Preface The ISWC 2014 Poster and Demonstration track complements the Research Paper track of the conference and o↵ers an opportunity for presenting late-breaking research results, ongoing research projects, and speculative or innovative work in progress. The informal setting of the track encourages presenters and participants to engage in discussions about the presented work. Such discussions can be a valuable input into the future work of the presenters, while o↵ering participants an e↵ective way to broaden their knowledge of the emerging research trends and to network with other researchers.
These proceedings contain the four-page abstracts of all accepted posters and demos presented at ISWC 2014. Posters range from technical contributions, reports on Semantic Web software systems, descriptions of completed work, and also work in progress. Demonstrations showcase innovative Semantic Web related implementations and technologies. This year we had 156 submissions, of which the program committee accept 71 posters and 49 demos. We would like to take this opportunity to thank all of the authors for their contributions to the ISWC 2014 programme!
We would also like to thank the members of the program committee and the additional reviewers for their time and e↵orts. A special thanks for respecting our deadlines, we know these fell in the middle of the summer holidays for many of you! All abstracts included here have been revised and improved based on your valuable feedback, and we feel the final result represents a wide variety of topics that will o↵er a vibrant and exciting session at the conference.
Finally, we would like to thank our local organisers Luciano Serafini and Chiara Ghidini for their invaluable help in sorting out the logistics of this track.
Stanford, Trento, Amsterdam
Alessandro Adamou Carlo Allocca Samantha Bail Pierpaolo Basile Eva Blomqvist Victor de Boer Stefano Bortoli Loris Bozzato Volha Bryl Marut Buranarach Jim Burton Elena Cabrio Annalina Caputo Vinay Chaudhri Gong Cheng Sam Coppens Oscar Corcho Francesco Corcoglioniti Claudia D’Amato Chiara Del Vescovo Aidan Delaney Daniele Dell’Aglio Chiara Di Francescomarino Mauro Dragoni Marieke van Erp Antske Fokkens Marco Gabriele Enrico Fossati Anna Lisa Gentile Aurona Gerber Jose Manuel Gomez-Perez Tudor Groza Gerd Gr¨oner Peter Haase Armin Haller Karl Hammar Michiel Hildebrand Pascal Hitzler Aidan Hogan Matthew Horridge Hanmin Jung Simon Jupp Haklae Kim Pavel Klinov Patrick Lambrix Paea Le Pendu Florian Lemmerich Yuan-Fang Li Joanne Luciano
Program Committee Muhammad Intizar Ali David Carral Marco Cremaschi Brian Davis Jangwon Gim Hegde Vinod Nazmul Hussain Myunggwon Hwang Amit Joshi Kim Taehong Kim Young-Min Fadi Maali Theofilos Mailis Nicolas Matentzoglu Jim Mccusker David Molik Raghava Mutharaju Alina Patelli Thomas Ploeger Riccardo Porrini Anon Reviewera Victor Saquicela Ana Sasa Bastinos Veronika Thost Jung-Ho Um Zhangquan Zhou 1 2 3 4 5 6 7 8 9 10 11 12 13 14
Life Stories as Event-based Linked Data: Case Semantic National Biography Eero Hyv¨onen, Miika Alonen, Esko Ikkala and Eetu M¨akel¨a News Visualization based on Semantic Knowledge Sebastian Arnold, Damian Burke, Tobias D¨orsch, Bernd L¨ober and Andreas Lommatzsch Sherlock: a Semi-Automatic Quiz Generation System using Linked Data Dong Liu and Chenghua Lin Low-Cost Queryable Linked Data through Triple Pattern Fragments Ruben Verborgh, Olaf Hartig, Ben De Meester, Gerald Haesendonck, Laurens De Vocht, Miel Vander Sande, Richard Cyganiak, Pieter Colpaert, Erik Mannens and Rik Van de Walle call: A Nucleus for a Web of Open Functions Maurizio Atzori Cross-lingual detection of world events from news articles Gregor Leban, Blaˇz Fortuna, Janez Brank and Marko Grobelnik Multilingual Word Sense Disambiguation and Entity Linking for Everybody Andrea Moro, Francesco Cecconi and Roberto Navigli Help me describe my data: A demonstration of the Open PHACTS VoID Editor Carole Goble, Alasdair J. G. Gray and Eleftherios Tatakis OUSocial2 - A Platform for Gathering Students’ Feedback from Social Media Keerthi Thomas, Miriam Fernandez, Stuart Brown and Harith Alani Using an Ontology Learning System for Trend Analysis and Detection Gerhard Wohlgenannt, Stefan Belk, Matyas Karacsonyi and Matthias Schett A Prototype Web Service for Benchmarking Power Consumption of Mobile Semantic Applications Evan Patton and Deborah McGuinness SPARKLIS: a SPARQL Endpoint Explorer for Expressive Question Answering S´ebastien Ferr´e Reconciling Information in DBpedia through a Question Answering System Elena Cabrio, Alessio Palmero Aprosio and Serena Villata Open Mashup Platform - A Smart Data Exploration Environment Tuan-Dat Trinh, Ba-Lam Do, Peter Wetz, Amin Anjomshoaa, Elmar Kiesling and Amin Tjoa 1 5 9 13 17 21 25 29 33 37 41 45 49 53 CIMBA - Client-Integrated MicroBlogging Architecture Andrei Sambra, Sandro Hawke, Timothy Berners-Lee, Lalana Kagal and Ashraf Aboulnaga The Organiser - A Semantic Desktop Agent based on NEPOMUK Sebastian Faubel and Moritz Eberl HDTourist: Exploring Urban Data on Android Elena Hervalejo, Miguel A. Martinez-Prieto, Javier D. Fern´andez and Oscar Corcho Integrating NLP and SW with the KnowledgeStore Marco Rospocher, Francesco Corcoglioniti, Roldano Cattoni, Bernardo Magnini and Luciano Serafini Graphical Representation of OWL 2 Ontologies through Graphol Marco Console, Domenico Lembo, Valerio Santarelli and Domenico Fabio Savo LIVE: a Tool for Checking Licenses Compatibility between Vocabularies and Data Guido Governatori, Ho-Pun Lam, Antonino Rotolo, Serena Villata, Ghislain Auguste Atemezing and Fabien Gandon The Map Generator Tool Valeria Fionda, Giuseppe Pirr`o and Claudio Gutierrez Named Entity Recognition using FOX Ren´e Speck and Axel-Cyrille Ngonga Ngomo A Linked Data Platform adapter for the Bugzilla issue tracker Nandana Mihindukulasooriya, Miguel Esteban-Gutierrez and Rau´l Garc´ıaCastro LED: curated and crowdsourced Linked Data on Music Listening Experiences Alessandro Adamou, Mathieu D’Aquin, Helen Barlow and Simon Brown 93 WhatTheySaid: Enriching UK Parliament Debates with Semantic Web Yunjia Li, Chaohai Ding and Mike Wald Multilingual Disambiguation of Named Entities Using Linked Data Ricardo Usbeck, Axel-Cyrille Ngonga Ngomo, Wencan Luo and Lars Wesemann The Wikipedia Bitaxonomy Explorer Tiziano Flati and Roberto Navigli Enhancing Web intelligence with the content of online video fragments Lyndon Nixon, Matthias Bauer and Arno Scharl EMBench: Generating Entity-Related Benchmark Data
