Elsevier has recently announced a new policy on text and data mining, or TDM. The implementation of the policy requires that libraries sign an agreement with Elsevier that will permit their patrons to use TDM tools on Elsevier’s content. Many libraries have questions on the implications of this policy – and on TDM in general. Below we outline the essential issues and offer some guidance on how libraries might consider responding to this new policy.
What is Text and Data Mining (TDM)?
TDM is a research process where computing tools are applied across multiple research articles to analyze their content and create new knowledge by combining information gathered from them. They look for facts, entities and relationships within the text, and extract that information for analysis. Importantly, they can analyze information across a broad range of fields and across many articles – hundreds of thousands if necessary – taking analysis to a different level than that which the human brain can manage. Although still in its relative infancy, these technologies hold great promise for the future and are already proving their worth in fields such as pharmaceutical, biomedical, and chemical research.
Some key points to keep in mind about TDM:
- Researchers carrying out TDM need to mine the contents of large numbers of articles.
- These articles may be in multiple fields of research.
- These articles may be published in many different journals and by many different publishers.
- These articles may exist on many different platforms.
What has the publishers’ general position been on enabling TDM?
Publisher’s positions vary, but in general, publishers consider that material licensed to libraries is made available for readers to download and read, but not to text-mine. In some cases, researchers who have tried to use TDM tools to crawl large numbers of articles from a single publisher have found that their IP address has been blocked; since the IP address used is usually that of their institution, the repercussions can be serious.
A number of publishers are developing their own, proprietary text-mining tools that they intend to provide to researchers. However, researchers who wish to carry out text-mining generally prefer to develop their own tools for the job (some of which are necessarily rather specialized) and would prefer to use those. It is not clear why researchers would expect publishers to provide research tools for them, though it is easier to see why publishers may wish researchers to use their proprietary tools.
In addition, some publishers have indicated that they are considering charging users to text-mine their content if a commercial product were to result. Those publishers presumably fear that what could be created by such activities (new databases) may damage their own business prospects.
What has the general position of libraries been on TDM?
Libraries generally consider that when they have licensed access to a publisher’s content, it should be available for their patrons to mine as well as to read. LIBER has published a useful and informative factsheet on TDM that sets out the case from the library community’s position.
What has Elsevier’s position been until now on TDM?
Until now, Elsevier has required researchers who wish to mine its content to seek specific permission from the company. This has been granted on a researcher-by-researcher, case-by-case basis. This solution may have helped individual researchers, but is not a sustainable one in the face of the growth in text-mining.
What is Elsevier’s new position on TDM?
Elsevier’s new policy enshrines TDM rights in the standard ScienceDirect subscription agreement for academic libraries. It applies to all new subscription agreements and can be added to existing agreements on request.
Researchers at a subscribing institution wishing to undertake TDM must register with Elsevier’s developer’s portal, upon which they will receive a key to access the ScienceDirect API for text and data mining. Outputs must be licensed using the Creative Commons CC-BY-NC licence (i.e. the product of the text-mining exercise cannot be used for commercial purposes). Researchers must also add the DOI of the original sources so that authors may get credit and there is a verifiable provenance to the data. The policy allows researchers to re-publish ‘snippets’ from the original papers, but these must not exceed 200 characters in length.
Are their research-based issues with Elsevier’s licensing scheme that libraries should be aware of?
Yes. For researchers wishing to undertake text-mining, licence-based solutions like Elsevier’s have the following shortcomings:
- Researchers wishing to mine across large collections of articles (e.g. from much of their library’s collection) find having to sign separate agreements with different publishers a barrier to their work. Emerging initiatives (such as Prospect, described below) may provide some progress towards resolving this administrative burden, but are not yet widely used.
- Research teams collaborating across multiple institutions where mining will take place in each institution will require all the individual libraries involved to be signed up to Elsevier’s licence terms.
- Researchers do not always wish to work through an API, preferring to mine the articles themselves. There are new technologies that enable mining even the PDFs of articles, a great step forward technically, but licences such as Elsevier’s will not allow this: researchers must use only ScienceDirect’s API.
- Elsevier’s TDM license only applies to text. It does not include images in the content that is minable. Researchers wishing to mine images must request specific permission for this and Elsevier will provide an image retrieval API in return. Elsevier’s reason for examining each case individually is that they note that there is some ambiguity about re-use rights for some images. This presumably implies that Elsevier will want to know the purpose for which the researcher will put the resulting mined data: some researchers may not wish to divulge this.
- Researchers must use a non-commercial (NC) licence when publishing the resulting information from the text-mining exercise. This may not suit all users purposes.
What else is going on regarding publisher permissions to text-mine?
A number of publishers have agreed to work towards a solution for researchers who wish to mine across a body of articles published by multiple companies. There is a developing initiative from CrossRef (called Prospect) that seeks to address this by providing a common API and a licence framework whereby researchers can agree to terms and conditions from multiple publishers through one portal.
Some governments are taking an interest in this issue. The UK Government intends to act in Spring 2014 on the recommendation in the report by the UK’s Intellectual Property Office to create an exemption from copyright law for text-mining for non-commercial purposes. The European Commission is also exercised by this issue. In 2013, it conducted a ‘stakeholder dialogue’ exercise on its Licences for Europe initiative, which proposed to have a specific European licence for TDM. The organizations representing the research community, led by library organizations, pulled out of the process in May, arguing against a licence-based solution, and for an exemption from copyright for TDM. The exercise has now concluded and it is generally agreed that in its intended form it has no future. The Commission has now convened a group to examine the economic impact of TDM in scientific research and the barriers to its use. The conclusions are expected at the end of February 2014.
So, should I sign a licence with Elsevier?
If you are about to renew or renegotiate your deal with Elsevier, the licence to text-mine will be part of the terms and conditions. You may of course choose to negotiate to omit that clause. If your deal still has some time to run, you can have the licence for your patrons to text-mine included straight away by a separate agreement.
Should you decide not to sign, your patrons will be unable to mine Elsevier’s content for the foreseeable future. If you do sign, your patrons will be able to undertake text-mining under Elsevier’s terms. This may not suit all of them, so a consultation with appropriate researchers is in order before the licence is signed. The issues are laid out here and should be fully explained to researchers so that they can take an informed position on this.
Other reading:
Trouble at the text mine. Nature News & Comment, 8 March 2012: http://www.nature.com/news/trouble-at-the-text-mine-1.10184
Elsevier opens its papers to text-mining. Nature News & Comment, 3 February 2014: http://www.nature.com/news/elsevier-opens-its-papers-to-text-mining-1.14659
