Researching a topic in Computer Science is a lot like doing a thorough
in English or History. When you explore a subject, you don't just use
the original, seminal paper or book that describes the fundamental
Instead, you use many resources that both refer to and are referred by
central paper or manuscript. The result is much more than just a
summary of a paper.
You have to form connections between what is presented and what you
How to do (database) research
Read and critique actively
Whenever you read a technical paper, it's important to read
This means not only understanding the details of what is presented, but
to understand the larger picture of why the ideas presented are a
the field. For any work, ask yourself:
Here are some suggestions of concrete things you can do to be an active
- Upon what work is the author building? What is the context?
- What is the author's motivation for looking at a novel, alternative
- What does the new idea accomplish? What advantage does it have or
what disadvantage does it avoid when compared to similar methods?
- What are its shortcomings?
- Are there certain circumstances under which the new idea performs
well or poorly?
Were assumptions made that, if lifted, would decrease the performance of
Next, we'll talk about how to find papers you are interested in.
- Keep notes. For some, this means keeping a log of answers to
the above questions for each paper. It can also help to paraphrase the
you do understand and to keep track of the confusing parts or terms. If
confusing part/term is critical to your understanding of the work, find
resources that will help you explain them.
- Do bibliographic backtracking. At times you'll want to find
article which is the original paper about a topic so you can learn about
fundamentals. You may think you've found it when, in fact, you haven't.
the introduction seems to assume the reader knows more background about
the topic than what's presented, you probably haven't found the original
In either the introduction or in a "related work" section, try to find
that would lead you to a promising earlier work. Find that paper and
You may have to backtrack through several bibliographies this way before
you find the material you need.
- Use abstracts prudently. Paper titles can be deceiving.
abstracts and section headings of a work first so you can see if
is really worth reading in depth.
Know the sources
In the database community, there's a core set of conference and journal
that researchers go to whenever one wants to know about the
the field. Here are some of them. This is not a complete list, but it
will help you
decipher the acronyms on the topic list.
- VLDB. Very Large Data Base. They publish both a conference
and a journal (the latter is most often called VLDBJ).
- SIGMOD. Special Interest Group on the Management Of Data.
I have recent (last 2 years) issues of the publication SIGMOD RECORD in
should you wish to borrow any.
- PODS. Principles Of Database Systems. This conference is
partly sponsored by and held concurrently with the SIGMOD conference.
- IEEE. Institute of Electrical and Electronics Engineers.
organization sponsors many database-related conferences and bulletins.
- FOCS. Foundations of Computer Science. Not completely
but still useful. You can find info from the IEEE link above.
- ICDE. International Conference on Data Engineering. This is
IEEE conference. http://www.ieee.org
Know the (real) sources
The "official" organizations above are useful for finding the
a work: complete author list, publisher, abstract, etc. These are all
usually free. This
is particularly useful when you're not sure if the paper will have the
info you're looking for,
and all you need is the abstract to find out. But for getting the
entire text, most of
these sites will charge, and you shouldn't have to pay. Typing in the
entire title to
Google is a hit-and-miss proposition. So here are more reliable ways of
a full paper:
- ACM digital library. Union College pays for its students
to have access to this. It's great for all ACM (Association for
sponsored publications, which includes SIGMOD and PODS. Access it
through our library.
- Citeseer. This digital library focuses
on scientific literature. It allows you to find full
bibliographic references and the references cited by a particular work
It will also list full URLs of where the complete work can be found.
- DBLP bibliography. This Computer Science bibliography lets
you do title and author searches. It also provides links to author home
who often post their most recent publications for download.
- Visit our library. If you have the full reference to a work
having trouble finding it, our library has many catalogs and indices
JSTOR to help you
out. And people who will gladly show you how to use them. Stop in.
- Email the author. Academic-type folks love to share (and
their own work. If all else fails, find the author's email address (by
using DBLP to
find their homepage, for example), tell the author you're a student
the work, and ask if you can have a copy. They'll be more than happy to
Know when to move on
It is important to be thorough when doing a literature review, but
that it is impossible to actually "finish" reading the literature about
a topic. There
will always be another paper you can read. Connections to other topics
are incessant. And new works are coming all the time. It's important
be able to "draw the line" at some point and move from active reading to
coalescing, writing, and presenting.
Thanks to Dianne Prost O'Leary's page on
study, from which several of the phrasings above draw their
Back to Project page