In the following, we give a few hints for preparing and giving a good talk at a seminar in the Computational Data Analytics group of Prof. Fürnkranz, öffnet eine externe URL in einem neuen Fenster. These also are points to which we pay attention when we grade your talk. While we believe that most of these recommendations are generally applicable, seminars in other groups may follow different criteria.
Your task in the seminar is to explain the topic that you received to the participants of the seminar. The topic is typically related to a paper, but you should not confine yourself to reading only this paper, or, conversely, to present every detail that is discussed in this paper.
Papers are often available on-line. Note that the download often only works when you are in the university network. If you have problems obtaining a paper, let us know (you are not supposed to buy them).
A prerequisite for being able to explain a paper is that you comprehend the paper yourself. This is a challenging task, research papers a generally written for other researchers in that area who have a lot of prior knowledge in the field of the paper. You (and the other seminar participants) may not have this. So in order to understand the paper you have to do a bit of research in order to understand the basics upon which the paper is built (Hint: Google and Wikipedia are often helpful, but there is a reason why research papers refer to other papers when they briefly explain the foundations of their paper). Survey papers or video tutorials are also often good sources of information.
Once you understood the paper, you need to prepare a presentation. If you follow the paper paragraph per paragraph and excerpt the most important information on to the slides, it will probably be a bad presentation. Try to structure your presentation first. Maybe it helps if you ask yourself questions like:
Also think about how your paper relates to the previous talks in the seminar. You generally do not need to explain things in detail that have been covered in a previous talk (although it might be a good idea to briefly remind the audience).
Good slides should not contain too much text. Long texts are generally hard to comprehend on slides. Try to make the text on slides as short as possible, while making sure that the key points are always mentioned on the slides. The audience should be able to read the slides while they are listening to you. Ideally, the slides should accompany your talk, i.e., they should, e.g., contain figures that illustrate what you say when you talk.
Conversely, however, the slides should also not contain too little text. Your fellow students should be able to use your slides for trying to understand what your paper was about when they need to prepare their presentation.
For similar reasons, it is typically not necessary that you present elaborate mathematical proofs if your talk. In most cases, it suffices if you only present the key result and the proof idea. Try to abstract from the math to the content.
It is, of course, good if you practice your talk, but you should not overdo it. If you learn everything by heart, it often makes a very boring talk. Conversely, do not just read off the slides. The slides should accompany you, but you should talk about the topic.
It is always good to involve the audience in some way (asking questions, etc.).
Try to make the presentation your own presentation. Avoid statements like "The authors of the paper do/say/claim ...".You need to explain the topic, it is your presentation.
Similarly, talk about the topic, not about the paper.
During your talk, you should keep an eye on the time (usually 20-30 min + questions). We do not strictly enforce time limits, so if you are a few minutes shorter or longer, this is no problem at all (we won't even notice it because we usually don't time the talks). It will become a problem if you run considerably over (or under) time, so much that it is noticable without a clock. If you realize that you are going much too long, you can also try to skip a few slides in your presentation on the fly.
In general, it is hard to give a good estimate of the number of slides that you need because different people need different amounts of time for a single slide. It is, however, usually a bad idea to have less than a minute per slide because then the audience can not really read it anymore. So if you have more than 30 slides, you should probably practice and see how long it takes you. In most cases, 20-25 slides should be enough for a 30 min presentation.
Feel free to use materials other than those given to you with the paper. If you find videos, pictures, even a different presentation on the topic on the Web, fine, you can use and adapt it to your needs. However, you must acknowledge all sources that you used either directly on the slides (typically for figures, tables, videos, etc.) or at the end of the slides (typically for additional papers or presentations that you used, etc.).
Needless to say, in the end, it should still be your presentation, and not you talking about somebody else's presentation.
You must be able to answer questions to all aspects of your presentation. If you write something on your slides, you must be able to explain it. If you don't understand a sentence you write on the slides, you should be alarmed. Pay particular attention to figures you include (do you understand what they say, can you explain all aspects, do you know what the axes of the graph mean, etc.).
"The authors did not explain this in the paper" is a very bad answer to a question. It typically means "I did not try hard enough to understand this.".
It is not required that you contact the organizers of the seminar prior to your talk, and most students don't. However, you may, of course ask them for assistance if you don't understand aspects of the paper, want feedback on your slides, etc. In this case, contact them per E-mail (there is typically a contact address specified on the seminar page) early and explain the problem so that they can prepare an answer. Note that they may need some time for that because they may also not know the answer off-hand.
We strongly encourage students to send us a preliminary version of the slides one week prior to the talk. This is not immediatelz relevant for grading, but we will try to give you some feedback on the slide prior to your presentation, so that you can still improve it in time before your talk.
At the point of this writing, we do not give much feedback on your talk during the seminar, because we do not want to discuss the performance of a student in front of other students (in particular if the performance was bad). However, based on feedback that we received, we may change this in the future.
In any case, you are very welcome of ask for feedback in private, e.g., by talking to organizers immediately after your talk or in their speaking hours.