Ekaterini Ioannou and Yannis Velegrakis
Demonstration of multi-perspective exploratory search with the Discovery Hub web application Nicolas Marie and Fabien Gandon Modeling and Monitoring Processes exploiting Semantic Reasoning Mauro Dragoni, Piergiorgio Bertoli, Chiara Di Francescomarino, Chiara Ghidini, Michele Nori, Marco Pistore, Roberto Tiella and Francesco Corcoglioniti WikipEvent: Temporal Event Data for the Semantic Web Ujwal Gadiraju, Kaweh Djafari Naini, Andrea Ceroni, Mihai Georgescu, Dang Duc Pham and Marco Fisichella Towards a DBpedia of Tourism: the case of Tourpedia Stefano Cresci, Andrea D’Errico, Davide Gazz`e, Angelica Lo Duca, Andrea Marchetti and Maurizio Tesconi Using Semantics for Interactive Visual Analysis of Linked Open Data Gerwald Tschinkel, Eduardo Veas, Belgin Mutlu and Vedran Sabol Exploiting Linked Data Cubes with OpenCube Toolkit Evangelos Kalampokis, Andriy Nikolov, Peter Haase, Richard Cyganiak, Arkadiusz Stasiewicz, Areti Karamanou, Maria Zotou, Dimitris Zeginis, Efthimios Tambouris and Konstantinos Tarabanis Detecting Hot Spots in Web Videos Jos´e Luis Redondo-Garc´ıa, Mariella Sabatino, Pasquale Lisena and Rapha¨el Troncy EUROSENTIMENT: Linked Data Sentiment Analysis J. Fernando S´anchez-Rada, Gabriela Vulcu, Carlos A. Iglesias and Paul Buitelaar Property-based typing with LITEQ Stefan Scheglmann, Martin Leinberger, Ralf L¨ammel, Ste↵en Staab and Matthias Thimm From Tale to Speech: Ontology-based Emotion and Dialogue Annotation of Fairy Tales with a TTS Output Christian Eisenreich, Jana Ott, Tonio Su¨ßdorf, Christian Willms and Thierry Declerck BIOTEX: A system for Biomedical Terminology Extraction, Ranking, and Validation Juan Antonio Lossio Ventura, Clement Jonquet, Mathieu Roche and Maguelonne Teisseire Visualizing and Animating Large-scale Spatiotemporal Data with ELBAR Explorer Suvodeep Mazumdar and Tomi Kauppinen A Demonstration of Linked Data Source Discovery and Integration Jason Slepicka, Chengye Yin, Pedro Szekely and Craig Knoblock 117 121 125 129 133 137 141 145 149 153 157 161 165
Developing Mobile Linked Data Applications Oshani Seneviratne, Evan Patton, Daniela Miao, Fuming Shih, Weihua Li, Lalana Kagal and Carlos Castillo A Visual Summary for Linked Open Data sources Fabio Benedetti, Laura Po and Sonia Bergamaschi EasyESA: A Low-e↵ort Infrastructure for Explicit Semantic Analysis Danilo Carvalho, Cagatay Calli, Andre Freitas and Edward Curry LODHub - A Platform for Sharing and Analyzing large-scale Linked Open Data Stefan Hagedorn and Kai-Uwe Sattler LOD4AR: Exploring Linked Open Data with a Mobile Augmented Reality Web Application Silviu Vert, Bogdan Dragulescu and Radu Vasiu PLANET: Query Plan Visualizer for Shipping Policies against Single SPARQL Endpoints Maribel Acosta, Maria Esther Vidal, Fabian Flo¨ck, Simon Castillo and Andreas Harth High Performance Linked Data Processing for Virtual Reality Environments Felix Leif Keppmann, Tobias K¨afer, Ste↵en Stadtmu¨ller, Ren´e Schubotz and Andreas Harth Analyzing Relative Incompleteness of Movie Descriptions in the Web of Data: A Case Study Wancheng Yuan, Elena Demidova, Stefan Dietze and Xuan Zhou A Semantic Metadata Generator for Web Pages Based on Keyphrase Extraction Dario De Nart, Carlo Tasso and Dante Degl’Innocenti A Multilingual SPARQL-Based Retrieval Interface for Cultural Heritage Objects Dana Dannells, Ramona Enache and Mariana Damova Extending Tagging Ontologies with Domain Specific Knowledge Frederic Font, Sergio Oramas, Gy¨orgy Fazekas and Xavier Serra DisambiguatingWeb Tables using Partial Data Ziqi Zhang On Linking Heterogeneous Dataset Collections Mayank Kejriwal and Daniel Miranker Scientific data as RDF with Arrays: Tight integration of SciSPARQL queries into MATLAB Andrej Andrejev, Xueming He and Tore Risch Measuring similarity in ontologies: a new family of measures Tahani Alsubait, Bijan Parsia and Uli Sattler 169 173 177 181 185 189 193 197 201 205 209 213 217 221 225
Towards Combining Machine Learning with Attribute Exploration for Ontology Refinement Jedrzej Potoniec, Sebastian Rudolph and Agnieszka Lawrynowicz ASSG: Adaptive structural summary for RDF graph data Haiwei Zhang, Yuanyuan Duan, Xiaojie Yuan and Ying Zhang Evaluation of String Normalisation Modules for String-based Biomedical Vocabularies Alignment with AnAGram Anique van Berne and Veronique Malaise Keyword-Based Semantic Search Engine Koios++ Bj¨orn Forcher, Andreas Giloj and Erich Weichselgartner Supporting SPARQL Update Queries in RDF-XML Integration Nikos Bikakis, Chrisa Tsinaraki, Ioannis Stavrakantonakis and Stavros Christodoulakis CURIOS: Web-based Presentation and Management of Linked Datasets Hai Nguyen, Stuart Taylor, Gemma Webster, Nophadol Jekjantuk, Chris Mellish, Je↵ Z. Pan and Tristan Ap Rheinallt The uComp Protege Plugin for Crowdsourcing Ontology Validation Florian Hanika, Gerhard Wohlgenannt and Marta Sabou Frame-Semantic Web: a Case Study for Korean Jungyeul Park, Sejin Nam, Youngsik Kim, Younggyun Hahm, Dosam Hwang and Key-Sun Choi SparkRDF: Elastic Discreted RDF Graph Processing Engine With Distributed Memory Xi Chen, Huajun Chen, Ningyu Zhang and songyang Zhang LEAPS: A Semantic Web and Linked data framework for the Algal Biomass Domain Monika Solanki Bridging the Semantic Gap between RDF and SPARQL using Completeness Statements Fariz Darari, Simon Razniewski and Werner Nutt COLINA: A Method for Ranking SPARQL Query Results through Content and Link Analysis Azam Feyznia, Mohsen Kahani and Fattane Zarrinkalam Licentia: a Tool for Supporting Users in Data Licensing on the Web of Data Cristian Cardellino, Serena Villata, Fabien Gandon, Guido Governatori, Ho-Pun Lam and Antonino Rotolo Automatic Stopword Generation using Contextual Semantics for Sentiment Analysis of Twitter Hassan Saif, Miriam Fernandez and Harith Alani The Manchester OWL Repository: System Description Nicolas Matentzoglu, Daniel Tang, Bijan