Web, and live decentralised approaches that can provide fresher results over the entire Web at the cost of slower response times. Herein, we propose a

hybrid query execution

approach that returns fresher results from a broader range of sources vs. the centralised scenario, while speeding up results vs. the live scenario. We first compare results from two public SPARQL stores against current versions of the Linked Data sources they cache; results are often missing or out-of-date. We thus propose using

coherence estimates

to split a query into a sub-query for which the cached data have good fresh coverage, and a sub-query that should instead be run live. Finally, we evaluate different hybrid query plans and split positions in a real-world setup. Our results show that hybrid query execution can improve freshness vs. fully cached results while reducing the time taken vs. fully live execution.

Determining trust of data available in the Semantic Web is fundamental for applications and users, in particular for linked open data obtained from SPARQL endpoints. There exist several proposals in the literature to annotate SPARQL query results with values from abstract models, adapting the seminal works on provenance for annotated relational databases. We provide an approach capable of providing provenance information for a large and significant fragment of SPARQL 1.1, including for the first time the major non-monotonic constructs under multiset semantics. The approach is based on the translation of SPARQL into relational queries over annotated relations with values of the most general m-semiring, and in this way also refuting a claim in the literature that the

OPTIONAL

construct of SPARQL cannot be captured appropriately with the known abstract models.

We introduce

SRBench

, a general-purpose

bench

mark primarily designed for

s

treaming

R

DF/SPARQL engines, completely based on real-world data sets from the Linked Open Data cloud. With the increasing problem of too much streaming data but not enough tools to gain knowledge from them, researchers have set out for solutions in which Semantic Web technologies are adapted and extended for publishing, sharing, analysing and understanding streaming data. To help researchers and users comparing streaming RDF/SPARQL (strRS) engines in a standardised application scenario, we have designed SRBench, with which one can assess the abilities of a strRS engine to cope with a broad range of use cases typically encountered in real-world scenarios. The data sets used in the benchmark have been carefully chosen, such that they represent a realistic and relevant usage of streaming data. The benchmark defines a concise, yet comprehensive set of queries that cover the major aspects of strRS processing. Finally, our work is complemented with a functional evaluation on three representative strRS engines: SPARQL

Stream

, C-SPARQL and CQELS. The presented results are meant to give a first baseline and illustrate the state-of-the-art.

First order logic (FOL) rewritability is a desirable feature for query answering over geo-thematic ontologies because in most geo-processing scenarios one has to cope with large data volumes. Hence, there is a need for combined geo-thematic logics that provide a sufficiently expressive query language allowing for FOL rewritability. The DL-Lite family of description logics is tailored towards FOL rewritability of query answering for unions of conjunctive queries, hence it is a suitable candidate for the thematic component of a combined geo-thematic logic. We show that a weak coupling of DL-Lite with the expressive region connection calculus RCC8 allows for FOL rewritability under a spatial completeness condition for the ABox. Stronger couplings allowing for FOL rewritability are possible only for spatial calculi as weak as the low-resolution calculus RCC2. Already a strong combination of DL-Lite with the low-resolution calculus RCC3 does not allow for FOL rewritability.