Some Thoughts About Writing of Journal Papers

Abstract

This editorial tries to stress the importance of a well structured approach to the writing of research papers, and especially of journal papers, which represents the crown of each and every publication effort. This is done through discussion of a research methodology being taught and used at the University of Belgrade, in the Department of Computer Engineering, and elsewhere. Both, when planing a new research activity, or when writing about the results of an already finalized research activity, this methodology insists that the following elements must be stressed, as precisely as possible, and as briefly as possible: (a) General research field that the research under consideration belongs to; (b) Specific problem attacked by the research, and why is that problem important; (c) A short survey of existing solutions of the same problem, and what are their drawbacks, under conditions of interest for the research under consideration, and its sponsor; (d) Proposed solution, its essence, and why is it expected to be better, compared to the best solution from the open literature; (e) What has been done, or what will be done, to demonstrate that the proposed solution is better—mathematical modeling, simulation analysis, prototype implementation, or a combination thereof; (f) If the research resulted only in research papers, or an industrial product has-been/will-be developed; (g) What are the errors made and lessons learned.

Whenever the author of this editorial gets assigned a duty to serve as a member of a thesis evaluation committee, he always asks the same seven questions, and he always insists that the answers fit into one (possibly complex) sentence (seven sentences for seven answers).

If the seven answer-sentences are essence-pointing and crisp—that is a good sign of a good research. If the candidate-student is ambiguous and starts beating around the bushes—that is a good sign of something going wrong (with the major professor—not with the graduate student).

The seven sentence-questions are defined and commented in the seven paragraphs to follow. The questions are based on the formal logic, and as such widely applicable. Of course, certain minor variations are possible, depending on the specific topic and general circumstances.

Question #1: What is the general research field of the research under consideration? This is definitely the easiest question to answer, but not necessarily the easiest one to answer precisely. What is needed here is the lower bound answer—an answer which encompasses the research as tightly as possible. An upper bound answer is often an indication that not enough attention was given to details of the work.

Question #2: What is the problem statement (including the conditions and assumptions of interest), and why is that problem important (so audience knows that the candidate was not wasting time with something irrelevant)? Experience teaches that a narrowly defined problem statement is more likely to result in useful results, which are later more likely to be used and referenced. A broadly defined problem statement is more likely to result in some concept or system in which it is often difficult to distinguish between contributions of the others and contributions of the candidate.

Question #3: What are the existing solutions of the problem, and what is not liked about these solutions, from the point of view of interest here (which is supposed to be well defined through the answer to the second question)? The answer to this question (#3) should demonstrate that the candidate is fully aware of other existing approaches. On one hand, some people say that there are two types of researchers—one which reads papers and one which writes papers. On the other hand, some other people say that knowing too much about the existing approaches destroys the creativity—one takes the well-walked paths, rather than trying his/hers own ones. Of course, extreme solutions are rarely the optimal solutions.

Question #4: What is the essence of the proposed solution, and why it is expected to be better compared to the best one from the open literature, in conditions of interest for the research under consideration? Here, the candidate is expected only to stress the logical or philosophical essence, without going into details. It is extremely difficult to come up with something which is universally better; however, the more narrow the set of initial conditions, the more likely it is to come up with something that really represents a step for state-of-the-art. Of course, the criteria of importance are the major factor which determines if and how much is the proposed solution better.

Question #5: What has been done to demonstrate that (under the given conditions and assumptions) the proposed solution is really better (for the given criteria)? The candidate is expected here to specify if the research was based on analysis, and/or simulations, and/or implementations, or a combination thereof. In engineering, performance (e.g., speed) and complexity (e.g., price) are inherently correlated, and the answer must refer to both (performance and complexity of something which is real are like soul and body of something which is alive).

Question #6: Was all that just an academic research, or a real product for a real market was generated? Implementation for market should be the crown of each research in engineering. This author believes that solid research must include an implementation phase, at least to show potential bottlenecks of a later industrial implementation (not to mention the fact that real pleasure comes after one can see that something is up and running—made and it works).

Question #7: What are the lessons learned and errors made? This author was taught that no research and development is completed if all that is done is that something is made and it works. The work is fully completed only after one has a chance to step back and think about all experiences gained during the work. Awareness of errors made and lessons learned (plus the ability and willingness to talk about them and to analyze them) is a higher-level achievement—that is what each graduate student should know. The student's answer to this (last) question tells if the student really understands what he/she was doing.

The bottom line is that when writing a paper, and especially a journal paper, all seven issues specified above have to be addresses explicitely, both in the abstract of the paper, and in the paper itself.

Papers written according to the guidelines specified in this editorial move faster through the reviewing process. After being published, more people read them (because they are easier to read and comprehend), and more people reference them, on average (because potentially good work is not hidden behind a not so good presentation).

Veljko Milutinovic
University of Belgrade