Cristina España-Bonet (Saarland University), Juliane Stiller (Humboldt University of Berlin), Roland Ramthun (ZPID - Leibniz Institute for Psychology Information), Josef van Genabith (Saarland University) and Vivien Petras (Humboldt University of Berlin) convinced the audience with their paper presentation "Query Translation for Cross-lingual Search in the Academic Search Engine PubPsych". The conference participants were given the opportunity to vote for the best full research paper submitted and presented to the conference.
"We are all pleased that our approach, which can be easily transferred to other domains of content, had received such a positive response from the international scientists," says Roland Ramthun, project staff at ZPID. "This encourages us to continue to develop easily implementable and at the same time effective approaches for multilingual retrieval and, above all, to evaluate them on real inventory and request data".
The aim of the CLUBS project is an empirical evaluation of four different approaches in the field of "Cross-lingual information retrieval" based on the psychological search engine PubPsych. Thousands of potentially relevant publications in various languages appear every year. Due to language differences, researchers and other experts are only able to retrieve and incorporate a fraction of the literature pertaining to their fields into their own work.
Project Enters Its Final Year
In 2019, the project, which is funded by the Leibniz Association, enters its final year. The structures for the scientific evaluation of the project were created. "We are currently programming the final prototype of the system," says Roland Ramthun. "Models for the translation of the content are being trained in Saarbrücken with the help of state-of-the-art hardware and software and can be evaluated in a few weeks."
Next, people should test the system to see if the quality of the search results is really improving for the end user. "Because we don't just want a multilingual system, but one that really delivers better, more relevant results in multiple languages," says Ramthun.
For further information about the CLUBS project, please visit www.clubs-project.eu. The slides of the presentation at MTSR are available via ZPID's repository PsychArchives. The full paper will be published in Communications in Computer and Information Science (CCIS), Volume 846.
MTSR is an annual international inter-disciplinary conference, which brings together academics, researchers and practitioners in the specialized fields of metadata, ontologies and semantics research.