2nd ATRIUM Researcher Forum: valuable user feedback and expert advice for the CLARIN Transcription Portal
By Carol Delmazo
The 2nd ATRIUM Researcher Forum focused on CLARIN Transcription Portal was successfully held on November 25, 2025, at the Ludwig-Maximilians-Universität München (LMU Munich) in Germany. This event follows the inaugural forum held in Poznań, Poland, in 2024, which focused on the OPERAS GoTriple platform. The primary goal of the ATRIUM Researcher Forum is to foster closer interactions between researchers using infrastructure services and the providers developing those services.
Core subject of Munich’s forum, the Transcription Portal is an easy-to-use web service designed for scholars across various domains. It allows users to transcribe, summarise, and translate audio files produced in Dutch, German, Italian, or English. Hosted by LMU, the service is discoverable through the CLARIN Language Resource Switchboard , part of the ATRIUM Services &Software Catalogue . The key question guiding the forum was: how can this service be improved? To answer the question, two different sessions were organised.
The morning session was dedicated to a usability test, where six students/researchers performed a set of tasks in the portal under the supervision of User Experience (UX) specialists from the Poznan Supercomputing and Networking Center (PCSS), providing detailed feedback. The afternoon session gathered six experts in fields such as data science, linguistics, phonetics, language and speech processing, and software in Digital Humanities to engage in strategic discussions regarding the current features and possible future developments of the portal.
Usability test
Participants generally found the interface to be intuitive and the process of uploading data and starting transcription straightforward. However, several points for improvement were identified.
Difficulties included the tiny size of the OCTRA window used for transcription and confusion surrounding the dropdown menus for selecting a suitable provider for the Automatic Speech Recognition (ASR) task. Users would appreciate more guidance, e.g. explaining which language a specific provider is best suited for. Participants also noted inconsistencies in the User Interface and felt there is room for improvement, e.g., the font sizes for key elements could be increased, whereas some instructions could be more concise to help users quickly identify where they need to click and why.
Despite these technical hurdles, the tool was judged to be very valuable for their academic work. Researchers focusing on German linguistics and dialect variation, as well as those handling time-consuming interview data, saw significant potential in the portal for making data more accessible and speeding up the processing workflows.
Expert Panel
The Expert Panel first focused on current features, starting with the Automatic Speech Recognition time stamps and speaker diarisation. More specifically, the experts debated the purpose of chunking the transcriptions into time intervals, whether it is needed and how to present it in the interface.
Experts then addressed the functionalities of summarisation and translation. While summarisation was confirmed as an ATRIUM project requirement, it was noted that its inclusion could “clutter” the user interface. They also debated the optimal workflow order for these features. A key structural recommendation was to cluster options based on user needs, such as historian versus phonetic expert.
The afternoon concluded by examining relevant future developments, for instance, the extension to other languages and dialects, including code switching. A crucial element raised was data security and GDPR restrictions, particularly that the Transcription Portal ensures that sensitive data remains within the academic network and does not leave through third-party providers.
Another theme was how to manage and monitor data processing in terms of expected processing duration and progress tracking. This discussion highlighted the dependency on external service providers. One perspective was that since the portal relies on external APIs for certain functions, this dependency is inherent. However, potential solutions to enhance control were explored, specifically by considering infrastructure partnerships and empowering the university to host services locally.
The final topic of the Expert Panel was the Integration with the CLARIN Language Resource Switchboard . Discussions centered on how users could export their results from the Transcription Portal and connect them for additional analysis and processing elsewhere. Furthermore, to help users decide how to proceed after transcription, suggestions were made to implement a wizard that could guide them.
Overall, the consensus was that the Transcription Portal is a tool with great functionalities and significant potential, especially for historians and for use in teaching, e.g. speech analysis.
According to Christoph Draxler, head of the Bavarian Archive for Speech Signals, the provider of the service, the ATRIUM Researcher Forum was very useful to gather feedback from both users and experts, which will result in a better user experience, and foster the development or incorporation of GDPR certificates. Henk van den Heuvel, as leader of Task 6.4 in ATRIUM and one of the service providers contributing to the Transcription Portal, concluded that the ATRIUM Researcher Forum showed the value of the Transcription Portal for a wide range of researchers and the Forum opened a treasure of feedback for further improvement within the ATRIUM project and beyond.
An internal report summarising the key findings will be produced to help the service providers with their next steps for improving the Transcription Portal.
