Presentations

During the semester, each student should give a presentation about an NLP-related topic.

The presentation is ungraded, but mandatory to pass the course.

Topic

The presentation can be about a tool, framework, technology, or just anything that you are interested in, as long as it fits into the broader context of NLP.

The idea is that you can explore an exciting NLP topic, hack around with some new framework, and share your findings with the class.

You can find the list of topics in the GitHub presentation topic pool issue.

If you are interested in a topic, please add a comment to the issue, stating the topic you would like to work on.

We will discuss the topic assignments in one of the first lectures.

Tip

You can also suggest your own topic! Just add a comment to the issue with your idea, and we can discuss it.

Note

If multiple students are interested in the same topic, we will figure out a way to assign the topics fairly. In case of doubt, a first-come-first-serve approach may be applied, so it's probably a good idea to be quick! 😉

Scope

The scope of the presentation should be around 15-20 minutes. Ideally, you should include a demo or practical example (but I am aware that probably not all topics are equally suitable for that).

For the practical part, try to aim for 25-50% of the time (depending on the topic). Feel free to use any tool you want. You can show something in the browser, use Postman or Jupyter notebooks, or do some live hacking on the command line. 🤓

Note

If you work on a paper, there may be fewer practiical parts as if you work on a tool or framework. That's totally fine!

Besides that, there are no strict rules about the presentation format. You can use slides if you want, but it is not mandatory.

Presentation Dates

We will discuss the dates of the presentations in one of the first lectures.

Submission

You don't need to submit the presentation slides.

For the practical part, please push your code to a GitHub repository and provide a README with instructions on how to run your code.

Ideally, your code should be directly executable with a setup that is self-explanatory and follows common conventions (e.g. npm start).

Example

Here are some examples of what I mean:

• If your practical part is a Jupyter Notebook, you could provide a mybinder link.
• If you build a web service, after I cloned the repository, the service should come up by running make or npm install && npm start.
• If you provide a Python script, it is acceptable to do a pip install followed by a python my_script.py, since this is conventional in Python.

Info

In case of doubt, feel free to ask in our discussion board or reach out to me directly.