Final Assignment
The point of the Project Assignments is to try out the skills you've learned in the course on your own dataset. The goal is ambitions because we want you to try to examine your dataset and learn something new. Something new that you've learned is also called a "research finding". This means that you can write a scientific paper about it. (You may be thinking. What!? What? That's too much!! But not to worry, we will guide you. More on this below.
The point is simply that we've been working on understanding networks and natural language processing, so the idea is to find a dataset to analyze that will let you show off that you can use these tools to learn about the world.
Here are some example datasets that it might be fun to work with.
- Specialized wikis for whatever you're into. E.g.
- Wookiepedia
- Game of Thrones Wiki
- Simpson's Wiki or one of the other Simpson's wikis.
- Bruce base
- Etc, etc.
- IMDb (networks of movies connected by actors), reviews of movies, movie durations, etc. For movie-datasets, also check out https://grouplens.org/datasets/movielens/latest/
- Wikileaks
- More examples can be found in Jim Bagrow's list
- And new ideas for datasets are very welcome. You should work with something that interests you - that way the project will be much more fun to work on.
Combining datasets is encouraged. This is a top way to learn new things (e.g. how the weather impacts network structures on Reddit or something)
You will be working together in groups just as for the first two assignments.
The first part of the final project is a 2 minute movie, which should explain the central idea/concept that you will investigate in your final project. You're making the movie so that the TAs and I can give you feedback, and so that other groups can 'steal' your ideas (and you can steal ideas from them). The movie must contain the following
- An explanation of the central idea behind your final project (what is the idea?, why is it interesting? which datasets did you need to explore the idea?, how did you download them)
- A walk-through of your preliminary data-analysis, addressing
- What is the total size of your data? (MB, number of rows, number of variables, etc)
- What is the network you will be analyzing? (number of nodes? number of links?, degree distributions, what are node attributes?, etc.)
- What is the text you will be analyzing?
- How will you tie networks and text together in your paper?
But other than that, there are no constraints on the video format. And we do appreciate funny/inventive/beautiful movies, although the academic content is most important. Note that we'll display the movie to the entire class.
Handing in the assignment: Simply upload your video to youtube (the higher the resolution the better) and submit the link to peergrade.
Ok. So a paper and a finding. What's that about?
What I'm trying to do is to push you to go beyond simply running the analyses we did on the superheros and actually use your skills to do some science. By asking you to present a finding, what I want is for you to think carefully about the results of your analyses. What is it that you learn through your analysis? Once you can answer that, you have the first steps towards a real research finding 🤓
It doesn't have to be something super fancy. Maybe your finding is just that the houses in Game of Thrones don't align with the communities in the networks. Then think about why that's the case given what you know about the show, and run further analyses to support and expand your theories. Or you find that the most important character in the Simpson's is Moe the bartender, then you confirm using additional centrality measures and investigate if this changes over time in the episodes. If you find change-points, when do they happen. Etc. Perhaps you have a hypothesis that you'd like to disprove. You can of course also be much more ambitious - I'm giving these examples to provide a sense that it's perhaps not so hard.
The deliverables for Project Assignment B are.
-
A Paper (
.pdfformat). The paper should contain your analysis, it should tell the story about the data and the research finding. -
An Explainer Notebook (
.ipynbformat). The Notebook should contain all the behind the scenes stuff, You should link to the notebook from the paper (in the Methods section).
Click on the link below to hear me talk about the elements of scientific papers.
IT IS SUPER IMPORTANT THAT YOU WATCH THIS VIDEO BEFORE YOU START WRITING THE PAPER. IT HAS A LOT OF INFORMATION WHICH WILL HELP YOU ACHIEVE A GOOD GRADE
The paper should have all the elements that are in the templates (links below):
- Abstract
- Significance Statement
- Introduction
- Results
- Discussion
- Methods
- References
In the video I discuss about tips & tricks that will bring success. Also use the template's author contributions (it is not OK simply to write "All group members contributed equally".).
BUT YOU ONLY HAVE MAX 5 PAGES (everything must be within those 5 pages, also the references) and max 5 figures. Less/shorter is OK, longer is not OK.
Link to templates here
Make sure that you use references when they're needed and follow academic standards.
TIP: When you have an idea for analysis, do a search to see if someone already studied your dataset - or the question you're interested in. There are lots of cool analyses out there that could be an inspiration. And it's OK to use that stuff, just remember to cite the work that you're drawing on (if you copy without citing, then it's cheating ... don't cheat).
The notebook should contain the code for your analyses. We appreciate if notebooks are commented and structured nicely. In addition, you can include the following information
- Data and Stats.
- Write in more detail about your choices in data cleaning and preprocessing
- Did you do analyses / calculate statistics that didn't make it to the main text, put them here.
Handing in the assignment: Simply upload your
- The main point is to show off what you've learned in the course, so the first thing is to make sure your dataset contains both networks and text.
- That you did a thorough analysis that shows what you've learned in the class. (And we can only know about this if you use and show key parts of that analysis in the paper itself.)
- Did you manage to get to a research finding about your dataset? (And not just reproduce the analyses from the lectures on your own dataset)
- All the formal things in the paper-writing video. Good abstract, Intro, etc. Informative figures with thoughtful captions, etc. The right references. Readable text.
- A well structured explainer notebook.
-
Q: May I use methods for analyses we didn't learn in class?
- A: Yes for sure! But remember that the point is to show off what you've learned in the class, so using tools from network science and NLP is essential.
-
Q: My network only has 50 nodes, is that enough?
- A: Hmm. That's a tough question to answer. My best guess is "probably not". But it could be that you have interesting temporal information and lots of textual data for your network, perhaps it could be great. If in doubt, come and talk to me.
This class has been hand crafted for you by Sune Lehmann.
This work is licensed under a Creative Commons Attribution 4.0 International License.
