Jump to main content

BMBF Collaborative Project KA³ – Cologne Center for the Analysis and Archiving of AV Data

The KA3 collaborative project was funded by the BMBF in two funding phases from 2015 to 2020. As part of the project, the infrastructure for curating and archiving audiovisual data at the Cologne site was fundamentally overhauled, thereby also adapting the technical foundation of the Language Archive Cologne (LAC) for long-term and more intensive use. The project is also testing the application of machine learning methods to identify specific phenomena that can facilitate—or even replace—the process of manually annotating large datasets. The results of the project form the current technical foundation of the LAC and the KA3 services for audio analysis.

Phase 1

Grant ID: 01UG1511
Duration: Oct. 2015–Sept. 2018

Phase 2

Grant ID: 01UG1511A
Duration: 10/2018–09/2020

Partner Institutions

  • Data Center for the Humanities (DCH), University of Cologne
  • Digital Averroes Research Environment (DARE), University of Cologne
  • Department of General Linguistics (ASW), Institute of Linguistics (IfL), University of Cologne
  • Thomas Institute, University of Cologne
  • Regional Computing Center (RRZK), University of Cologne
  • German Memory Archive (ADG), FernUniversität Hagen
  • Fraunhofer Institute for Intelligent Analysis and Information Systems (IAIS), Sankt Augustin
  • (First phase only: Max Planck Institute for Psycholinguistics (MPI-PL), Nijmegen)

Management

  • Prof. Dr. Nikolaus P. Himmelmann (IfL)
  • Prof. Dr. Andreas Witt (DCH/IDS)
  • Prof. Dr. Ulrich Lang (RRZK)

Team

  • Jonathan Blumtritt (CCeH)
  • Anke Debbeler (DCH; starting in 2019)
  • Mark Eschweiler (DARE; starting in 2019)
  • Anne Gerlach (DCH; starting in 2019)
  • Jochen Graf (RRZK)
  • Lukas Mönch (DCH; starting in 2020)
  • Miguel Ramirez Peña (DCH)
  • Felix Rau (DCH)
  • Christoph Stollwerk (RRZK)

Contact

Felix Rau (DCH)
f.rau(at)uni-koeln.de
+49 221 470-5832
 

Phase 2 of the funding program

In the second funding phase, the focus is on completing and testing a dynamic repository for audiovisual data. During implementation, the project will take into account the sometimes very different requirements of various disciplinary approaches regarding discoverability, access control, and metadata and annotation formats. This will ensure that the repository is open and versatile enough to be used across different disciplines. A dynamic repository for audiovisual data remains the most urgent need, as there is currently no stable global offering in this area that allows user-controlled uploads and downloads while simultaneously ensuring secure regulation and control of access permissions.

With regard to the components of audio mining and document analysis, the focus is on stabilizing and providing user-friendly access to the tools developed by the Fraunhofer Institute IAIS—which were tested during the first funding phase and modified for scientific research. In the area of audio mining, these include the automatic transcription and tagging of German-language audio-visual recordings, as well as language-independent speaker identification (assigning the speech signal to different speakers). In document analysis, the focus is on text line detection, which serves as the basis for all subsequent analysis steps in the study of handwritten manuscripts.

As in the first funding phase, the development of the repository and the analysis tools will be accompanied by three pilot studies in linguistics, oral history, and philosophy. The pilot studies ensure that the user perspective is consistently integrated into the development process and that all developments are immediately tested for their practical utility.

Conference Papers (Selection)

Research Data and Humanities – RDHum 2019
Oulu, August 14–16, 2019. “Challenges and Developments in the Preservation and Publication of Large Audio/Video Data.” Workshop: Jonathan Blumtritt*, Johan Frid, Jens Larsson, Martin Matthiesen, Felix Rau*.

6th Annual Conference on Digital Humanities in the German-Speaking World (DHd) 2019 “Multimedia & Multimodal”
Frankfurt & Mainz, March 25–29, 2019.

“Quality Standards and Interdisciplinarity in the Curation of Audiovisual (Speech) Data.” Workshop: Thomas Schmidt, Jonathan Blumtritt, Hanna Hedeland, Jan Gorisch, Felix Rau, Kai Wörner. doi:10.5281/zenodo.2596094 “Metadata

in the Age of Google Dataset Search.” Presentation: Jonathan Blumtritt, Felix Rau. doi:10.5281/zenodo.2596094

INEL Workshop (Indigenous Northern Eurasian Languages) “Linguistic Diversity, Minority Languages, and Digital Research Infrastructures.”
Hamburg, September 20–21, 2018. “Applications and Limits of Machine Learning for Language Documentation Resources.” Presentation: Felix Rau.

