Sent by José Borbinha <email@example.com> to the FRBR mailing list today.)
This is an invitation to take part in an evaluation of an experiment in the TELplus project (http://www.theeuropeanlibrary.org/telplus)
Please evaluate the system described below and fill the questionnaire, according to the instructions.
The questionnaire is anonymous, and we estimate it can be answered in less than 5 minutes. However, it has to be preceded by a “test drive” with the system, which can take as long as you want…
We appreciate the answers to the questionnaire until the end of the day of December 16 (Wednesday), please!
The aim of this task is to explore possible improvements for search in The European Library (http://www.theeuropeanlibrary.org), under the assumption that a presentation of the results, following the general ideas of the FRBR model, could help.
Considering that The European Library amounts to about 150 million records in 32 different languages, this might have the potential to narrow down the search results by clustering similar records (including of multiple languages).
In order to evaluate that, we developed a case with a sample collection of the works of the Literature Nobel Prize winners. For that, we used records extracted from 11 libraries, thus with a high number of works with many translations in multiple languages. We are now inviting you to evaluate this system and answer to the questionnaire available at:
Description of the system
The system to experiment is available here:
The system was prepared like this:
1- All the records were received in MARC21 or UNIMARC
2- All the records were transformed to the “TEL Application Profile” and indexed by a Search Engine, just like The European Library is doing nowadays
3- In parallel, all the records were also processed to build a new “FRBR database”, built according to a set of “FRBR motivated” rules that try to pre-build clusters according to what is perceived to be the same work.
The way the system works is like this:
1- Each search is first done by the Search Engine! But instead of presenting the results immediately…
2- …they are first sent to the FRBR database, which returns the same results but now grouped according to what this database already knows from the initial work’s clustering (this might comprise more records than those initially found by the search engine, and also might include enhancements in author’s names and dates). These are the results that finally are shown to the user!
For easy evaluation of the prototype, and answering the questionnaire, the search results can also be seen without the FRBR clustering at:
Important things to know in advance
- The unique purpose of this experiment is to explore new ways to present search results in order to help ANONYMOUS users to better find resources, AND NOT TO PROPOSE ANY “FRBR OPAC”!!! We only looked for inspiration in the main concepts of the FRBR model, and we never intended to build a formal FRBR database!!! In this sense, we are hiding from the presentation of results any reference to any FRBR concept, for not distracting/confusing the user.
- However, this prototype also is intended to mainly be evaluated by librarians or professionals with knowledge of the domain, so we also were not concerned in providing a “final product”. This means that the presentation of the results was decided with the main purpose of supporting evaluation, and not usability. For example, in order to see a specific record, we just present
the original MarcXchange record, and did not care to provide any other visualization…
- Please do not search with diacritics. For example, to search by
“L’étranger”, please use “L’etranger”. We found a problem with this in the last moment with the search engine that might be solved at any moment, but until there is remain an annoying limitation…
- This service is hosted in a “working server”, so you might experience a slow performance if too many users are trying it on the same time… If that occurs, sorry, but please be patient…
- After this evaluation we will produce a public document with the results, as also the detailed description of the architecture of the system and of the processes…
This collection contains 92.535 records extracted from the following Libraries:
- British Library
- French National Library
- German National Library
- National Library of Hungary
- National Library of Latvia
- National Library of Lithuania
- National Library of Portugal
- National Library of Serbia
- National Library of Spain
- National Library of the Czech Republic
- Russian State Library (only records from Russian authors)
- We also received a collection form the Royal Library of Belgium, but it was not possible to use it due to a detail in the records for which we had no time to adapt the FRBR system…
The generic criteria for extraction was “all the records where the author was a Literature Nobel Prize winner” (we got a lot, but as you can imagine, we also got some noise… we did not care too much about that, as we don’t think it was important…).
José Borbinha (IST ~V INESC-ID)
Nuno Freire (National Library of Portugal)