Artificial Intelligence Research Methodologies

Reading a Paper

Tutorial for Week Beginning 23rd October 2000

For your literature survey you will need to read 10-30 papers. Some of these you will read in detail, but some papers it will
only be necessary (or, indeed, possible in the timescale) to skim, ie those papers that are not part of the main development
of the topic, but represent tributaries. This tutorial will teach you how to skim a paper, which may not be a skill you have
previously learnt. 

Preparatory Task

Read the tips in the tutorial task below. Practice skimming a paper and summarising its main theme. 

Tutorial Task

You will be given a short paper to skim for 5 minutes. Your task will be to identify the main theme of the paper. You will
then be asked to explain this theme to your fellow tutees. The tutorial will end with a discussion of techniques for reading
papers at different levels of detail. 

Here are some tips on skimming papers to identify the main theme in a restricted time. 

When you first get a paper to skim you are usually trying to identify its key contribution. Look at the review form,
especially sections 1 and 3, for tips about how to unpack what this means. To find the contribution look first at the title and
the abstract. These may tell you all you need to know and you can stop there. If you decide to go on, read the introduction,
the section heads and the conclusion. You might also look at the bibliography to see which camp the author(s) come from
and whose material they build on. 

Failing this, the key idea may be found, for instance, in: a figure in which the structure of a system is outlined; a worked
example in which the main procedure is illustrated; a theorem which constitutes the main result; a table or graph where
the experimental results are summarised; related work where the key advantages over rival approaches are listed; or a
summary paragraph. 

If you decide you want to read the whole paper but do not have much time then try speed reading. Read the paper very
fast skipping any technical details, e.g. proofs, formal definitions, background material, program outputs, etc. Try to
understand the role of these bits without bothering with the details. Look for the main new ideas. Slow down if you get to
an important bit then speed up again afterwards. 

If you learn nothing else during your time at Edinburgh but how to skim papers then it will be time well spent. Almost any
job you get in the future will involve dealing with mountains of paperwork. If you can rapidly extract the sense from these
papers then you will survive. Otherwise, you will become bogged down and depressed. 

When you write a paper bear in mind that people may want to skim it. Make it easy for them by highlighting the theme of
your paper in title, abstract, introduction and conclusion and by clearly identifying the key passages. Your reward will be
that readers remember your paper with fondness.