Tips for Reading Academic Literature on Computer Science

Posted 21 April, 2025

The literature review process was one of the most difficult parts of writing my undergraduate dissertation. I had barely read academic literature about computer science before and it was an enormous task to find papers that were relevant to my research. It didn't help that I thought that reading more papers would inherently get me further in my research, and I often got sidetracked by papers that were interesting but not useful for my work.

I came up with a strategy for how to find and read academic papers more efficiently, and shared it with some of my friends at UST. Now I'll publish it here in hopes that it can be useful to other students.

Academic Reading

Academic reading can be intimidating, but it can be easily distilled down to a formula because academic articles / papers often have a similar format in computer science:

ABSTRACT: gives you a brief overview of the problem space and what the researchers are attempting to tackle
BACKGROUND: gives you a broader view of the problem space, then narrows down to the EXACT issue that the researchers are trying to solve. Can also describe how this research is different from other research (answers the question, 'what makes this research novel/cutting edge?')
RELATED WORK: gives you a list of papers/research that are similar to this paper/research
CONCLUSION: gives summary of the whole paper, including the problem space and areas where future research can be done

Those sections are present in essentially every single good paper that is published in academia. But there are a few types of research papers (this is not a conclusive list):

• Analytical research papers: papers that talk about the existing research -- how are things 'going' in this field? What are people currently working on, and what problems still exist? This can also be about the history of a field. If you see a 'survey' article that means it's most likely a review of the existing research in a particular area.
• Experimental research papers: present a new methodology or implementation of something in order to solve a problem. These will also include descriptions of how the researchers implemented a solution/new way of doing something programmatically, and then an evaluation of how successful the implementation was. In CS, you see a lot of experimental papers.

Here's my personal strategy for reading academic papers:

1. Read the title and abstract. This tells me almost immediately if the paper is related or completely unrelated to what I'm interested in
2. Read the introduction -- this tells me more about what the authors plan to tackle, i.e. the problem space
3. Read the beginning and end paragraphs of the background section -- this tells me more about the existing research in the field on this problem/objective and how well that's going. This lets me know if the researchers are interested in the same niche problem that I'm looking at. If I know the paper is very relevant at this point, sometimes I read more of the background section if it explains something I'm confused about -- there should be subsection titles that tell you what each set of paragraphs will talk about
4. Read the conclusion -- this quickly tells you what problem the paper was talking about, how they tried to solve it, and how successful their solution was. It also tells you more about 'future work': based on this paper, what are some good research areas or things to try in the future to solve this problem?

Finding Relevant Research Papers

Often there is a kindly researcher out there who has compiled a list of seminal works in your research niche. Most of the research I seek out to read in depth and cite directly is from the past five years (go any older and you'll likely miss out on the state of the art). But reading the most influential and high-quality works in a specific field will provide valuable context for your research and help you determine what is common knowledge, especially if you are new to that niche. Beyond lists compiled by academics, I mostly use Google Scholar to find relevant papers.

Additionally, if you are trying to understand a specific research topic, e.g. 'using supervised machine learning models to detect network attacks', it is a waste of time to try and fully read every single paper/article on the subject. Many are not really relevant, are outdated, or are simply bad. Instead, scan the abstract, intro, conclusion, and part of the background to determine if the paper is relevant/important enough to fully read. It is normal to read even part of a paper and not understand it. Academic writing is quite dense and can be pretty time-consuming to try and decode. It is very normal to have to read a paper many times before you fully understand it. Still, with time and practice, you can certainly learn to read computer science academic literature with high efficiency.

Finally, here is a link that helped me understand how to read/write academic stuff. It's specific to systems security research, but also contains a lot of good general advice:
https://fabio.pierazzi.com/blog/2021/literature-review-systems-security/