Despite many e orts for making data about scholarly publications available on the Web of Data, lots of information about academic conferences is still contained in (at best) free-text format. When available in a structured format, these data would provide an essential input for the decisions researchers, libraries, publishers, funding and evaluation bodies take every day. In this paper we present a vision for having such data available as Linked Open Data (LOD), and we argue that this is only possible { and for the mutual bene t { in cooperation between researchers and publishers. We also present a pilot project aimed at publishing data about 8,500 computer science conferences as LOD.
{ Shall I submit a paper to this conference? How good and relevant is it? What are the alternatives? (younger researcher) { Shall I accept a PC membership invitation, give an invited talk, send a workshop proposal to this conference? (more senior researcher) { Shall we publish the proceedings of this conference? (publisher) { Is it worth sponsoring this conference? (sponsor) What do you do when you face these questions? You google, read many documents and webpages, ask people, and you are never sure whether you have found all relevant data and numbers.
Data about conferences are spread across several sources in a largely chaotic
and non-structured way, being duplicated multiple times. Let us take the
example of the PC (Program Committee) membership: being involved in paper
reviewing and other activities related to a conference organization is hard work
that should be credited [
The key issues to address here are data exchange between various systems
involved in conference organizations and the lack of trusted sustainable5
largescale data sources providing detailed conference data. Currently, the LOD cloud
includes several resources that contain conference metadata. Semantic Web
Conference Corpus6 includes information on major Semantic Web conferences (37)
and workshops (235), therefore, providing high quality but low coverage data.
Another example is COLINDA [
In the following, we show how the reliability, coverage and sustainability of such data can be improved by cooperation between publishers and researchers.
5 Not many resources and tools outlive the research projects they originate from. 6 http://data.semanticweb.org/ 7 http://www.colinda.org/
The issues outlined above motivated the launch of the Springer LOD pilot, which aims at publishing data about Computer Science conferences as a linked open dataset. The availability of such a dataset will contribute to the broader goals of publishing the scholarly data as LOD: { accessible science: data about publications, authors, topics, and conferences should be easy to explore; { transparent science: the data on productivity and impact of authors, research institutions, and conferences should be open and easy to analyze. But these goal are only marginally relevant for publishers, whose primary goal is, not surprisingly, commercial bene ts. So, how do the interests of publishers and researches align?
Publishing conference data as LOD would allow Springer to enrich bibliographic data provided via data services to libraries, data agencies and aggregators. This would also allow linking to other data, thus increasing the visibility of the proceedings in SpringerLink digital library. This would provide bene ts for conference community (i.e. researchers): more readers, more downloads, more citations, conference submissions and participants. Moreover, Springer sees this as a way of collaborating with the research community and other stakeholders (libraries, indexing services, conference-related systems) to get new insights on the data. Also, the data would allow detecting trends in the conference business, and plan accordingly: knowing that many conferences go to Russia or China, publishers need to establish agreements with local printers, take into account customs regulations, etc. As with any LOD resource, sustainability is crucial: and in our opinion, it is directly related to the economic value the data brings. Moreover, the bene ts of boosting the content usage and discoverability, and data enrichment via linking, outweigh potential pro ts from selling these data.
The pilot has started in 2013 and is ongoing. In the conference dataset that will be made available as a result of the pilot (later in 2014), for each conference the following information is provided: conference series name and ID; conference ID, acronym, and number in the series; city country, start and end dates. See Figure 1 for an internal XML representation. The starting point are the conference data that are present in the subtitles of the proceedings, i.e. in a freetext format: e.g. \12th International Semantic Web Conference, Sydney, NSW, Australia, October 21-25, 2013, Proceedings, Part I". In the pilot, the Springer internal conference data management system was extended with a module that extracts and structures this information from the subtitles. Then, the quality of the extracted data is manually assured with the help of an interactive GUI, following the same philosophy for the data quality standards as the one of DBLP. The resulting conference data are stored in a database, which makes their conversion to RDF straightforward. As the example shows, the data contains some elds, e.g., conference acronym, number, city, country, link to the proceedings, which are either not available in COLINDA or the Semantic Web Conference Corpus or available there in free text form.
Currently, the data for 8,500 conferences (which correspond to around 2,000 conference series) published in LNCS, LNAI, LNBI, LNBIP, CCIS, IFIP-AICT and LNICST series since 1973 was processed following the above procedure. As every year 650 new conferences are published in these series, the information about them will be added to the system, structured and exported into RDF. The data are curated at publisher's end, using well-established (over the course of the last 40 years) processes: the process of producing metadata for SpringerLink was augmented with an additional step, during which the conference metadata is extracted and its quality is assured. Services such as scholarly search engines will be able to use the conference data directly from SpringerLink. The very same conference metadata will be then published as LOD, in RDF format. Such separation of data from formats allows for adding third party LOD conference data (for conferences not published by Springer) in the future.
In the future we plan to provide richer metadata that includes the number of submitted and accepted papers, acceptance rates, information on the best paper awards, PC and chairs, co-located workshops, links to CORE rankings8, etc. Moreover, in the future the data would go beyond the computer science scope, extending to approximately 350 conferences published annually by Springer in other disciplines.
According to the internal Springer statistics, the 8,500 conferences contain almost 300,000 articles published in the proceedings, and slightly over 300,000 distinct authors contributing to the papers. Making publication and author metadata available is not the focus of the current stage of the pilot, but such information can be provided in the future by linking to other datasets, such as DBLP. Linking to citation gures (e.g. from CrossRef9) and ORCIDs will further enrich the data.
The result of this initial data publishing stage is a well-structured carefully maintained conference dataset, which can be interlinked with other datasets (DBLP or national libraries' data, GeoNames and DBpedia for locations, etc.) and used in applications. However, the initial data publishing stage will hardly go any further unless both researchers and publishers actively participate in providing more data, linking them, and developing new applications supporting the questions we posed in the introduction.
One example of application is based on the Rexplore [
Another application is using the conference data during the conference lifecycle. Once entered in a CMS, the data about PC membership could be exported10 to become part of LOD cloud and then displayed on the website in one of n standard ways (e.g. using speci c plugins), or be included in the preface of the proceedings, various conference apps, etc. Such coordination between researchers and publishers would prevent data duplication and enable data reuse. 4
This work has been supported by the LOD2 and DM2E EU FP7 projects. We thank Max Schmachtenberg, University of Mannheim, for providing the metadata on the scholarly domain in LOD, and Markus Richter for developing the Springer conference data management tools.