Open Data Heterogeneity, Quality and Scale Presentation of the Open - PowerPoint PPT Presentation

Open Data Heterogeneity, Quality and Scale Presentation of the Open Data Research Group Zohra Bellahsene, Anne Laurent, François Scharffe, Konstantin Todorov PhD students : Manel Achichi, Mohamed Ben Ellefi, Abdel Nasser Tigrine Master students: Imène Chentli, Mykael Vigo LIRMM / University of Montpellier July 2015 Z. Bellahsene, A. Laurent, F . Scharffe, K. Todorov 1 / 29

Outline 1 The Big Picture 2 Ontology Matching 3 Data Lifting and Linking 4 Dataset and Vocabulary Recommendation 5 Data Access Z. Bellahsene, A. Laurent, F . Scharffe, K. Todorov 2 / 29

Outline 1 The Big Picture 2 Ontology Matching 3 Data Lifting and Linking 4 Dataset and Vocabulary Recommendation 5 Data Access Z. Bellahsene, A. Laurent, F . Scharffe, K. Todorov 3 / 29

The Big Picture The Open Data Research Group Z. Bellahsene, A. Laurent, F . Scharffe, K. Todorov 4 / 29

The Big Picture The Open Data Research Group Z. Bellahsene, A. Laurent, F . Scharffe, K. Todorov 5 / 29

Outline 1 The Big Picture 2 Ontology Matching 3 Data Lifting and Linking 4 Dataset and Vocabulary Recommendation 5 Data Access Z. Bellahsene, A. Laurent, F . Scharffe, K. Todorov 6 / 29

Ontology Matching Borrowed from a tutorial by S. Staab and A. Hotho. Z. Bellahsene, A. Laurent, F . Scharffe, K. Todorov 7 / 29

Ontology Matching A Generic Framework for Ontology Matching and Evaluation Ontologies are created in a decentralized , strongly human biased manner. Many ontologies describing the same domain of interest => ontology heterogeneity: • syntactic • terminological • conceptual / structural => Ontology Matching: detect the semantic correspondences between the elements of two ontologies. Z. Bellahsene, A. Laurent, F . Scharffe, K. Todorov 8 / 29

Ontology Matching A Generic Framework for Ontology Matching and Evaluation [Ngo, Bellahsene, Todorov. ESWC 2013] Z. Bellahsene, A. Laurent, F . Scharffe, K. Todorov 9 / 29

Ontology Matching YAM++ (not) Yet Another Matcher Many matching systems are out there. Here are some of the pluses of YAM++: • Automatic configuration: similarity measures selection, tuning, and combination • A novel terminological measure based on Tversky’s similarity • Able to deal with large ontologies [Ngo, Bellahsene, EKAW 2012], [http://oaei.ontologymatching.org] Z. Bellahsene, A. Laurent, F . Scharffe, K. Todorov 10 / 29

Ontology Matching YAM++ (not) Yet Another Matcher Among the best performing systems in the current state-of-the-art (Cf. reports of the Ontology Alignment Evaluation Initiative (OAEI) 1 ) 1 http://oaei.ontologymatching.org Z. Bellahsene, A. Laurent, F . Scharffe, K. Todorov 11 / 29

Ontology Matching A Fuzzy Framework for Ontology Matching Consider the (inherently) vague nature of concepts and their alignemnts • Provide the missing implicit background knowledge • Most matching procedures produce 1:1 mappings: often we will not be interested in the best (exact) match, but would like to find related yet not equivalent concepts • A fuzzy set representation of the concepts, construction of a fuzzy common ontology • Infer (fuzzy) relations between cross-ontology concepts [Todorov, Hudelot, Popescu, Geibel. IJUFKS 2014] Z. Bellahsene, A. Laurent, F . Scharffe, K. Todorov 12 / 29

Ontology Matching Cross-lingual Ontology Matching Motivation • No one-to-one correspondence between the majority of terms across different languages • Machine translation still tolerates low precision levels • Machine learning ? – No large training corpora with OM data Use of background knowledge PhD project of Abdel Nasser Tigrine Z. Bellahsene, A. Laurent, F . Scharffe, K. Todorov 13 / 29

Outline 1 The Big Picture 2 Ontology Matching 3 Data Lifting and Linking 4 Dataset and Vocabulary Recommendation 5 Data Access Z. Bellahsene, A. Laurent, F . Scharffe, K. Todorov 14 / 29

Data Lifting and Linking The Datalift projet [Scharffe et al. 2012, http://datalift.org] Z. Bellahsene, A. Laurent, F . Scharffe, K. Todorov 15 / 29

Data Lifting and Linking A General Data Linking Framework [Ferrara, Nikolov, Scharffe. IJSW 2011] Z. Bellahsene, A. Laurent, F . Scharffe, K. Todorov 16 / 29

Data Lifting and Linking The DOREMUS project Semantic web technologies for application-oriented use and reuse of musical data Leading cultural partner institutions: BnF, Radio France, Philharmonie de Paris Collaboration with Eurecom (Nice). PhD project of Manel Achichi Z. Bellahsene, A. Laurent, F . Scharffe, K. Todorov 17 / 29