Parsia and Uli Sattler 229 233 237 241 245 249 253 257 261 265 269 273 277 281 285 A Fully Parallel Framework for Analyzing RDF Data Long Cheng, Spyros Kotoulas, Tomas Ward and Georgios Theodoropoulos289 Objects as results from graph queries using an ORM and generated semantic-relational binding Marc-Antoine Parent Hedera: Scalable Indexing and Exploring Entities in Wikipedia Revision History Tuan Tran and Tu Ngoc Nguyen Evaluating Ontology Alignment Systems in Query Answering Tasks Alessandro Solimando, Ernesto Jimenez-Ruiz and Christoph Pinkel Using Fuzzy Logic For Multi-Domain Sentiment Analysis Mauro Dragoni, Andrea Tettamanzi and C´elia Da Costa Pereira AMSL — Creating a Linked Data Infrastructure for Managing Electronic Resources in Libraries Natanael Arndt, Sebastian Nuck, Andreas Nareike, Norman Radtke, Leander Seige and Thomas Riechert Extending an ontology alignment system with BioPortal: a preliminary analysis Xi Chen, Weiguo Xia, Ernesto Jimenez-Ruiz and Valerie Cross How much navigable is the Web of Linked Data? Valeria Fionda and Enrico Malizia A Framework for Incremental Maintenance of RDF Views of Relational Data Vˆania Vidal, Marco Antonio Casanova, Jose Monteiro, Narciso Arruda, Diego S´a and Valeria Pequeno Document Relation System Based on Ontologies for the Security Domain Janine Hellriegel, Hans Ziegler and Ulrich Meissen Representing Swedish Lexical Resources in RDF with lemon Lars Borin, Dana Dannells, Markus Forsberg and John P. Mccrae QASM: a Q&A Social Media System Based on Social Semantic Zide Meng, Fabien Gandon and Catherine Faron-Zucker A Semantic-Based Platform for Ecient Online Communication Zaenal Akbar, Jos´e Mar´ıa Garc´ıa, Ioan Toma and Dieter Fensel
SHAX: The Semantic Historical Archive eXplorer Michael Feldman, Shen Gao, Marc Novel, Katerina Papaioannou and Abraham Bernstein SemanTex: Semantic Text Exploration Using Document Links Implied by Conceptual Networks Extracted from the Texts Suad Aldarra, Emir Mun˜oz, Pierre-Yves Vandenbussche and Vit Novacek345 293 297 301 305 309 313 317 321 325 329 333 337 341 Towards a Top-K SPARQL Query Benchmark Shima Zahmatkesh, Emanuele Della Valle, Daniele Dell’aglio and Alessandro Bozzon 349 Exploring type-specific topic profiles of datasets: a demo for educational linked data Davide Taibi, Stefan Dietze, Besnik Fetahu and Giovanni Fulantelli 353 TEX-OWL: a Latex-Style Syntax for authoring OWL 2 ontologies Matteo Matassoni, Marco Rospocher, Mauro Dragoni and Paolo Bouquet357 Supporting Integrated Tourism Services with Semantic Technologies and Machine Learning Francesca Alessandra Lisi and Floriana Esposito Towards a Semantically Enriched Online Newspaper Ricardo Kawase, Eelco Herder and Patrick Siehndel Identifying Topic-Related Hyperlinks on Twitter Patrick Siehndel, Ricardo Kawase, Eelco Herder and Thomas Risse Capturing and Linking Human Sensor Observations with YouSense Tomi Kauppinen, Evgenia Litvinova and Jan Kallenbach An update strategy for the WaterFowl RDF data store Olivier Cur´e and Guillaume Blin Linking Historical Data on the Web Valeria Fionda and Giovanni Grasso User driven Information Extraction with LODIE
Anna Lisa Gentile and Suvodeep Mazdumar
QALM: a Benchmark for Question Answering over Linked Merchant Websites Data Amine Hallili, Elena Cabrio and Catherine Faron Zucker GeoTriples: a Tool for Publishing Geospatial Data as RDF Graphs Using R2RML Mappings Kostis Kyzirakos, Ioannis Vlachopoulos, Dimitrianos Savva, Stefan Manegold and Manolis Koubarakis 393 New Directions in Linked Data Fusion Jan Michelfeit and Jindich Mynarz Bio2RDF Release 3: A larger, more connected network of Linked Data for the Life Sciences Michel Dumontier, Alison Callahan, Jose Cruz-Toledo, Peter Ansell, Vincent Emonet, Fran¸cois Belleau and Arnaud Droit Infoboxer: Using Statistical and Semantic Knowledge to Help Create Wikipedia Infoboxes Roberto Yus, Varish Mulwad, Tim Finin and Eduardo Mena 361 365 369 373 377 381 385 389 397 401 405
The Topics they are a-Changing — Characterising Topics with TimeStamped Semantic Graphs Amparo E. Cano, Yulan He and Harith Alani Linked Data and facets to explore text corpora in the Humanities: a case study Christian Morbidoni Dexter 2.0 - an Open Source Tool for Semantically Enriching Data Diego Ceccarelli, Claudio Lucchese, Salvatore Orlando, Ra↵aele Perego and Salvatore Trani A Hybrid Approach to Learn Description Logic Ontology from Texts Yue Ma and Alifah Syamsiyah Identifying First Responder Communities Using Social Network Analysis John Erickson, Katherine Chastain, Zachary Fry, Jim Mccusker, Rui Yan, Evan Patton and Deborah McGuinness Exploiting Semantic Annotations for Entity-based Information Retrieval Lei Zhang, Michael F¨arber, Thanh Tran and Achim Rettinger Crawl Me Maybe: Iterative Linked Dataset Preservation Besnik Fetahu, Ujwal Gadiraju and Stefan Dietze A Semantics-Oriented Storage Model for Big Heterogeneous RDF Data Hyeongsik Kim, Padmashree Ravindra and Kemafor Anyanwu Approximating Inference-enabled Federated SPARQL Queries on Multiple Endpoints Yuji Yamagata and Naoki Fukuta VKGBuilder – A Tool of Building and Exploring Vertical Knowledge Graphs Tong Ruan, Haofen Wang and Fanghuai Hu Using the semantic web for author disambiguation - are we there yet? Cornelia Hedeler, Bijan Parsia and Brigitte Mathiak SHEPHERD: A Shipping-Based Query Processor to Enhance SPARQL Endpoint Performance Maribel Acosta, Maria Esther Vidal, Fabian Fl¨ock, Simon Castillo, Carlos Buil Aranda and Andreas Harth AgreementMakerLight 2.0: Towards Ecient Large-Scale Ontology Matching Daniel Faria, Catia Pesquita, Emanuel Santos, Isabel F. Cruz and Francisco Couto Extracting Architectural Patterns from Web data Ujwal Gadiraju, Ricardo Kawase and Stefan Dietze Xodx — A node for the Distributed Semantic Social Network Natanael Arndt and Sebastian Tramp 409 413 417 421 425 429 433 437 441 445 449 453 457 461 465
An Ontology Explorer for Biomimetics Database Kouji Kozaki and Riichiro Mizoguchi Semi-Automated Semantic Annotation of the Biomedical Literature Fabio Rinaldi Live SPARQL Auto-Completion Stephane Campinas 469 473 477 Life Stories as Event-based Linked Data:
Case Semantic National Biography Eero Hyvo¨nen, Miika Alonen, Esko Ikkala, and Eetu Ma¨kela¨
