This instruction briefs instruction about writing a literature review as your Bachelor thesis. Remember that to read a wide range of articles, have a summary, and keep writing and revising are the keys to a successful literature review. Please actively discuss with your advisor if you have any problems during this procedural circle. Your advisor will try his/her best to give advice or help you solve some problems if necessary.

Reading

Keep reading research papers. You need to skim >=100 articles and read ~50 selected papers in detail. Try the three-pass method. As I advisor, I will recommend several articles to get started. However, you should also search for more by yourself and decide the useful articles to be included in your thesis.

Where to Find Papers?

Many use Google Scholar. Here is a nice blog about how to search effectively with Google Scholar. Generally speaking, the CS community focuses more on conferences. Finding paper from top venues is a safe way to find papers.

Conferences: computational linguistics(ACL, EMNLP, COLING, NAACL), artificial intelligence and machine learning(NeurIPS, ICLR, AAAI, IJCAI), data mining (SIGKDD, ICDM, CIKM, WSDM, The Web Conference)

Journals: computational linguistics and data mining(TACL, TKDE, TNNLS, etc.), healthcare (Bioinformatics, Journal of biomedical and health informatics, Journal of Biomedical Informatics, etc.)

When you’re reviewing some informal publications (e.g., preprints in arXiv), you must carefully tell the quality. Always remember to discuss this with your advisor.

Writing

Academic writing is not easy. I recommend to check out some nice references for academic writing for computer science, for instance, instructions from Renée J. Miller, Richard Zanibbi, and my personal collection of references for scientific writing.

Some quick tips:

  • Avoid mechanistic reporting when you are reviewing papers.
  • Don’t overuse enumeration.

You will use Overleaf and the Aalto thesis template for editing. I have shared / will share the template with you via Aalto email.

Categorization for Literature

The most important task of a literature review-based thesis is to organize your review in a fancy way.

You can learn the categorization from published surveys or review articles (For example, Fig.3 of my recent survey). You need to propose your outlines, discuss with me, propose your coherent taxonomy, and conduct the qualitative and qualitative analysis of current literature. I will give you feedback and help you to refine it. The following example shows you a simple outline of how to structure your thesis.

(Describe in the introduction) What is general background knowledge of your topic (give a brief introduction, find a formal definition or some informal description)? Why are specific techniques such as deep learning useful in this task (think about your topic’s challenges, what has been solved, what is ongoing, and the research trends)? What categorization of current literature would you propose in your thesis (briefing in the introduction, and explaining in the next section in detail)?

(Consider more in-depth in categorization as Sec. 2 ) What is the framework of your literature review in detail? This section is the divide step in the divide-and-conquer strategy. Perhaps you can categorize existing papers into two folds, i.e., methodologies and applications. Or you can also propose more complex and reasonable categorization (this will be your main contribution, that is, a better proposal, a higher grade).

(Sec 3. conquer, i.e., address specific subcategories) What are the main issues of each subcategory? What models or methods each paper proposes, and what challenges those models to solve (reviewing the articles your advisor sent and you searched)? What makes those models different (summarize their pros and cons)?

(Sec. 4) Are there any practices in the real-world applications (For example, connecting DL models to the industry applications)? Are there any other issues to consider like privacy, ethics, fairness, bias, etc.?

(Sec. 5 Provide some useful resources) Are there any resources, such as datasets, software, etc.

(Sec 6) Are there any unsolved problems or limitations of current research? What is the future direction? What conclusion do you want to make?

Where to Start?

Introduction is for sure the first place to start. However, it would be hard to write the first paragraph as you may not be familiar with a specific field. In this situation, you can write a very draft version, which can be a working plan or a brief plan to categorize your thesis. With this categorization, you’ll have a big picture (maybe not very clear). Do not try to write a perfect introduction at the very beginning. Come back to revise the introduction section after you have finished the main body or you’re confident about the topic.

Improve Your References

  • use google scholar to generate bibtex items for your references. Here is a useful Chrome plugin called BibTex Quick Copy for Google Scholar

  • Check your references if everything is complete. Don’t miss pages, volumes, or publication venues.

  • BibTeX (in many bibliography styles, including ACL’s) lowercases the titles of conference papers and needs to be told which letters not to lowercase. So if your title has letters that should always be capitals, please protect them with curly braces, for example:

    title={ {Can LSTM Learn to Capture Agreement? The Case of Basque} }

Checklist Before Sending Your Draft

  1. Check all terms with a capital letter; make sure they’re consistent.
  2. Check your spelling and use American English
  3. Enumerate all symbols and abbreviations and check if they are consistent.
  4. Check grammar using Grammarly.com. You can also arrange academic writing consultations with Writing Clinics at Aalto Language Center
  5. Use Turnitin to check the originality for your thesis and make sure the similarity score is less than 20%.

Computing

Some of you may conduct small-scale experiments. Check out the university computing resources via this link. For heavy jobs, I will invite you to the CSC cluster (optional). If you use CSC, refer to the CSC documentation.

The End

I hope this may help. Please keep reading and begin to write whatever you can. You never know what’s gone happen unless you start to write something. Enjoy your writing!