Informatics Research Methodologies
Literature Survey
Each student must prepare a literature survey on some topic in
Artificial Intelligence. The topic will be chosen to reflect the
student's interests. The
first tutorial is about selecting your literature survey topic.
Some suggestions for survey topics can be found here, but you are not restricted to these.
The survey should be based on between 10 to 30 papers. Students
are not expected to read all these papers in detail; in fact, one
of the skills to be learnt is the ability to read papers at
different depths. Students will be expected to locate the key
papers and read them in detail and to get a general enough idea of
the contents of the other papers to place them in context.
The literature survey is an assessable component of the module
carrying 40% of the total mark. Between 15 and 30 hours are
allocated to it. The mark will be based on the written report which
should be submitted electronically by 5pm on Sunday of week 10 as
a postscript file.
submit {aimsc,ai4} airm 1 <filename> <time spent>
A copy with feedback and a mark will be returned by Monday of week 1, term
2. Note that the mark awarded is provisional and must be
confirmed by the Board of Examiners.
The literature survey is designed to help you develop the
following skills:
- Reading at different levels of depth. You cannot read
all the papers your survey must cover. Identify the key papers and
read them all in the traditional way. If you are using some of them
for your presentation you may want to read these twice and/or try
the ideas on some examples of your own. Other papers you might only
skim (see more detail below). The second tutorial is about
reading research papers.
- Locating relevant papers. You will be given a few
pointers to papers to start you off, but then you are on your own.
To discover some more there are various methods. Look in the
bibliography of those papers you have already got to identify other
authors in the field. This will only take you backwards, however.
To go forward look for recent publications by the same authors or
other papers on this topic. You can search the library catalogue
(some are on CD-ROM). There are publications which give forward
pointers, i.e. list papers which cite some paper. You can
access the www pages of the authors or you can email them. You can
skim through relevant journals or conference proceedings. You may
find the Library's
online resources useful for tracking down papers
electronically.
- Organising the material. It will not be enough to
merely record who said what when. You must try and relate the
papers you have read. What were the key ideas and when and by whom
did they appear? How were they propagated? What were the dead-ends?
What are the rival approaches and the hot issues? An issue-based
organisation of the survey will be preferred to a merely historical
account, i.e. try to tell a story about the development of
the research.
- Managing your time. 15-30 hours is not a lot of time
and can easily be dissipated in undirected reading and fruitless
searching. You must structure your activities to make efficient use
of your time. Leave plenty of time for writing the report. Do not
leave everything until the week before the deadline.
The literature survey should be about 10-20 pages in length and
cover the following points:
- What papers were surveyed? Put a bibliography at the end.
- Which were the key papers in the area? Give an overview of the
content of these.
- How are the other papers related to these key papers? Tell a
story about the development of the field explaining how everything
fits together.
- What is the current situation? What are the hot issues? Where
is progress being made and what has proved to be dead end?
Alan Bundy