Semantic Computing Research Group (SeCo), Aalto University http://www.seco.tkk.fi/, firstname.lastname@aalto.fi Abstract. This paper argues, by presenting a case study and a demonstration on the web, that biographies make a promising application case of Linked Data: the reading experience can be enhanced by enriching the biographies with additional life time events, by proving the user with a spatio-temporal context for reading, and by linking the text to additional contents in related datasets.
This paper addresses the research question: How can the reading experience of biographies be enhanced using web technologies? Our research hypotheses is to apply the Linked Data (LD) approach to this, with the idea of providing the reader with a richer reading context than the biography document alone. The focus of research is on: 1) Data linking. Biographies can be linked with additional contextual data, such as links to the literal works of the person. 2) Data enriching. Data from different sources can be used for enriching the life story with additional events and data, e.g., with metadata about a historical event that the person participated in. 3) Visualization. LD can be visualized in useful ways. The life story can, e.g., be shown on maps and timelines. We tested the hypoheses in a case study1 where the Finnish National Biography2 (NB), a collection of 6,381 short biographies, is published as LD in a SPARQL endpoint with a demonstrational application based on its standard API. 2
To enrich and link biographical data with related datasets the data must be made
semantically interoperable, either by data alignments (using, e.g., Dublin Core and the
dumb down priciple) or by data transformations into a harmonized form [
A simple custom event extractor was created for transforming biographies into this model represented in RDF. The extractor first lemmatizes a biography and then analyzes its major parts: a textual story followed by systematically titled sections listing major achievements of the person, such as “works”, “awards”, and “memberships” as snippets. A snippet represents an event and typically contains mentions of years and places. For example, the biography of architect Alvar Aalto tells “WORKS: ...; Church of Muurame 1926-1929;...” indicating an artistic creation event. The named entity recognition tool of the Machinese4 NLP library is used for finding place names in the snippets, and Geonames is used for geocoding. Timespans of snippet events are found easily as numeric years or their intervals, and an actor of the events is the subject person of the biography. The result of processing a biography is a list of spatio-temporal CIDOC CRM events with short titles (snippet texts) related to the corresponding person. At the moment, the extractor uses only the snippets for event creation—more generic event extraction from the free biography narrative remains a topic of further research.
For a domain ontology, we reused the Finnish History Ontology HISTO by transforming it into CIDOC CRM. The new HISTO version contains 1,173 major historical events (E5 Event in CIDOC CRM) covering over 1000 years of Finnish history, and includes 80,085 activities (E7 Activity) of different kinds, such as armistice, election etc. Linked to these are 7,302 persons (E21 Person) and a few hundred organizations and groups, 3,290 places (E53 Place), and 11,141 time spans (E52 Time-span). The data originates from the Agricola timeline5 created by Finnish historians.
The extracted events were then enriched with events from external datasets as follows: 1) Persons in SNB and HISTO were mapped onto each other based on their names. This worked well without further semantic disambiguation since few different persons had similar names. NB and HISTO shared 921 persons p, and the biography of each p could therefore be enriched with all HISTO events that p was involved in. 2) There were 361 artistic creation events (e.g., publishing a book) of NB persons that could be extracted from Europeana Linked Open Data6 using the person as the creator. Related biographies could therefore be enriched with events pointing to Europeana contents. 3) The NB persons were involved in 263 instances of publications of the Project Gutenberg data7. Corresponding events could therefore be added into the biographies, and links to the original digitized publications be provided. 4) The NB persons were also linked to Wikipedia for additional information; again simple string matching produced good results. These examples demostrate how linked events can be extracted from other datasets and be used for enriching other biographical events. In the experiment, 116,278 spatio-temporal events were finally extracted for the NB biography records. 3
Biographies Enriched in a Spatio-temporal Context
Based on the enriched and linked biography data, a demonstrator was created
proving the end user with a spatio-temporal context for reading NB biographical data as
well as links to addtional content from related sources. Fig. 1 depicts the user
interface online8 with architect Alvar Aalto’s biography selected; the other 6,400 celebrities
can be selected from the alphabetical list above. On the left column, temporal events
extracted from the biography and related datasets are presented (in Finnish), such as
“1898 Birth”, and “1908-1916 Jyva¨skyla¨ Classical Lyceum”. The event “1930–1939:
Alvar Aalto created his famous functionalist works (Histo)” shows an external link to
HISTO for additional information. The events are also seen as bubbles on a timeline
at the bottom. The map in the middle shows the end-user the places related to the
biography events. By hovering the mouse over an event or its bubble the related event is
high-lighted and the map zoomed and centered around the place related to the event. In
this way the user can quickly get an overview about the spatio-temporal context of
Alvar Aalto’s life, and get links to additional sources of information. The actual biography
text can be read by clicking a link lower in the interface (not visible in the figure). The
user interface also performs dynamic SPARQL querying for additional external links.
