Publishing and Pushing Linked Open Data

Embed Size (px)

DESCRIPTION

This presentation outlines the need to invest intellectual and expert human effort in data publication in order to see compelling research outcomes. I gave this presentation on April 10th, 2014 at the University of Pennsylvania in an event sponsored by the Penn Humanities forum (http://humanities.sas.upenn.edu/13-14/dhf_opendata.shtml)

Text of Publishing and Pushing Linked Open Data

  • 1. Publishing and Pushing Linked Data in Archaeology Unless otherwise indicated, this work is licensed under a Creative Commons Attribution 3.0 License Eric C. Kansa (@ekansa) UC Berkeley D-Lab & Open Context

2. Introduction Challenges in Reusing Data 1. Background 2. Data publishing workflow 3. Data curation and dynamism 3. Gold Standard of professional contribution 4. My Precious Data: Dysfunctional incentives (poorly constructed metrics), limit scope, diversity of publications Image Credit: Lord of the Rings (2003, New Line), All Rights Reserved Copyright 5. Need more carrots! 1. Citation, credit, intellectually valued 2. Research outcomes (new insights from data reuse!) 6. Need more carrots! 1. Citation, credit, intellectually valued 2. Research outcomes (new insights from data reuse!) Why linked data is so important 7. EOL Computable Data Challenge (Ben Arbuckle, Sarah W. Kansa, Eric Kansa) 8. Large scale data sharing & integration for exploring the origins of farming. Funded by EOL / NEH 9. 1. 300,000 bone specimens 2. Complex: dozens, up to 110 descriptive fields 3. 34 contributors from 15 archaeological sites 4. More than 4 person years of effort to create the data ! 10. Relatively collaborative bunch, Ben Arbuckle cultivated relationships & built trust over years prior to EOL funding. 11. Introduction Challenges in Reusing Data 1. Background 2. Data publishing workflow 3. Data curation and dynamism 12. 1. Referenced by US National Science Foundation and National Endowment for the Humanities for Data Management 2. Data sharing as publishing metaphor 13. Raw Data: Idiosyncratic, sometimes highly coded, often inconsistent 14. Raw Data Can Be Unappetizing 15. Publishing Workflow Improve / Enhance 1. Consistency 2. Context (intelligibility) 16. Sometimes data is better served cooked 17. - Documentation - Review, editing - Annotation 18. - Documentation - Review, editing - Annotation 19. - Documentation - Review, editing - Annotation 20. - Documentation - Review, editing - Annotation 21. - Documentation - Review, editing - Annotation 22. Ovis orientalis Code: 14 Wild sheep Code: 70 Code: 16 Ovis orientalis Code: 15 Sheep, wild O. orientalis Sheep (wild) 23. - Documentation - Review, editing - Annotation 24. Ovis orientalis http://eol.org/pages/311906/ Code: 14 Wild sheep Code: 70 Code: 16 Ovis orientalis Code: 15 Sheep, wild O. orientalis Sheep (wild) 25. Controlled vocabulary Linked Data applications 26. Sheep/goat http://eol.org/pages/32609438/ 1. Needed to mint new concepts like sheep/goat 2. Vocabularies need to be responsive for multidisciplinary applications 27. Linking to UBERON 1. Needed a controlled vocabulary for bone anatomy 2. Better data modeling than common in zooarchaeology, adds quality. 28. Linking to UBERON 1. Models links between anatomy, developmental biology, and genetics 2. Unexpected links between the Humanities and Bioinformatics! 29. 7000 BC (many pigs, cattle) 7500 BC (sheep + goat dominate, few pigs, few cattle) 6500 BC (few pigs, mixing with wild animals?) 8000 BC (cattle, pigs, sheep + goats) Not a neat model of progress to adopt a more productive economy. Very different, sometimes piecemeal adoption in different regions. Separate coastal and inland routes for the spread of domestic animals, over a 1000-year time period. 30. Easy to Align 1. Animal taxonomy 2. Bone anatomy 3. Sex determinations 4. Side of the animal 5. Fusion (bone growth, up to a point) 31. Hard to Align (poor modeling, recording) 1. Tooth wear (age) 2. Fusion data 3. Measurements Despite common research methods!! 32. Professional expectations for data reuse 1. Need better data modeling (than feasible with, cough, Excel) 2. Data validation, normalization 3. Requires training & incentives for researchers to care more about quality of their data! 33. Nobody expected their data to see wider scrutiny either.. 34. and not just academic researchers, linked open data involves many sectors! 35. Digital Index of North American Archaeology (DINAA) 1. State site files created to comply with federal preservation laws 2. Main record of human occupation in North America 3. PIs: David G. Anderson and Josh Wells 36. DINAA 1. Stable URI for each site file. 2. CC-Zero (public domain) 3. Beginning to link to controlled vocabularies 37. Data are challenging! 1. Decoding takes 10x longer 2. Data management plans should also cover data modeling, quality control (esp. validation) 3. More work needed modeling research methods (esp. sampling) 4. Editing, annotation requires lots of back-and-forth with data authors 5. Data need investment to be useful! 38. Introduction Challenges in Reusing Data 1. Background 2. Data publishing workflow 3. Data curation and dynamism 39. Investing in Data is a Continual Need 1. Data and code co-evolve. New visualizations, analysis may reveal unseen problems in data. 2. Data and metadata change routinely (revised stratigraphy requires ongoing updates to data in this analysis) 3. Problems, interpretive issues in data (and annotations) keep cropping up. 4. Is publishing a bad metaphor implying a static product? 40. Data sharing as publication Data sharing as open source release cycles? 41. Data sharing as publication Data sharing as open source release cycles? 42. Data sharing as publication AND Data sharing as open source release cycles 43. Data are challenging! 1. Decoding takes 10x longer 2. Data management plans should also cover data modeling, quality control (esp. validation) 3. More work needed modeling research methods (esp. sampling) 4. Editing, annotation requires lots of back-and-forth with data authors 5. Data need investment to be useful! 44. Image Credit: Brainchildvn via Flickr (CC-By) http://www.flickr.com/photos/brainchildvn/3957949195 45. Image Credit: Brainchildvn via Flickr (CC-By) http://www.flickr.com/photos/brainchildvn/3957949195 Not an easy environment to seek new investments. 46. Contingent Employment Source: Washington Monthly (http://ecleader.org/2012/02/21/nation-wide-trend-towards- adjuncts-threatens-higher-ed/) 47. Bethany Nowviskie (University of Virginia) Shifts in Career Paths and Professions (#alt-academy), different publishing incentives, emerging as data assume a greater emphasis 48. Bethany Nowviskie (University of Virginia) Alt-Acs (contingent, low status) not a good answer, but reflect wider need for institutional reform. 49. One does not simply walk into Mordor Academia and share usable data Image Credit: Copyright Newline Cinema 50. Final Thoughts Data require intellectual investment, methodological and theoretical innovation. Institutional structures poorly configured to support data powered research New professional roles needed, but who will pay for it? 51. Thank you! University of Pennsylvania Digital Humanities Forum and other Sponsors!