Course development report 2026
This is the 2026 course development report for TDDE09 Natural Language Processing (6 credits).
Statistics
The 2025 session had 36 registered participants. Based on social security numbers, 32 out of these were men and 4 were women. There were 5 exchange students (Austria, Germany, Italy, Spain). The examiner was me, Marco Kuhlmann.
After the first examination (2025–03–28), 18 out of 28 students (64%) had passed the course. This number had increased to 27 out of 32 (84%) after the last examination (2025–08–30).
Course evaluation
The Evaliuate for the 2025 session was open between 2025–03–17 and 2025–04–13. It received responses from 15 out of 36 respondents (response rate: 42%). The overall grade for the course was 4.40 out of 5 (median: 5.00).
Below is a summary of the free-text replies.1
What students like about the course
Effective and well-structured course design: Students appreciated the thoughtful structure of the course, especially the combination of in-person lectures, high-quality video content, and quizzes. This multi-format approach helped make the material accessible and supported a deeper understanding of the subject matter.
High-quality labs and practical learning: The lab series received strong praise for being well-organised, educational, and supported by helpful assistants. Students found that the labs, projects, and oral examinations significantly reinforced their understanding of NLP, machine learning, and tools like PyTorch.
Skilled instruction and positive learning experience: The course instructor was frequently highlighted as a key strength, with students recognising his dedication and skill. Many noted that, despite a heavy workload, the overall experience was highly rewarding, engaging, and among the best in their academic journey.
What students think can be improved
Overwhelming workload: Students greatly value the course’s structure, comprehensive content, and the variety of learning methods (labs, quizzes, lectures, project). However, many felt the number of deliverables, especially when concentrated in a short period, was overwhelming. Suggestions included spreading the workload more evenly and slightly reducing the number of assessment components, or extending the course over a longer period to maintain quality while reducing pressure.
Time-intensive labs and project: Labs were consistently praised as the most beneficial part of the course for understanding complex concepts. The project was also appreciated as a meaningful, research-oriented task. However, both were considered very time-consuming, particularly in combination with each other. Some students proposed reducing the number or depth of labs to better support project work, or starting the project earlier to ease pressure during the exam period.
Communication, assessment clarity, and support: Several students noted that grading criteria, especially for quizzes and the oral exam, could be more clearly communicated from the start. Others requested more guidance for the oral exam and more flexible or continuous retake opportunities for quizzes. Some also felt that group project dynamics were hard to manage effectively and suggested more individual accountability or better-defined evaluation methods.
Examiner comments
I would like to thank everyone who responded to the course evaluation and took the time to share their experiences and suggestions!
Overall, the evaluation shows that students are very satisfied with the course. This is reflected both in the high overall rating and in the qualitative feedback, where students particularly highlighted the course structure, the quality of the labs, and the overall learning experience. I am especially pleased to see that the changes introduced for the 2025 session appear to have had a positive effect. In particular, the proportion of students who felt that the workload was too high has decreased to 60%, compared to 83% in 2024. While this indicates that the workload is still perceived as demanding (a point also clearly articulated in the free-text feedback) it suggests that the adjustments made so far are moving the course in the right direction.
Based on the student feedback, I have implemented the following changes for the 2026 session:
Content revision: I have decided to remove the unit on structured prediction (tagging and parsing), which had been part of the course for many years. While this topic remains close to my heart, the feedback made it clear that the course feels broad and intense, and at the same time there is strong interest in recent NLP methods. Removing structured prediction creates space for more material on language models and results in a clearer lecture narrative, organised around the idea of “language models, from basics to development”.
Workload reduction: Students repeatedly pointed out that the number and timing of deliverables can feel overwhelming. By removing the structured prediction unit, I was able to reduce the number of basic labs from five to four, directly addressing this concern. I have also removed the requirement to submit each lab for a written review, putting more focus on the feedback during the lab sessions. Each lab is now designed to take about eight hours on average. With roughly twelve tasks per lab, this corresponds to about 40 minutes per task. I hope that this makes the expected workload more transparent and manageable.
Project structure: Students consistently described both the labs and the project as very valuable, but also as difficult to combine in terms of time and energy. To address this, I have tried to give the project a more prominent and better-structured role. With the course now one unit shorter, students have an additional week dedicated entirely to project work. I have also increased the number of preparatory seminars from two to three, to provide a smoother ramp-up before the most intense part of the project work. My goal is to help students distribute their effort more evenly and to get more out of the project experience.
Footnotes
The summary was generated by ChatGPT and then checked and edited by the examiner.↩︎