In our demonstration, the BookSampo dataset and SPARQL endpoint [
The user interface for spatio-temporal lifeline visualization was implemented using AngularJS9 and D310 on top of the Linked Data Finland (LDF) data service11. 6 http://pro.europeana.eu/linked-open-data 7 http://datahub.io/dataset/gutenberg 8 http://www.ldf.fi/dataset/history/map.html 9 http://angularjs.org 10 http://d3js.org 11 Cf. http://www.ldf.fi/dataset/history/ for dataset documentation and SPARQL endpoint
Discussion, Related Work, and Future Research Our case study suggests that biography publication is a promising application case for LD. The event-based modeling approach was deemed useful and handy, after learning basics of the fairly complex CIDOC CRM model. The snippet events could be extracted and aligned with related places, times, and actors fairly accurately using simple stringbased techniques. However, the results of event extraction and entity linking have not been evaluated formally, and it is obvious that problems grow with larger datasets and when analysing free text—these issues are a topic of future research.
Biographical data has been studied by genealogists (e.g., (Event) GEDCOM12), CH
organizations (e.g., the Getty ULAN13), and semantic web researchers (e.g., BIO
ontology14). Semantic web event models include, e.g., Event Ontology [
Bernd Loeber, and Andreas Lommatzsch
Technische Universita¨t Berlin
Ernst-Reuter-Platz 7, D-10587 Berlin, Germany {sarnold,damian.burke,tobias.m.doersch,bernd.loeber, andreas.lommatzsch}@mailbox.tu-berlin.de Abstract. Due to the overwhelming amount of news articles from a growing number of sources, it has become nearly impossible for humans to select and read all articles that are relevant to get deep insights and form conclusions. This leads to a need for an easy way to aggregate and analyze news articles efficiently and visualize the garnered knowledge as a base for further cognitive processing.
The presented application provides a tool to satisfy said need. In our approach we use semantic techniques to extract named entities, relations and locations from news sources in different languages. This knowledge is used as the base for data aggregation and visualization operators. The data operators include filtering of entities, types and date range, detection of correlated news topics for a set of selected entities and geospatial analysis based on locations. Our visualization provides a time-based graphical representation of news occurrences according to the given filters as well as an interactive map which displays news within a perimeter for the different locations mentioned in the news articles. In every step of the user process, we offer a tag cloud that highlights popular results and provide links to the original sources including highlighted snippets. Using the graphical interface, the user is able to analyze and explore vast amounts of fresh news articles, find possible relations and perform trend analysis in an intuitive way. 1
Comprehensive news analysis is a common task for a broad range of recipients. To overlook the overwhelming amount of articles that are published in the Web every hour, new technologies are needed that help to classify, search and explore topics in real time. Current approaches focus on automated classification of documents into expertdefined categories, such as politics, business or sports. The results need to be tagged manually with meta-information about locations, people and current news topics. The simple model of categories and tags, however, is not detailed enough to suit temporal or regional relationships and it cannot bridge the semantic gap that the small subset of tagged information opens. The challenge for machine-driven news analysis consists of two parts. First, an extractor needs to be able to identify the key concepts and entities mentioned in the documents and to find the most important relationships between them. Second, an intuitive way for browsing the results with support for explorative discovery of relevant topics and emerging trends needs to be developed.
We present a semantic approach that abstracts from multi-lingual representation of facts and enriches extracted information with background knowledge. Our implementation utilizes natural language processing tools for the extraction of named entities, relations and semantic context. Open APIs are used to augment further knowledge (e.g. geo-coordinates) to the results. Our application visualizes the gained knowledge and provides time-based, location-based and relationship-based exploration operators. The relationship between original news documents and aggregated search results is maintained throughout the whole user process.
In Section 2, we give an overview on existing projects of similar focus. Our knowledgebased approach and the implementation is introduced in Section 3. The user interaction and visualization operators are discussed in Section 4. We conclude in Section 5. 2
We start with an overview on existing projects in the field of semantic news visualization. The following projects are related to our approach on a conceptual or visual level.
MAPLANDIA1 visualizes news for a specific date beginning in 2005 on a map. The system uses the BBC news feed as its only source to deliver markers and outlines for the countries that were mentioned in the news on a specified date. Additionally, it offers a list of the news for the day. However, by using only one marker the application is unable to visualize news on a more detailed and fine-grained level. MAPLANDIA also does not offer any possibility to limit the displayed visualizations to a certain region of interest. The application offers news in only one language and source for a specific day.
The idea behind the SPIGA-SYSTEM [
In this section, we explain our semantic approach to news aggregation. In contrast
to classical word- or tag-based indexing, we focus on semantic features that we
extract from daily news documents. To handle the linguistic complexity of this problem,
we utilize information extraction techniques for natural language [
The alignment selection and repair module ensures that the final alignment has
the desired cardinality and that it is coherent (i.e., does not lead to the
violation of restrictions of the ontologies) which is important for several applications.
AML’s approximate alignment repair algorithm features a modularization step
which identifies the minimal set of classes that need to be analyzed for coherence,
thus greatly reducing the scale of the repair problem [
Performance
AML 1.0 achieved top results in the 2013 edition of the Ontology Alignment
Evaluation Initiative (OAEI) [
AML 2.0 is more e↵ective than its predecessor (thanks to the improved handling of background knowledge, the richer data structures and the addition of new matching algorithms) without sacrificing eciency, so we expect it to perform even better in this year’s edition of the OAEI.
Acknowledgments DF, CP, ES and FMC were funded by the Portuguese FCT through the SOMER project (PTDC/EIA-EIA/119119/2010) and the LASIGE Strategic Project (PEst-OE/EEI/UI0408/2014). The research of IFC was partially supported by NSF Awards CCF–1331800, IIS–1213013, IIS–1143926, and IIS–0812258 and by a UIC-IPCE Civic Engagement Research Fund Award.
Ujwal Gadiraju, Ricardo Kawase, and Stefan Dietze L3S Research Center, Leibniz University Hannover, Germany
{gadiraju, kawase, dietze}@L3S.de Abstract. Knowledge about the reception of architectural structures is crucial for architects or urban planners. Yet obtaining such information has been a challenging and costly activity. With the advent of the Web, a vast amount of structured and unstructured data describing architectural structures has become available publicly. This includes information about the perception and use of buildings (for instance, through social media), and structured information about the building’s features and characteristics (for instance, through public Linked Data). In this paper, we present the first step towards the exploitation of structured data available in the Linked Open Data cloud, in order to determine well-perceived architectural patterns. 1
Introduction and Motivation Urban planning and architecture encompass the requirement to assess the popularity or perception of built structures (and their evolution) over time. This aids in understanding the impact of a structure, identify needs for restructuring, or to draw conclusions useful for the entire field, for instance, about successful architectural patterns and features. Thus, information about how people think about a building that they use or see, or how they feel about it, could prove to be invaluable information for architects, urban planners, designers, building operators, and policy makers alike. For example, keeping track of the evolving feelings of people towards a building and its surroundings can help to ensure adequate maintenance and trigger retrofit scenarios where required. On the other hand, armed with prior knowledge of specific features that are well-perceived by the public, builders and designers can make better-informed design choices and predict the impact of building projects.
