At it’s recent annual meeting in Baltimore, the Linguistic Society of America (LSA) passed a resolution on data sharing that is the result of a series of discussions that took place last year, for example at the meeting of the Cyberlinguistics group in Berkeley last June.
Here’s the text (snip):
Whereas modern computing technology has the potential of advancing linguistic science by enabling linguists to work with datasets at a scale previously unimaginable; and
Whereas this will only be possible if such data are made available and standards ensuring interoperability are followed; and
Whereas data collected, curated, and annotated by linguists forms the empirical base of our field; and
Whereas working with linguistic data requires computational tools supporting analysis and collaboration in the field, including standards, analysis tools, and portals that bring together linguistic data and tools to analyze them,
Therefore, be it resolved at the annual business meeting on 8 January 2010 that the Linguistic Society of America encourages members and other working linguists to:
make the full data sets behind publications available, subject to all relevant ethical and legal concerns; annotate data and provide metadata according to current standards and best practices; seek wherever possible institutional review board human subjects approval that allows full recordings and transcripts to be made available for other research; contribute to the development of computational tools which support the analysis of linguistic data; work towards assigning academic credit for the creation and maintenance of linguistic databases and computational tools; and when serving as reviewers, expect full data sets to be published (again subject to legal and ethical considerations) and expect claims to be tested against relevant publicly available datasets.
I think it’s great that the LSA is throwing its weight behind this effort and supporting the idea of data sharing. The only minor complaint that I have concerns the wording – what exactly does make available mean? It could mean real Open Access, but also that you’ll email me your datasets if I ask nicely. Or it could mean that a publisher will make your datasets available for a fee – any of these approaches qualify as making data available in this terminology.
So, while I think this is good starting point, more discussion is needed. Especially when it comes to formats, means of access and licensing we need to be more explicit.
Imagine this scenario for a moment: you want to compare the semantic prosody of the verb cause across a dozen languages. If data sharing (and beyond that, resource sharing) were already a reality, we could do something like this:
1. Send a query to WordNetAPI* to identify the closest synonyms of cause in the target languages.
2. Send a query to UnversalCorpusAPI* using the terms we have just identified and specifying a list of megacorpora that we want to search in.
3. Retrieve the result in TEI-XML.
4. Analyze the results in R using the XML package.
The decisive advantage here would be that I only get the data I need, not everything else that’s in those megacorpora that is unrelated to my query. Things just need to be in XML and openly available and I can continue to process them in other ways. This would not just be sharing, but embedding your data in an infrastructure that makes it usable as part of a service. And that would be neat because what good is the data really if it doesn’t come with the tools needed to analyze it? And in 2010 tools=services, not locally installed software.
Now that would be awesome.
(*) fictional at this point, but technically quite feasible.
