We want reports that copy ideas from papers, and which cite the original paper to give them credit for the idea. If you do not cite the original paper, you are plagiarising it.
We do not want you to copy text from papers. If you feel it is important to copy the exact sentence(s) used in the original paper, you can put it in quotation marks and cite the original paper. However, this does not require any understanding so it does not get you many marks.
If you copy the text from a paper but do not put it in quotation marks and cite it, that is plagiarism.
To repeat: we want you to copy the ideas (and cite the author of the ideas) but not the words.
We want you to write things in your own words because we want to see if you understand the material. If you use other people's words we can't tell whether you understand the material.
Here are some examples of what you should and should not do. This is part of a paper:
XCS introduced a number of innovations, foremost among them its accuracy-based fitness under which rule fitness is related to its classification accuracy and not the magnitude of the reward it receives as in earlier systems. For lack of space we do not include the details of the XCS updates, but suffice it to say that XCS evaluates the prediction and fitness of each rule. Prediction is, for concept learning tasks such as those we study here, an estimate of the proportion of inputs matched by the rule which belong to the positive class. Prediction is used in conflict resolution, when matching rules perform a weighted vote on the classification of a data point. Accuracy is a measure of the consistency of prediction. Rules with prediction near the maximum or minimum value have high fitness. Higher fitness rules are allocated more reproductive opportunities by the genetic algorithm in XCS, and fitness is also factored into the classification vote.
We can cite the paper in a bibliography like this:
[1] Tim Kovacs and Manfred Kerber. High classification accuracy does not imply effective genetic search. Proceedings of the Genetic and Evolutionary Computation Conference (GECCO 2004), pages 785-796, Springer, 2004.
If you want to cut-and-paste some exact text from the paper, you should quote and cite it:
Kovacs and Kerber said: ``XCS introduced a number of innovations, foremost among them its accuracy-based fitness under which rule fitness is related to its classification accuracy and not the magnitude of the reward it receives as in earlier systems'' [1] p. 2.
However, as I said you should avoid using quotations in your report. We much prefer you to write things in your own words. For example, it would be ok to write:
XCS made many innovations, of which accuracy-based fitness was the most important [1]. With accuracy-based fitness, a rule's fitness is based on its classification accuracy, not on the magnitude of the reward. XCS evaluates the prediction and fitness of each rule in the system. For concept learning tasks, prediction is an estimate of the proportion of inputs matched by the rule which belong to the positive class.
Notice that although all of the above paragraph is based on the paper we only cite it once, at the end of the first sentence. It is reasonable for the reader to assume that the rest of that paragraph is based on the same cited paper, since the whole paragraph is about the details of the same system.
However, this is not a good paragraph, because it follows the original text very closely and does not show a deep understanding of the subject. It would not get many marks.
It is best to not look at other papers at the time that you are writing your report, unless you need to verify something. In other words, you should read the original papers and make notes on their ideas, and then hide the papers when writing your report. This way, you will copy the ideas (and reference them) but you will not copy the exact text.
If you find you are tempted to copy the words, this may be a sign you may not understand the material very well. (Another reason is that if English is not your first language it can be difficult to say things the way you want.)
The worst case is to write paragraphs like this:
XCS introduced a number of innovations, foremost among them its accuracy-based fitness under which rule fitness is related to its classification accuracy and not the magnitude of the reward. XCS evaluates the prediction and fitness of each rule. Prediction is an estimate of the proportion of inputs matched by the rule which belong to the positive class. Accuracy is a measure of the consistency of prediction. Higher fitness rules are allocated more reproductive opportunities by the genetic algorithm.
That is just a shorter version of the paragraph in the original paper. You can copy that without understanding very much of the original. So it deserves a low mark. More importantly, we also consider it plagiarism because it is an exact copy of parts of the original and there is no reference or quotation marks. If you write reports by cutting-and-pasting parts of the original papers, the least serious consequence is that you will get 0 marks for the coursework.
So: