Roeland Ordelman

Nik Vaessen, Roeland Ordelman, David A. van Leeuwen

This paper explores the use of Dutch archival television broadcast data for self-supervised learning of speech foundation models, specifically wav2vec 2.0. We first study data quality assumptions for pre-training, and show how music, noise and speaker overlap affect SSL convergence and downstream fine-tuning performance. Secondly, we explore effectively pre-processing strategies to convert the noisy broadcast dataset into a qualitative dataset for pre-training, by using Whisper and WhisperX. Thirdly, we compare mono-lingual and multi-lingual pre-training with equivalent amounts of data, and show that mono-lingual pre-training is more robust to out-of-domain data. Lastly, we achieve a state-of-the-art LARGE wav2vec 2.0 model for the Dutch language, by a continuation of pre-training a wav2vec 2.0 XLS-R model checkpoint with our 55k hour archival dataset.

T. Beelen, E. Velner, R. Ordelman et al.

In this paper, we identify challenges in children's current information retrieval process, and propose conversational robots as an opportunity to ease this process in a responsible way. Tools children currently use in this process, such as search engines on a computer or voice agents, do not always meet their specific needs. The conversational robot we propose maintains context, asks clarifying questions, and gives suggestions in order to better meet children's needs. Since children are often too trusting of robots, we propose to have the robot measure, monitor and adapt to the trust the child has in the robot. This way, we hope to induce a critical attitude with the children during their information retrieval process.

Robin Aly, Maria Eskevich, Roeland Ordelman et al.

This report describes metrics for the evaluation of the effectiveness of segment-based retrieval based on existing binary information retrieval metrics. This metrics are described in the context of a task for the hyperlinking of video segments. This evaluation approach re-uses existing evaluation measures from the standard Cranfield evaluation paradigm. Our adaptation approach can in principle be used with any kind of effectiveness measure that uses binary relevance, and for other segment-baed retrieval tasks. In our video hyperlinking setting, we use precision at a cut-off rank n and mean average precision.