Frank van Harmelen’s tweet drew my attention on a paper of Jan Hannemann and Jürgen Kett on Linked Data for Libraries. I hope Jan and Jürgen will not be upset if I copy some quotes from their paper, but I thought that giving more publicity to some of their experiences in deploying linked data at the German National Library is worthwhile. Reproduced here without change though somewhat shortened:
- Setting up a service is not trivial. […] the essential software solutions (tools) involved have not reached full maturity yet. […] documentation may be lacking the required depth. […] multiple software components need to be setup to work together […] which requires appropriate expertise.[…]
- Data modeling can be complex. When publishing data on the web, it is advantageous to use existing, registered ontologies. Unfortunately, these ontologies do not always match the data representation of each individual library […] the definitions of individual properties can vary considerably. […] There is no simple answer to the question which is the right thing to do.[…]
- Open data exchange mentality does not exist everywhere. Even before linked data, libraries have exchanged and aligned their data sets. The results of such projects could be prime information sources for connecting linked data sets. Sadly, not all institutions involved share the open exchange mentality, and shared ownership may make it difficult to publish these results.
- Best practices are seen as rules. Linked open data is based largely on best practices rather than rules. However, this pragmatic aspect is not seen as essential in all areas of the linked data community. Deviations from perceived standards tend to be criticized, which can cause institutions new to the semantic web to doubt their decisions – even if they make sense for the organization in question. Libraries should not be deterred by such feedback and rather see this as a motivation to contribute their own experiences and knowledge to the community. Guidelines and best practices should be carefully considered in the context of each institution’s needs, especially in this early forming phase of the semantic cloud.[…]
- Properly modeled data is very useful. Once the data modeling is completed and the data made available, it can be used by others. A colleague at the Technical University of Braunschweig has shown that with properly modeled data, this can result in very useful applications: within a day, he imported our data into a database, added a web interface and had thus created a searchable access to our data.