The Trouble with TDM
Download as PPTX, PDF0 likes71 views
The document discusses the difficulties of text and data mining (TDM) for a research project between the British Library and several universities. It summarizes that: 1) While the UK copyright law contains a TDM exception, it was not drafted with research partnerships in mind and does not allow the British Library to share digitized content with research partners. 2) Practical difficulties in assessing and obtaining permissions for the huge amount of copyrighted content were addressed through high-level content analysis, spot checks, and managing risks. 3) The British Library addressed legal difficulties by having partners temporarily join the library and access digitized content on a secure server, then deleting the content after use.
The Trouble with TDM
- 1. The Trouble with TDM* … (*Text & Data Mining) Matthew Lambert
- 2. 2 bl.uk Matthew Lambert • Head of Copyright Policy & Assurance The British Library • National Library • 150-200M items • Legal deposit Library Introductions
- 3. 3 bl.uk The project – Living with Machines • Living with Machines is research project between The British Library, The Turing Institute, Universities of Cambridge, East Anglia, Exeter and London (QMUL) • Funded by the UK Research and Innovation (UKRI) Strategic Priority Fund • Focussing on the period 1780-1918 the project uses computational analytical tools to examine the ways in which technology impacted society in the UK
- 4. 4 bl.uk British Library’s Contribution • The British Library holds a vast collection of content, including newspapers and maps detailing changes in the period being looked at. • We provided over 100 newspaper titles and digitised more than 600,000 pages for the project.
- 5. 5 bl.uk One quick caveat… This is only the copyright aspect – there are many other difficulties associated with a project of this size
- 6. 6 bl.uk Difficulties Legal: • The TDM exception not drafted with research partnerships in mind • No other useful exceptions • Need to be able to make sample datasets available following research Practical: • Assessing the material at scale • Tracking down rights holders • Managing content while enabling analysis
- 7. 7 bl.uk The law – CDPA 1988 – s29A The UK’s Copyright legislation does contain an exception covering Text and Data Mining • It is fantastic that we have the exception, however it was not written with partnerships in mind Let’s take a closer look…
- 8. 8 bl.uk The law – CDPA 1988 – s29A • Great! The British Library holds physical copies of the items and so has lawful access to them • The project is non-commercial, and will provide acknowledgement • Content is not being sold or ‘let for hired’
- 9. 9 bl.uk The law – CDPA 1988 – s29A So while the British Library can make the copies we can not give it to anyone else… …where does this leave our partners who are doing the research?
- 10. 10 bl.uk The law – CDPA 1988 – s29A • So on the face of it the exception isn’t all that helpful... • Still all is not lost • It does allow us to digitise the material • It does also allow the BL to use the content (more on that later) • We will just need to use other approaches to fill the gaps • Assessing content • Licensing material • Managing risk • In addition to the initial legal difficulties, research of this type is expected to publish datasets so research can be checked
- 11. 11 bl.uk Practical Difficulties • There is a vast amount of content the project wishes to use, hundreds of thousands of pages • Some of the content will be out of copyright and so free to use, but which? • Often the content was published by organizations which no longer exist • Is the content anonymous? Article by-lines are relatively recent in newspapers
- 12. 12 bl.uk Practical Difficulties • If we decide to go down a licensing route who do we speak to? • Given all the unknowns there will inevitably be risks, so that needs to be managed. • How do we get the content to our partners?
- 13. 13 bl.uk Approaches – Practical Difficulties • As each article and image in a newspaper is a distinct copyright work they potentially need to be individually looked at – very resource intensive • Assessing content at such scale has significant difficulties, so had to take a high level approach • Content was looked at in detail to determine safe dates to help with assessment • Spot checks to find out if / when by-lines became common in newspapers • Risk appetite for the project determined and processes built to manage them
- 14. 14 bl.uk Approaches – Legal Difficulties • Once copyright status was determined we needed to work out how to deal with the in copyright material • The British Library can’t give the content to our partners and still remain within the exception • But what if the partners temporarily joined the British Library? Then they would have lawful access and work would fall under the exception… • Digitised content is held on a dedicated, secure server which the BL controls, researchers have limited access to the content they need
- 15. 15 bl.uk • When content no longer needed, it is deleted from the secure server • Out of copyright datasets will be made available to test research results • Provided guidelines to researchers regarding publication of their results to ensure copyright was not infringed. Approaches – Legal Difficulties
- 16. 16 bl.uk • After Brexit the DSM - will not change UK legislation (so articles 3 and 4 no help). • Recent consultation by the UK IPO on Copyright and AI did specifically cover TDM. • For UK research to remain globally competitive we need this exception to be widened. The Future