This week, Eva is presenting her and Martin's paper "Content-based User Models: Modeling the Many Faces of Musical Preferences" at the 19th International Society of Music Information Retrieval Conference (ISMIR) in Paris. This year, Eva is also co-chair for the Women in Music Information Retrieval activities at ISMIR (together with Audrey Laplante from the University of Montreal, Canada).
As a co-organizer of the annual PAN workshop Michael attended the Conference and Labs of the Evaluation Forum (CLEF) last week in Avignon, France. The style change detection task, organized by DBIS, had six successful submissions with very promising results, which were presented at the conference. Also in 2018, PAN could again top the number of registrations (227) and the number of active participants(41), making it one of the mostly active tasks at CLEF in general. Looking forward to next year!
This week, Eva will be presenting Martin and her work on multi-context music recommendation at the International Conference on Content-Based Multimedia Indexing in La Rochelle, France. We are particularly proud to say that our paper has received the Best Student Paper Award :)
Together with Univ.-Prof. Dr. Theo Hug (Department of Media, Society and Communication), Eva will be leading the project "Digitization and Visualization of Archives and Collections" that aims for developint and applying concepts, methods and visual tools for the interactive presentation and analyses of archivals. The project is generously funded by the Region of Tyrol (Leuchtturmprojekt).
The project builds on new possibilities of the development of generic interfaces and the visual exploration of archives and collections and combines them with regard to a modular set of open source software for visual presentation and examination of archival documents. The development and application of the concepts, methods and tools takes place prototypically on the example of the radical constructivism and the Ernst von Glasersfeld Archive, on the other hand, the modular tool set can generally be used in memory institutions in the cultural field (GLAM).
It's a busy week for DBIS. Eva is presenting her work on Culture-aware Music Recommender Systems at the User Modeling and Personalization Conference (UMAP) in Singapore. She is also the co-organizer of the IUADaptMe workshop at UMAP, which was successfully held on Sunday. At the same time, Martin is presenting Carl: A Sports Award Recommender at the eCom workshop at the 41st International ACM SIGIR Conference on Research and Development in Information Retrieval (SIGIR) in Ann Arbour, US.
We welcome our newest team member Mike, who started working for DBIS this week. He will effectively be succeeding Martin, who finished his Ph.D. last month.
Mike has written his Master thesis at DBIS, in which he explored the possibilities to use autoencoders for next-track music recommendations.
Together with cooperation partners from Taiwan, Eva has looked into how information about the user's current mood can be used to improve the ranking of musical tracks in a recommendation scenario. The resulting paper has just been accepted at the IEEE Journal of Affective Computing and is already published in an online first version. You can find a local copy of the work here.
Publication details:
Eva Zangerle, Chih-Ming Chen, Ming-Feng Tsai and Yi-Hsuan Yang: Leveraging Affective Hashtags for Ranking Music Recommendations. In IEEE Transactions on Affective Computing. 2018 (online first).
Today, Robert Binna will present our paper "HOT: A Height Optimized Trie Index for Main-Memory Database Systems" at this year's SIGMOD/PODS Conference. The developed main-memory index structure (called HOT) was joint work of Robert Binna, Eva Zangerle, Martin Pichl and Günther Specht together with Viktor Leis from the Technical University of Munich. The SIGMOD conference is one of the two major database conference (core ranking A*) and we are very proud to be part of this year's conference :)
You can find the paper here or in the official ACM SIGMOD proceedings.
This week, Günther Specht opened the 30th Grundlagen von Datenbanken-Workshop (GvDB) (22.-25.5.2018) in Wuppertal. This workshop is part of the oldest workshop series in the field of databases of the German Informatics Society. This year, the workshop will also feature keynotes by Prof. Rainer Mantey and Prof. Ralf Schenkel.
Our paper "#nowplaying-RS: A New Benchmark Dataset for Building Context-Aware Music Recommender Systems", co-authored by Asmita Poddar (National Institute of Technology, Rourkela, India), Eva Zangerle (DBIS) and Yi-Hsuan Yang (Research Center for IT Innovation, Academia Sinica, Taiwan) has been accepted at this year's Sound and Music Computing Conference.
Our research group is featured in the current issue of the Datenbank-Spektrum journal, which is the central outlet of the Fachgruppe Datenbanksysteme (working group for databases) of the German Informatics Society (GI). In this article, we present a short retrospective of the group's history and, more importantly, briefly present our research activities and accomplishments. You can find the article here.
It's a busy week for DBIS - Eva and Benjamin are presenting their work at The Web Conference in Lyon, France.
Eva is presenting her paper "Recommendation-Assisted Data Curation for Wikidata" together with Claudia Müller-Birn of Freie Universität Berlin at the Wikipedia Workshop. Benjamin is presenting his paper "Detecting Music Genre Using Extreme Gradient Boosting" at the Challenges track, participating in the WWW 2018 Challenge: Learning to Recognize Musical Genre.
This week, Eva presents the paper ALF-200k: Towards Extensive Multimodal Analyses of Music Tracks and Playlists at the European Conference on Information Retrieval (ECIR) in Grenoble, France. In this paper, we (Michael Tschuggnall, Stefan Wurzinger, Günther Specht and Eva Zangerle) present the ALF-200k dataset, which comprises lyrics and audio features of 226,000 songs. The dataset and code can be found here.
DBIS is represented twice on the 19th CICLing conference in Hanoi, Vietnam. On March 19th, Benjamin presented his paper "On the Influence of Machine Translation on Language Origin Obfuscation", which explains how the original language of translated documents can be detected. Later that day, Michaels work "Algorithmic Segmentation of Job Ads Using Textual Analysis" was presented as part of the poster session, yielding interesting discussions from many different fields.
Two DBIS papers were accepted at the International Conference on Computational Linguistics and Intelligent Text Processing (CICLing) 2018, a well-known and established conference in the field of natural language processing. Benjamin evaluated the automatic detection of the origin of a machine translated text, and Michael proposed a new algorithm to automatically segment job ads. Both papers will be presented at the conference at the end of this month in Hanoi, Vietnam and will be published in a Springer LNCS volume.