The Web contains structured information about particular building features, for example, size, architectural style, built date, etc. of certain buildings through public Linked Data. Here in particular, reference datasets such as Freebase1 or DBpedia2 o↵ er useful structured data describing a wide range of architectural structures.
The perception of an architectural structure itself has historically been studied to
be a combination of the aesthetic as well as functional aspects of the structure [
In this paper, we present our work regarding the alignment of Influence Factors with structured data. Firstly, we identified the influence factors for a predefined set of architectural structures. Secondly, we align these factors with structured data from DBpedia. This work serves as a first step towards semantic enhancement of the architectural domain, which can support semantic classification of architectural structures, semantic analysis, and ranking, amongst others. 2
Crowdsourcing Influential Factors and Ranking Buildings
Recent research works in the field of Neuroscience [
An initial survey was conducted using LimeService3 with a primary focus on the expert community of architects, builders and designers in order to determine influential factors. The survey administered 32 questions spanning over the background of the participants and their feelings about certain buildings, of di↵ erent types (bridges, churches, skyscrapers, halls and airports). We received 42 responses from the expert community. The important influential factors that surfaced from the responses of the survey are presented below.
For bridges, churches, skyscrapers and halls: history, surroundings, materials, size, personal experiences, and level of detail. For airports: Ease of access, e ciency, appearance, choice/availability, facilities, miscellaneous facilities and size.
Based on these influential factors we acquired perception scores of buildings on a Likert-scale, through crowdsourcing. By aggregating and normalizing these scores between 0 and 1, we arrived at a ranked list of buildings of each type within our dataset. 3
Correlating Influential Factors with Relevant Structured Data In order to determine patterns in the perception of well-received structures (as per the building rankings), we correlate the influential factors of buildings with concrete properties and values from DBpedia. 3 http://www.limeservice.com/
Table 1 depicts some of the properties that are extracted from the DBpedia knowledge graph in order to correlate the influence factors corresponding to each structure with specific values.
By doing so, we can analyze the well-received patterns for architectural structures at a finer level of granularity, i.e., in terms of tangible properties. In order to extract relevant data from DBpedia for each structure in our dataset, we first collect a pool of properties that correspond to each of the influence factors as per the building type (see Table 1). In the next step, by traversing the DBpedia knowledge graph leading to each structure in our dataset, we try to extract corresponding values for each of the properties identified. The properties thus extracted semi-automatically, are limited to those available on DBpedia. In addition, it is important to note that not all structures of a particular type have the same properties available on DBpedia. Therefore, although all the identified values accurately correspond to the structure, the coverage itself is restricted to the data available on DBpedia (see Table 2). By correlating the influence factors to specific DBpedia properties, we can identify patterns for well-perceived architectural structures. In order to demonstrate how such observed patterns for architectural structures can be used, we choose the influence factor ‘size’ of the structure. Although, this approach can be directly extended to other influence factors and across di↵ erent kinds of architectural structures, due to the limited space we restrict ourselves to showcasing this influence factor.
We observe that for each airport, we can extract indicators of size using the DBpedia property dbpedia-owl:runwayLength. Similarly, in case of bridges the influence factor ‘size’ can be represented using the DBpedia properties dbpedia-owl:length, dbpedia-owl:width and dbpedia-owl:mainspan, for halls we can use the DBPedia properties dbprop:area and dbprop:seatingCapacity, while we can use dbpedia-owl:floorCount, and dbprop:height to consolidate the well-perceived patterns for Skyscrapers. We thereby extract corresponding property values for each structure in our dataset4 using the DBpedia knowledge graph.
Figure 1 depicts the influence of size in the perception of halls. We observe that halls with a seating capacity between 1000-4000 people are well-perceived with the positive perception, varying between 0.1 and 1. The perception scores are obtained through the aggregation of results from the crowdsourcing process. Similarly, as a result of the quantitative analysis of churches, by leveraging the rankings and correlating with the property dbpedia-owl:architecturalStyle, we find that the most well-received styles of churches in Germany are (i) Gothic, (ii) Gothic Revival, and (iii) Romanesque.
With this, we demonstrated that by correlating building characteristics with extracted data from DBpedia, one is able to compute and analyze architectural structures quantitatively. Thus, our main contribution includes semantic analysis and quantitative measurement of public perception of architectural structures based on structured data. As future work, we plan to develop algorithms that exploit properties from the structured data on the web in order to provide multi-dimensional architectural patterns like ‘skyscrapers with x size,y uniqueness, and z materials used are best perceived’, which architects and urban planners can benefit from. 4 Our dataset and building rankings: http://data-observatory.org/building-perception/
Kouji KOZAKI1 and Riichiro MIZOGUCHI2 1The Institute of Scientific and Industrial Research, Osaka University 8-1 Mihogaoka, Ibaraki, Osaka, 567-0047 Japan
kozaki@ei.sanken.osaka-u.ac.jp 2Japan Advanced Institute of Science and Technology 1-1 Asahidai, Nomi, Ishikawa 923-1292, Japan
mizo@jaist.ac.jp Abstract. Biomimetics contributes to innovative engineering by imitating the models, systems, and elements of nature. For biomimetics research, it is important to develop biomimetics database including widely varied knowledge across different domains such as biology and engineering. Interoperability of knowledge among those domains is necessary to create such a database. For this purpose, the authors are developing a biomimetics ontology which bridge gaps between biology and engineering. In this demo, the authors shows an ontology exploration tool for biomimetics database. It is based on linked data techniques and allows the users to find important keywords so that they can search meaningful knowledge from various databases.
Keywords: ontology, linked data, biomietics, database, semantic search
1
Learning from nature aids development of technologies. Awareness of this fact has
been increasing, and biomimetics1 [
Because ontologies clarify concepts that appear in target domains [
A Biomimetics Ontology Before we began developing a biomimetics ontology, we conducted interviews with engineers working with biomimetics regarding their requirements for biomimetics database search. When we asked, “What do you want to search for in a biomimetic database?” they said they wanted to search for organisms or organs that perform functions that they were trying to develop in their new products. In fact, most successful examples are imitations of capabilities that organisms possess, such as the water repellency of a lotus and the adhesiveness of a gecko’s feet. Therefore, we proposed that it is important to search the biomimetic database for functions or goals that they want to achieve.