Data Lifting and Linking The DOREMUS project Entity of interest: a musical work: — physical manifestations (recordings, scores) and — all the events that define them (creation, publication, performance). — relations between works — relations between events Z. Bellahsene, A. Laurent, F . Scharffe, K. Todorov 18 / 29

Data Lifting and Linking The DOREMUS project www.doremus.org Applications: tools in support of the selection of musical works, able to suggest original musical programming for specialized radia, choosing works and interpretations to illustrate the biography of a composer, a historical period, culture or genre. Z. Bellahsene, A. Laurent, F . Scharffe, K. Todorov 19 / 29

Data Lifting and Linking Lifting data from Tweets TEWS: Twitter Events on the Semantic Web: extraction and modeling of events from the Twitter stream Use of the Wikitimes ontology for events representation. Master’s project of Mykael Vigo Z. Bellahsene, A. Laurent, F . Scharffe, K. Todorov 20 / 29

Outline 1 The Big Picture 2 Ontology Matching 3 Data Lifting and Linking 4 Dataset and Vocabulary Recommendation 5 Data Access Z. Bellahsene, A. Laurent, F . Scharffe, K. Todorov 21 / 29

Dataset Recommendation for Linking ...Any candidates? Towards an automatic discovery and recommendation of candidate datasets for linking PhD project of Mohamed Ben Ellefi Z. Bellahsene, A. Laurent, F . Scharffe, K. Todorov 22 / 29

Dataset Recommendation for Linking Given a dataset d , return a (possibly) ranked set of WoD datasets with respect to their relevance to the dataset d in view of the linking task. Towards dataset profiling: definition of a collection of characteristics that allow to • describe in the best possible way a dataset • separate this dataset in the best possible way from other datasets • many (statistical) characteristics of interest (scale, coverage, data values range, degree of connectedness, attribute entropy, etc...) Collaboration avec L3S Hannover. Z. Bellahsene, A. Laurent, F . Scharffe, K. Todorov 23 / 29

Vocabulary Recommendation with Datavore The Datalyse project Data modeling with Datavore , the data vo cabulary re commender. http://www.datalyse.fr Z. Bellahsene, A. Laurent, F . Scharffe, K. Todorov 24 / 29

Outline 1 The Big Picture 2 Ontology Matching 3 Data Lifting and Linking 4 Dataset and Vocabulary Recommendation 5 Data Access Z. Bellahsene, A. Laurent, F . Scharffe, K. Todorov 25 / 29

Data Access The AgroLD project (Agronomic Linked Data) Collaboration with IBC—the Institute of Computational Biology (Montpellier). Master’s project of Imène Chentli Z. Bellahsene, A. Laurent, F . Scharffe, K. Todorov 26 / 29

The Big Picture The Open Data Research Group Z. Bellahsene, A. Laurent, F . Scharffe, K. Todorov 27 / 29

References [Bizer, Heath, Bernes-Lee. IJWS 2009] Christian Bizer, Tom Heath, Tim Berners-Lee: Linked Data - The Story So Far. Int. J. Semantic Web Inf. Syst. 5(3): 1-22 (2009) [Ferrara, Nikolov, Scharffe. IJSW 2011] Alfio Ferrara, Andriy Nikolov, Franois Scharffe: Data Linking for the Semantic Web. Int. J. Semantic Web Inf. Syst. 7(3): 46-76 (2011) [Ngo, Bellahsene, EKAW 2012] DuyHoa Ngo, Zohra Bellahsene: YAM++ : A Multi-strategy Based Approach for Ontology Matching Task. EKAW 2012: 421-425 [Ngo, Bellahsene, Todorov. ESWC 2013] DuyHoa Ngo, Zohra Bellahsene, Konstantin Todorov: Opening the Black Box of Ontology Matching. ESWC 2013: 16-30 [Nikolov et al. JIST 2011] Andriy Nikolov, Mathieu d’Aquin, Enrico Motta: What Should I Link to? Identifying Relevant Sources and Classes for Data Linking. JIST 2011: 284-299 [Scharffe et al. AAAI 2012] Franois Scharffe, Ghislain Atemezing, Raphal Troncy, Fabien Gandon, Serena Villata, Bndicte Bucher, Fayal Hamdi et al. Enabling linked-data publication with the datalift platform. In Proc. AAAI workshop on semantic cities. 2012. [Todorov, Hudelot, Popescu, Geibel. IJUFKS 2014 (in print)] Konstantin Todorov, Celine Hudelot, Adrian Popescu, Peter Geibel. Fuzzy Ontology Alignment Using Background Knowledge. Intl. Journal on Uncertainty, Fuzziness and Knowledge-Based Systems. 2014. Z. Bellahsene, A. Laurent, F . Scharffe, K. Todorov 28 / 29

Thank you for listening! Z. Bellahsene, A. Laurent, F . Scharffe, K. Todorov 29 / 29