2nd Workshop “Research Infrastructures for the Humanities,”
Berlin, April 10, 2018. “eHumanities Center: KA3.” Presentation by Jochen Graf.

5th Annual Conference on Digital Humanities in the German-Speaking World (DHd) 2018: “Critique of Digital Reason,”
Cologne, February 26–March 2, 2018.

“Audio Mining for the Humanities and Cultural Studies: Use Cases and Challenges.” Workshop: Joachim Köhler, Almut Leh, Nikolaus Himmelmann, Felix Rau. doi:10.18716/KUPS.8085 “User

Support and Recent Developments in Research Data Repositories for Audiovisual (Speech) Data.” Workshop: Jonathan Blumtritt, Felix Rau. doi:10.18716/KUPS.8085

8th DINI/nestor Workshop “Research Data Repositories,”
Stuttgart, November 27, 2017. “Generic Components and Discipline-Specific Requirements in the KA3 Project.” Presentation: Jonathan Blumtritt, Christoph Stollwerk.

3rd Annual Conference on Digital Humanities in the German-Speaking World (DHd) 2016 “Modeling – Networking – Visualization: The Digital Humanities as an Interdisciplinary Research Paradigm.”
Leipzig, March 7–11, 2016. “User Experience of Language Archives: A Reassessment of the Interaction Between Archives and Users.” Presentation: Jonathan Blumtritt, Felix Rau. Link to the Book of Abstracts | Link to the slides

*Presenter* in cases of multiple co-authors | external co-authors

Publications by the project partners (selection)

  • Gref, Michael, Christoph Andreas Schmidt, Sven Behnke, and Köhler, Joachim. “Two-Staged Acoustic Modeling Adaptation for Robust Speech Recognition: The Case of German Oral History Interviews.” In IEEE International Conference on Multimedia and Expo, ICME 2019 Proceedings. Shanghai, China/Piscataway, NJ: IEEE, 2019. pp. 796–801. publica.fraunhofer.de/documents/N-555493.html
  • Gref, Michael, Joachim Köhler, and Almut Leh. “Improved transcription and indexing of oral history interviews for digital humanities research.” In European Language Resources Association (ELRA), Paris: LREC 2018, Eleventh International Conference on Language Resources and Evaluation. Proceedings. May 7–12, 2018. Phoenix Seagaia Conference Center, Miyazaki, Japan. Paris: ELRA, 2018. pp. 3124–3131. publica.fraunhofer.de/documents/N-494202.html
  • Gref, Michael, Christoph Andreas Schmidt, and Joachim Köhler. “Improving robust speech recognition for German oral history interviews using multi-condition training.” In Informationstechnische Gesellschaft (ITG): Speech Communication. 13th ITG Symposium on Speech Communication 2018: October 10–12, 2018, Oldenburg, Berlin: VDE-Verlag, 2018 (ITG Technical Report 282). http://publica.fraunhofer.de/documents/N-531366.html
  • Köhler, Joachim, Nikolaus P. Himmelmann, and Almut Leh. “KA3: Speech Analytics for Oral History and the Language Sciences,” ERCIM NEWS, 111 (2017), pp. 13–14.
  • Köhler, Joachim, Michael Gref, and Almut Leh. “KA³. Further Development of Language Technologies in the Context of Oral History,” BIOS – Journal of Biographical Research, Oral History, and Life Course Analysis, 1–2/2017. pp. 43–59. https://doi.org/10.3224/bios.v30i1-2.05
  • Leh, Almut, Joachim Köhler, Michael Gref, and Nikolaus Himmelmann. “Speech Analytics in Research Based on Qualitative Interviews: Experiences from KA3.” VIEW Journal of European Television History and Culture 7, no. 14 (2018). pp. 138–49. https://doi.org/10.18146/2213-0969.2018.jethc158
  • Leh, Almut, Michael Gref, and Joachim Köhler. “Audio Mining: Advanced Speech Analytics for Oral History.” Palabras y silencios = Words & Silences (2019). 9 pp. http://publica.fraunhofer.de/eprints/urn_nbn_de_0011-n-5690113.pdf
  • Trilsbeek, Paul. “Migrating The Language Archive to a New Repository Solution.” In Open Repositories 2019. Hamburg, 2019. pp. 41–44.
  • Trilsbeek, Paul, and Menzo Windhouwer. “FLAT: A CLARIN-Compatible Repository Solution Based on Fedora Commons.” In CLARIN Annual Conference 2016. Aix-en-Provence, 2016. hdl.handle.net/20.500.11755/b72c4df0-9f35-4f4e-9725-a36bcecd5723.