Hints for written work
- Make sure you give references to the literature you use -- the reader may
want to learn more, or may not believe what you have said! The exact
style of the references is not important (although you should be
consistent). What is important is that you have at least the following
Conference Papers: author(s), title, conference name, page numbers, publisher, year.
Journal Papers: author(s), title, journal name, journal volume, journal issue (if there is one), page numbers, year. Note: issue number always comes after volume number. E.g. 3(4) = volume 3, issue 4.
See the departmental publications repository for examples. Note that other types of publications (e.g. books, technical reports, theses) have slightly different requirements. Essentially, the reader should be able to find the paper given the information you have provided.
There is a difference between giving a reference and quoting. References have the following form:
"the following definition/theorem/algorithm/idea was given in [XYZ]; here is a summary in my own words".
Everything in your report that is or could have been the result of cut-and-paste is a quotation
and should be clearly identified as such:
- if you use a figure from another document, add "(from [XYZ, fig.5])" to the caption;
- the same method applies to other items with captions, e.g. tables or algorithms;
- if you reproduce a figure with small changes, add "(modified from [XYZ, fig.5])";
- if you use one or two sentences from another source, enclose them in quotation marks and include the page number [XYZ, p.123];
- if you use more than two sentences, format it as a separate indented paragraph and add quotation marks and exact reference.
You should try to use quotations sparingly. First of all, we ask you to write reports in order to test your understanding, and quotations don't really allow that. Secondly, there is usually some mismatch between your terminology, notation, and level of detail and that of the referenced paper, which means that summarising it in your own words is almost always better. As for figures, ask yourself: "does this figure convey exactly the right message?" If not, draw your own -- this shouldn't take too much time. It is also easy to cut and paste in the wrong format, e.g. a 72 dpi bitmap may look good on screen but not on paper. What you should definitely not cut and paste are formulas.
In summary, it is essential that you reference everything -- text, figures, programs, ideas -- that you use in your reports but did not create yourself; and that you clearly indicate literal quotations. Otherwise you are guilty of plagiarism and the consequences can be very severe.
- Make sure you introduce any notation you use. E.g. ``My knowledge
of machine learning (k) increases as follows: k = p**2, where p is the
number of papers I have read.'' Also note that for each paper there is
an appropriate level of notation and that using too many symbols and
equations can make things more difficult for the reader: sometimes it
is better to say it in English. On the other hand, formal statements
often simplify things: sometimes it is better to say it
- Put your own perspective on things. We would like to hear
something about what you think, not just a summary of the paper. Note
that it is ok to criticise or disagree with the paper. E.g.:
- ``I don't think this part of the paper was written in a clear way.''
- ``I don't know why the author said...'' or ``I don't understand this part...''
- ``I think the author's conclusions are incorrect because he has not considered...''
- ``An interesting extension/application of this work would be...''
- ``A better way to do this would be...''
- ``I thought of another example of this, which is...''
- ``I found another paper which disagrees with this one regarding...''
- ``For me, the most interesting part of the paper was...''
- ``The most important thing I learned was ...''
Normally written work is more formal than oral presentations. (And a dissertation or published paper is more formal than the reports you are doing here.) With more formal work, you should avoid subjective statements like ``I think that...''. See published papers for the style you should use with your dissertation.
- Don't try to impress the reader by making things seem
complicated. A good scientific paper should make things seem simple
(as much as possible). This takes a lot of effort, and not everyone
can do it well.
- Introduce and define any unusual technical terms. Spell out
all acronyms the first time you use them
(e.g. Machine Learning (ML), Inductive Logic Programming (ILP),
Receiver Operating Characteristic (ROC)).
- Use numbered sections (and, optionally, subsections) in your
report. Some software will do this automatically. You may wish to
outline the structure of your report at the end of the introduction:
``The remainder of this report is structured as follows: section 2
provides xxx, section 3 outlines xxx, and section 4 discusses xxx.''
- It's better not to write on things you don't understand. Don't try
to fake it!
- You may want to get some tips on writing style from a little book
called ``The Elements of Style'' by William Strunk Jr. and E.B. White.
Published by Allyn and Bacon.
- Proofread your work before submitting it, preferably after not
working on it for a few days. Try to put yourself in the reader's position.
Does it make sense? Is any information assumed and not actually given?
- Run a spell check!