Homework 4 HT2023

Homework 4 - Reading a scientific article

Due Thursday Oct 12 22:00

Part 1 -  How to read an article

a. Read the following short article:

Mary Purugganan and Jan Hewitt, How to read a scientific paper   Links to an external site.(or found here as a pdf-file) Download pdf-file)

The following is additional voluntary reading on the same topic:

S. Keshav, How to read a paper, Links to an external site.

Philip W. L. Fong, How to read a computer science research paper Links to an external site.(page 1-3 only).

These articles reflect the background of their authors in different ways:

The first of these contains very good advice, and a template that you will use in Part 2 of the homework assignment. However, the authors have a background in biochemistry rather than computer science, which makes a few comments less relevant.

The second article is written by the computer scientist Srinivasan Keshav, now at the University of Cambridge.  He recommends three stages in reading a paper, where the last consists of working through the paper to virtually re-implement the results. In theory-based research this is excellent advice, but it may be less feasible in experimental fields.

The third article is also written by a computer scientist, Philip Wong. The last part of that article deals with advice on reading a paper as referee, and can be skipped.

Some additional optional reading on reading is:

Erren TC, Cullen P, Erren M (2009). How to surf today’s information tsunami: on the craft of effective reading. Links to an external site. Med Hypotheses 73: 278–279. 

Part 2 - Reading an article

a. Choose one of these scientific articles and read it, keeping the advice from part 1 in mind:

1. Aditya Ramesh et al., Hierarchical Text-Conditional Image Generation with CLIP Latents

2. Ricardo Vinuesa et al., The Role of Artificial intelligence in Achieving the Sustainable Development Goals.

3. Albert Ziegler et al., Productivity Assessment of Neural Code Completion.

4. Allison Koenecke et al., Racial Disparities in Automated Speech Recognition.

5. Stephanie Lin et al, TruthfulQA: Measuring How Models Mimic Human Falsehoods.

6. César Soto-Valero et al., A Longitudinal Analysis of Bloated Java Dependencies.

Full references and access information were not included - finding the article is part of the exercise.

All these articles are quite recent (2020 or later), two are written by KTH authors.

And please answer the following questions in your written homework assignment:

b. Discuss how and where you accessed the article, and provide a correct reference in the ACM (Association for Computing Machinery) format. Links to an external site.If there are several choices of access or reference, motivate which one you have chosen.

c. Where was the article published - in a journal, conference proceedings, or a preprint archive for example? Characterize briefly the medium of publication (or media, if it appeared in more than one way) - what was the scientific field of the journal (or other), what impact/recognition does it have, is it peer-reviewed? 

d. Apply the technique by Purugganan and Hewitt, and copy and fill in the template at the end of their article.  Pay particular attention to the fields Context and Significance.

e. List at least one article among the references of the article that you would like to follow up, and make a short note of why. Take a quick look at this article to see that it actually is relevant. Also, choose at least one newer articles that makes a reference to the article you have chosen and you would like to read, and make a short note of why you selected it.

f. Even though the article by Purugganan and Hewitt contains some good advice on reading, it mostly relates to other scientific fields than computer science, and contains some statements that may not be entirely relevant to articles in CS. Sample at least two of the articles listed, and check whether they follow the IMRD format. Discuss your findings: in particular, are there subfields of computer science where this format is more appropriate, and others where it is not really appropriate?

g. All the articles listed above are considered significant recentcontributions to their subfield. This may be reflected in how often they have been cited by others - how many times has this happened in the case of the article you have chosen?

h. Finish this exercise by stating, as clearly as possible in at most two sentences, the essence of the contribution of the article you have chosen to computer science, or science in general.

Note - your answers are allowed to be brief and concise, for example in part d.

Some of the articles may have been published in journals that are not open source, and need to be accessed through the publisher. The KTH Library has subscriptions to these journals, which lets you access them. If you have trouble accessing articles when not connected to eduroam, the KTH Library can help you, see:

• Databases and search tools 
• Access electronic resources off campus

Handing in your solution

Please save your solution as a pdf file and submit it in Canvas 

Peer grading

There will be no peer grading of this exercise.

Feedback from your TA

Your TA will grade your submission and report the result in Canvas.

Complete means you have passed the assignment.

Incomplete means you have to hand in a revised version.