On the other hand, someone engaged in cooperative research with engineers and biologists reported that engineers do not have knowledge that is very familiar to biologists. For instance, when an engineer had a question about functions of projections shown in an electron microscopy image of an insect, a biologist (entomologist) suggested that it could have an anti-slip capability, because the insect often clings to slippery surfaces. This suggests that a biomimetic ontology must bridge knowledge gaps between engineers and biologists.
Considering the requirements discussed in the above, we set the first requirement for biomimetics ontology as to be able to search for related organisms by the function the user wants to perform. At the same time, we propose that it should support various viewpoints to bridge gaps among domains. As a result, we built a biomimetics ontology that includes 379 concepts (classes) and 314 relationships (properties), except for the is-a (sub-class-of) relation. For example, Organism may have relationships such as Ecological environment, Characteristic behavior, Characteristic structure, Characteristic function, Region Part, and Goal may have relationships such as Structure on which to base and Related function. Other top level concepts includes Behavior, Act, Function, Process, Structure, Living environment, and so on. 3
An Ontology Explorer for Biomimetics Database
We developed the ontology explorer for biomimetics database based on an ontology
exploration techniques proposed in our previous work [
Exploration of an ontology can be performed by choosing arbitrary concepts from which multi-perspective conceptual chains can be traced, according to the explorer’s
Fig.1 A snapshot of the Ontology Explorer for Biomimetics Database. databases. Though the current version supports only a few LODs and databases, it can be easily extended to others. 4
Conclusion and Future work This article outlined an ontology explorer for biomimetics database. Since the current version of the system is a prototype, it uses only a small ontology and has limits on the conditions of exploration. However, it was well received by researchers on biomimetics. In fact, one of them said that the resulting path from Antifouling to Sandfish shown in Fig.1 was unexpected one for him. This suggests that the proposed system could contribute innovations in biomimetics. The researchers also plan to use the biomimetics ontology and system as an interactive index for a biomimetics textbook.
Future work includes extensions of the biomimetics ontology and the exploration system. For the former, we plan to use documents on biomimetics and existing linked data related to biology and considering some methods for semi-automatic ontology building using them. For later, we are exploring potentially useful patterns through discussion with biomimetics researchers and ontology engineers.
There are many approaches to Semantic Search using SPARQL. For example,
Ferré proposes QFS (Query-based Faceted Search) for support in navigating faceted
search using LISQL (Logical Information System Query Language) [
The current version of the proposed system is available at the URL; http://biomimetics.hozo.jp/ontology_db.html .
Acknowledgements This work was supported by JSPS KAKENHI Grant Number 25280081 and 24120002. Semi-Automated Semantic Annotation of the Biomedical Literature
Fabio Rinaldi Institute of Computational Linguistics, University of Zurich
fabio.rinaldi@uzh.ch Abstract. Semantic annotations are a core enabler for ecient retrieval of relevant information in the life sciences as well in other disciplines. The biomedical literature is a major source of knowledge, which however is underutilized due to the lack of rich annotations that would allow automated knowledge discovery.
We briefly describe the results of the SASEBio project (Semi Automated Semantic Enrichment of the Biomedical Literature) which aims at adding semantic annotations to PubMed abstracts, in order to present a richer view of the existing literature. 1 The scientific literature contains a wealth of knowledge which however cannot be easily used automatically due to its unstructured nature. In the life sciences, the problem is so acutely felt that large budgets are invested into the process of literature curation, which aims at the construction of structured databases using information mostly manually extracted from the literature. There are several dozens of life science databases, each specializing on a particular subdomain of biology. Examples of well-known biomedical databases are UniProt (proteins), EntrezGene (genes), NCBI Taxonomy (species), IntAct (protein interactions), BioGrid (protein and genetic interactions), PharmGKB (drug-gene-disease relations), CTD (chemical-gene-disease relations), and RegulonDB (regulatory interactions in E. coli).
The OntoGene group1 aims at developing text mining technologies to support the process of
literature curation, and promote a move towards assisted curation. By assisted curation we mean a
combination of text mining approaches and the work of an expert curator, aimed at leveraging the
power of text mining systems, while retaining the high quality associated with human expertise.
We believe that it is possible to gradually automate much of the most repetitive activities of the
curation process, and therefore free up the creative resources of the curators for more challenging
tasks, in order to enable a much more ecient and comprehensive curation process. Our text
mining system specializes in the detection of entities and relationships from selected categories,
such as proteins, genes, drugs, diseases, chemicals. OntoGene derives some of its resources from
life sciences databases, thus allowing a deeper connection between the unstructured information
contained in the literature and the structured information contained in databases. The quality of
the system has been tested several times through participation in some of the community-organized
evaluation campaigns, where it often obtained top-ranked results. We have also implemented a
platform for assisted curation called ODIN (OntoGene Document INspector) which aims at serving
the needs of the curation community. The usage of ODIN as a tool for assisted curation has been
tested within the scope of collaborations with curation groups, including PharmGKB [
Assisted curation is also of utility in the process of pharmaceutical drug discovery. Many text mining tasks in drug discovery require both high precision and high recall, due to the importance of comprehensiveness and quality of the output. Text mining algorithms, however, cannot often achieve both high precision and high recall, sacrificing one for the other. Assisted curation can be paired with text mining algorithms which have high recall and moderate precision to produce results that are amenable to answer pharmaceutical problems with only a reasonable e↵ort being allocated to curation. Methods The Ontogene system is based on a pipeline architecture (see figure 1), which includes, among others, modules for entity recognition and relation extraction. Some of the modules are rule-based (e.g. lexical lookup with variants) while others use machine-learning approaches (e.g. maximum entropy techniques). The initial step consists in the annotation of names of relevant domain entities in biomedical literature (currently the system considers proteins, genes, species, experimental methods, cell lines, chemicals, drugs and diseases). These names are sourced from reference databases and are associated with their unique identifiers in those databases, thus allowing resolution of synonyms and cross-linking among di↵erent resources.
One of the problems with sourcing resources
from several databases is the possible
inconsistencies among them. The fact that domain knowledge
BioC XML input is scattered across dozens of data sources,
occa# sionally also with some incompatibilities among
(0) Wait request / validate BioC them, is a severe problem in the life sciences.
