Cover Image

PAPERBACK
$48.00



View/Hide Left Panel

that you make a representation of your thoughts through language. These are the bases with which I would like to start.

Currently, the scientific community is storing data in repositories. We link to publications and vice-versa. The example that is commonly used is that people add PDFs and spreadsheets to their papers. This is pretty useless because we are not doing anything with these documents. Having them does not mean we can find the dataset.

In general, I believe that datasets should all be available for server search and that sets and subsets of that data should be made freely accessible, whenever possible. Overall, commercial publishers are not interested in owning or charging for research data or running those repositories. There might be exceptions, but in general, this is the case.

In my view, most publishers are very interested in working with data repositories and believe that it would be very useful if there were one place where we can find data items. It would be useful if an identifier is persistent and unique and that if the content changes, the identifier changes as well. Also, it would be very useful if the data would link back to the publication. It would be more interesting if we have data in a repository and can link them to some content from within a publication. Not only from the top level, but from within the publication. There are some examples of this. What my lab has been doing currently is tagging entities and linking them to databases. This involves some manual as well as some automated work.

More interesting, I think, is the fact that we can now create claim evidence networks that span across documents, so we can have a statement that can be backed up in a table or a reference in another publication or in another data center. At least at Elsevier, we are very invested in the idea of linked data. We have developed something that we call a satellite, which is essentially a way to describe a Linked Data annotation, in RDF. We are using Dublin Core and SWAN’s provenance and authoring/version ontology to identify the provenance.

We are very happy to develop this with people like Paul Groth and Herbert van de Sompel and others to have an ontology that connects to their work. The idea is that we can have some files that link to our XML at any level of granularity. There are files that sit outside the publication or the data center but we can still link one to the other. I think this is a very promising way to move forward.

What would be really interesting is if we had the opportunity to completely re-think science publishing. Why only change where the data is located: why not change the whole process? In my opinion, what is key is that scientists should be allowed to do their research process the way they want. We do not want to put more obstacles in front of the busy scientists who are already struggling to do their work. In fact, I think that the publishers would like to help them. So, if they have an experimental design, perhaps they can put a copy of it in the repository and put a link to it in their paper. Similarly, there are reports of observations. Perhaps there can be some way to deposit these reports in a repository and to pull them into their paper, code their statistics in a same way, and then draw the conclusions.

For the publisher and probably for the reader, it is incredibly important to maintain the context that the data have (e.g., the experimental context, the reason you did the experiment, the time



The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement