The Pragmatics of Ontology and Heterogeneous Data Sources The Ins and Outs of CTSAsearch David Eichmann School of Library and Information Science University of Iowa
Research Networking Programmatic support for discovery and use of research and scholarly information regarding people and resources. They are essentially special purpose institutional knowledge management systems.
Representative RN Systems Profiles (Harvard) VIVO (VIVO Consortium) Loki (Iowa) SciVal Experts (aka Pure – Elsevier) A number of others
Why Bother with VIVO (the ontology)? Words in a profile are just sequences of characters carrying no meaning –Try asking Google Scholar what grant funded a given hit… With structure and relationship comes meaning, aka semantics –Enter the Semantic Web!
Connecting the Dots The real challenge here is translation of information already in existence in scattered sources –Research networking tools –Citation databases (e.g., PubMED) –Award databases (e.g., NIH Reporter) –Curated archives (e.g., GenBank) –Locked up in text (the research literature)
CTSAsearch – version 1 10 SPARQL endpoints 19 institutions 124,945 individuals Proved challenging for some sites to handle the queries
CTSAsearch – version 1 subclass | count NonFacultyAcademic | FacultyMember | NonAcademic | EmeritusFaculty | 2134 EmeritusProfessor | 2070 Postdoc | 1226 Librarian | 232 Student | 89 GraduateStudent | 71
CTSAsearch – version 2 10 SPARQL endpoints (19 institutions) 15 VIVO sites –Harvested with customized crawler 14 Profile sites –Harvested with customized crawler
CTSAsearch – version 2 subclass | count NonFacultyAcademic | FacultyMember | NonAcademic | Student | GraduateStudent | EmeritusFaculty | 3096 EmeritusProfessor | 2072 Postdoc | 1410 Librarian | 264
CTSAsearch – architecture 1 VIVO-based SPARQL harvester 2(!) VIVO-based crawlers 1 Profiles-based crawler 2 Platform-specific HTML crawlers 1 CSV-based loader
CTSAsearch – architecture
CTSAsearch – current 45,456,417 VIVO-derived triples 48,569,115 Profiles-derived triples
Recent Work Cross-linkage across sites –Resolving ‘stubs’ –Formation of a single ecosystem Macro concerns –Institution-scale analytics –Pondering reflection
Current “profile”
CTSAsearch/Polyglot – version x Temporary SPARQL endpoint: – Shared visualization widgets –Intended for embedding in institutional sites Community-wide sameAs assertions
Pattuelli’s Spectrum of Relationships (2012)
Pattuelli’s Spectrum of Relationships (2012) RN Tools
Pattuelli’s Spectrum of Relationships (2012) RN Tools Linked In
Pattuelli’s Spectrum of Relationships (2012) Ontologies used –foaf (Friend of a Friend) –rel (Relationship) –mo (Music) Echos of Trigg’s link taxonomy –Trigg, R Network-Based Approach to Text Handling for the Online Scientific Community. Ph.D. dissertation, Department of Computer Science, University of Maryland, technical report TR-1346
Connecting the Dots – Take 2 Figure courtesy of Melissa Haendel, OHSU
PubMed Central Open Access 886,172 papers (as of 1/1/15) 423,764 with acknowledgements 994,931 sentences 4,329,972 parses
The Simple Cases PMCID: SeqNum: 2 SentNum: 6 Sentence: EK analysed the data. POS: [EK/NNP, analysed/VBD, the/DT, data/NNS,./.] Parse: [S [NP EK/NNP ] [VP analysed/VBD [NP the/DT data/NNS ] ]./. ]
And the Not So Simple… PMCID: Sentence: We thank Sheila Harvey, Clinical Trials Unit Manager at ICNARC, and Ruth Canter, Trials Administrator at ICNARC, for their assistance in chasing completed surveys; Dr Kevin Gunning for early advice and project development; Drs Neill K. J. Adhikari and Gordon D. Rubenfeld for feedback and discussion of analysis plan; Dr Chris AKY Chong for his valuable comments on the initial draft of this manuscript; and our Responders: Addenbrooke’s Hospital ( Dr Kevin Gunning ), Airedale General Hospital ( Dr John Scriven ), Alexandra Hospital ( Dr Tracey Leach ), Arrowe Park Hospital ( Dr Lawrence Wilson ), Barnet Hospital ( Dr AH Wolff ), … 8,245 character long sentence
Extract Entities/Relationships with Syntactic Queries [S [NP:Author NN:Author ] [VP NN [NP:Person ] [PP ], [PP ] ] ] S <1NP:Author <2[VP <1/thank/ <2(NP) <3(PP) ] –For the sentence having this pattern, match the object noun phrase and the next prepositional phrase NP <#2 <1(NNP) <2(NNP) –For the noun phrase, extract two proper nouns PP <#2 <1DT <2(NP) –For the prepositional phrase, match the noun phrase
Person Results Snippet IDTitleFirst NameMiddle NameLast Name 76HansMatrin 77JeffVieira 78P.ZAMORE 79Prof.EricSchon 80CarlosLois 81AndreaMöll 82ElenaGovorkova 83K.M.Pollard 84Dr.MichaelBerton
Relationships for Person 77 PMCIDCategoryPP Supportthe kind gift of rKSHV Supportthe kind gift of rKSHV.219 and for helpful discussions Collaborationhelpful discussions
Relationships for Person 79 PMCIDCategoryPP Resourcethe rabbit polyclonal antibody Resourcethe ECFP and EYFP plasmids Collaborationhis helpful advice and discussions
Category Frequencies CategoryCount Collaboration47,052 46,327 Technique33,598 Resource8,894 Support6,836 Event3,744 Project854 Place Name229 Publication Component 210 Place186 Organization93
Next Steps Continue slogging through extraction pattern definition Define patterns for –funding declarations –chairs, fellowships, etc. Merge data into CTSAsearch visualizations Align current category scheme with Melissa Haendel’s current draft ontology for CASRAI taxonomy and then merge with VIVO-ISF
In the Next Year Joint work with Melissa Haendel (OHSU) on administrative supplement to OHSU’s CTSA bridging RNs and NIH’s SciENcv –Map SciENcv data model to VIVO-ISF –Enable bi-directional data exchange –Integrate clinical/trial data sources –Integrate SciENcv, ORCID data into CTSAsearch –Multi-granularity search and visualization
Questions?