HS Wie kann Qualität automatisch beurteilt werden und warum brauchen wir das? (MA)
Instructors: Moritz SchaefferShortname: 06.008.0518
Course No.: 06.008.0518
Course Type: Hauptseminar
Requirements / organisational issues
Most of the research we will be reading is published in English, which is why good English skills are important. Basic mathematical knowledge is not a prerequisite, but a call to action for me to provide you with the necessary tools which allow you to understand the metrics, models and approaches.Anwesenheitspflicht
It is not compulsory to attend classes.Recommended reading list
Moorkens, J. et al. (eds) (2018) Translation Quality Assessment: From Principles to Practice. Cham: Springer International Publishing (Machine Translation: Technologies and Applications). Available at: https://doi.org/10.1007/978-3-319-91241-7.Contents
Determining the quality of a translation is not easy. It often depends on subjective criteria, and it is difficult to formalise. However, it has long been extremely important for both human and machine translation to do so. We will first look at classic error typologies and quality metrics before delving into newer methods based on word vectors before we focus on how to automatically predict the quality of a translation. In this context, we will deal with the question of what AI actually is and how current models actually work, on which theoretical and on which data basis these models are developed, what machine learning actually is, to what extent human and machine learning are (dis)similar, what all this means for us as translators and humans – but above all, we will look at how to implement old and new quality metrics in practice, i.e., how to automatically determine the quality of a translation (whether of human origin or not), how well these metrics express what is necessary in certain situations or for certain purposes, we will explore the potential and limits of each metric. You will deal with theoretical foundations – relevant not only for quality metrics and practical implementation – but also for the implementation in industrial practice.Active participation (weekly readings and a mini-term paper) and a small research effort are necessary to pass.
Dates
| Date (Day of the week) | Time | Location |
|---|---|---|
| 04/19/2023 (Wednesday) | 13:30 - 15:00 | N.210 Hörsaal 8512 - FAS Hörsaalgebäude |
| 04/26/2023 (Wednesday) | 13:30 - 15:00 | N.210 Hörsaal 8512 - FAS Hörsaalgebäude |
| 05/03/2023 (Wednesday) | 13:30 - 15:00 | N.210 Hörsaal 8512 - FAS Hörsaalgebäude |
| 05/10/2023 (Wednesday) | 13:30 - 15:00 | N.210 Hörsaal 8512 - FAS Hörsaalgebäude |
| 05/17/2023 (Wednesday) | 13:30 - 15:00 | N.210 Hörsaal 8512 - FAS Hörsaalgebäude |
| 05/24/2023 (Wednesday) | 13:30 - 15:00 | N.210 Hörsaal 8512 - FAS Hörsaalgebäude |
| 05/31/2023 (Wednesday) | 13:30 - 15:00 | N.210 Hörsaal 8512 - FAS Hörsaalgebäude |
| 06/07/2023 (Wednesday) | 13:30 - 15:00 | N.210 Hörsaal 8512 - FAS Hörsaalgebäude |
| 06/21/2023 (Wednesday) | 13:30 - 15:00 | N.210 Hörsaal 8512 - FAS Hörsaalgebäude |
| 06/28/2023 (Wednesday) | 13:30 - 15:00 | N.210 Hörsaal 8512 - FAS Hörsaalgebäude |
| 07/05/2023 (Wednesday) | 13:30 - 15:00 | N.210 Hörsaal 8512 - FAS Hörsaalgebäude |
| 07/12/2023 (Wednesday) | 13:30 - 15:00 | N.210 Hörsaal 8512 - FAS Hörsaalgebäude |
| 07/19/2023 (Wednesday) | 13:30 - 15:00 | N.210 Hörsaal 8512 - FAS Hörsaalgebäude |