Ide# ally these resources should be integrated in a
sin(1) Read input with PyBioC reader gle repository, as some projects are attempting
# to do (e.g. OpenPhacts [
# We train our system using the knowledge
pro(4) Sentence splitting + tokenization vided by life sciences databases as our gold
stan# dard, instead of hand-labeled corpora, since we
(5) Term annotation believe that the scope and size of manually
anno# tated corpora, however much e↵ort has been
in(6) Extract terms vested in creating them, is not sucient to capture
# the wide variety of linguistic phenomena that can
(7) Merge tokens be encountered in the full corpus of biomedical
lit# erature, let alone other types of documents, such
(8) Entity disambiguation as internal scientific reports in the pharma
indus# try, which are not represented at all in annotated
(9) Compute concept relevance corpora. For example, PubMed currently contains
# more than 23 million records, while the entire
(10) Filter concepts by score set of all annotated publications probably barely
# reaches a few thousands, most of them sparsely
(11) Compute relation relevance annotated for very specific purposes.
We generate interaction candidates using
co(12) Filter relati#ons by score occurence of entities within selected syntactic
units (typically sentences). An additional step of
# syntactic parsing using a state-of-the-art
depen(13) Annotate OGXML for visualization dency parser allows us to derive specialized
fea# tures in order to increase precision. The details of
(14) Add annotations to PyBioC writer the algorithm are presented in [
BioC XML output connects the two participating entities. All
availFig. 1. Schema of the OntoGene pipeline able lexical and syntactic information is used in
order to provide an optimized ranking for
candidate interactions. The ranking of relation
candidates is further optimized by a supervised machine learning method described in detail in [
The OntoGene pipeline identifies all relevant entities mentioned in the paper, and their
interactions, and reports them back to the user as a ranked list, where the ranking criteria is the system’s
own confidence for the specific result. The confidence value is computed taking into account
several factors, including the relative frequency of the term in the article, its general frequency in
PubMed, the context in which the term is mentioned, and the syntactic configuration among two
interacting entities (for relationships). A detailed description of the factors that contribute to the
computation of the confidence score can be found in [
The user can choose to either inspect the results, using the ODIN web interface, or to have them delivered back via the RESTful web service in BioC XML format, for further local processing. ODIN (OntoGene Document Inspector) is a flexible browser-based client application which interfaces with the OntoGene server. The curator can use the features provided by ODIN to visualize selected annotations, together with the statements from which they were derived, and, if necessary, add, remove or modify them. Once the curator has validated a set of candidate annotations, they can be exported, using a standard format (e.g. CSV, RDF), for further processing by other tools, or for inclusion in a reference database, after a suitable format conversion. In case of ambiguity, the curator is o↵ered the opportunity to correct the choices made by the system, at any of the di↵erent levels of processing: entity identification and disambiguation, organism selection, interaction candidates. The curator can access all the possible readings given by the system and select the most accurate.
As a way to verify the quality of the core text mining functionalities of the OntoGene
system, we have participated in a number of text mining evaluation campaigns [
Since internally the original database identifiers are used to represent the entities and interactions detected by the system, the annotations can be easily converted into a semantic web format, by using a reference URI for each domain entity, and using RDF statements to express interactions. While it is possible to access the automatically generated annotations for further processing by a reasoner or integrator tool, we strongly believe that at present a process of semi-automated validation is preferable and would lead to better data consistency.
Acknowledgments. The OntoGene group is partially supported by the Swiss National Science Foundation (grant 105315 130558/1 to Fabio Rinaldi) and by the Data Science Group at Ho↵mann-La Roche, Basel, Switzerland. References Live SPARQL Auto-Completion
St´ephane Campinas Insight Centre for Data Analytics, National University of Ireland, Galway
stephane.campinas@insight-centre.org Abstract. The amount of Linked Data has been growing increasingly. However, the ecient use of that knowledge is hindered by the lack of information about the data structure. This is reflected by the diculty of writing SPARQL queries. In order to improve the user experience, we propose an auto-completion library1 for SPARQL that suggests possible RDF terms. In this work, we investigate the feasibility of providing recommendations by only querying the SPARQL endpoint directly. 1 The Linking Open Data movement has brought a tremendous amount of data available to the general user. The available knowledge spans a wide range of domains, from life sciences to films. However, using SPARQL to search through this knowledge is a tedious process, not only because of the syntax barrier but mainly due to the schema heterogeneity of the data. The expression of an information need in SPARQL is dicult due to the schema being generally unknown to the user as well as an heterogeneous of several vocabularies.
A common solution is for the user to manually gain knowledge about the data
structure, i.e., what predicates and classes are used, by executing additional queries
in parallel to the main one. The paper [
In Section 2 we discuss related works regarding auto-completion for SPARQL.
In Section 3 we present the proposed approach. In Section 4 we report an evaluation
of the system based on query logs of DBpedia.
2
Over the years, many contributions have been done towards facilitating the use of
SPARQL, either visually [
Recommendation Scope. While building a SPARQL query, not all triple patterns are relevant for the recommendation. Therefore, we define the scope as the connected component that contains the POF. Figure 1a depicts a SPARQL query where the POF is associated with the variable “?s”: it seeks possible predicates that occur with a “:Person” having the predicate “:name”. Figure 1b depicts the previous SPARQL query reduced to its recommendation scope. Indeed, the pattern on line 4 is removed since it is not part of the connected component containing the POF. 1 2 3 4 5
SELECT * { ? s a : Person ;
: name ? name ; < .
? o a : Document SELECT ? POF { ? s a : Person ; : name ? name ; ? POF [] . (a) A query with ‘<’ as the POF (b) Scope of the query
Fig. 1: Query auto-completion
Recommendation Capabilities. The scope may include content-specific terms, e.g,
resources and filters, unlike to [
Evaluation
Systems. In this section, we evaluate the recommendations returned by the proposed
system, that we refer to as “S1”, against the ones provided by the approach in [
Graphs. We loaded into an endpoint the English part of the Dbpedia3.34 dataset, which consists of 167 199 852 triples. The graph summary consists of 29 706 051 triples, generated by grouping resources sharing the same set of classes.
Endpoint. We used a Virtuoso5 SPARQL endpoint. The endpoint is deployed on a server with 32GB of RAM and with SSD drives.
Comparison. For each group of query complexity QC, we report in Table 1 the
results of the evaluation, with J 1 (resp., J 2) the average Jaccard similarity for the
system S1 (resp., S2); and T 1 (resp., T 2) the average response-time in ms for the
system S1 (resp., S2). The reported values are the averages over 5 runs. We can see
that as the LIMIT gets larger, the higher the Jaccard similarity becomes.Since the
graph summary used in S2 is a concise representation of the graph structure, the
data sample at a certain LIMIT value contains more terms than in S1. However,
this impacts negatively on the quality of S2 as reflected by the values of J2. This
shows the graph summary is subject to errors [
T 1 T 2
2 107 81 108 81 141 91
1-3 101 108 103 105 126 105
Acknowledgement This material is based upon works supported by the European FP7 projects LOD2 (257943).
Bibliography