How to Write a Good Journal Paper Abhijit A. Gosavi, Ph.D. Missouri S & T Outline • • • • Doing the research Writing the paper Potential revisions Some final thoughts Disclaimer • • • • • What I am going to present is what I have learned from my teachers, coauthors, and from personal experience; and also from helpful anonymous reviewers: there are many people with generous hearts who are very helpful when they realize you are a new author This presentation should not be considered to be a magic pill, since I don’t have “ready solutions.” There is no such thing as a ready solution here Your advisor(s) should be able to provide you guidance in many ways. Always seek advice from well-published professors in your department or university. As a matter of fact, although I am very happy to do this presentation, there are many people in my community, who are far, far more successful than I am and have more useful pearls of wisdom. Do not listen to folks who tell you that getting a journal paper published in a good journal is all about who you know, where you got your Ph.D. from, and whether your advisor is on the editorial board of the journal – rather than about the quality of your research. Fortunately, good research gets published in good journals (at least, I believe this to be true). Doing the research • Identifying a problem that has not been solved (perhaps obvious!) • In order to get your paper accepted to a good journal, some acceptable mechanisms for research are: 1. Solving a problem that has been solved before by others, but you want to use a new method and show that the new method has advantages over existing methods 2. Solve a new problem that has not been solved before (sounds easy, but chances are that when you look hard, someone else has solved a similar or related problem). Going to the industry to hunt for an unsolved problem has led to many a path-breaking paper. Do note however that if it is too trivial a problem, it may be hard to get that paper accepted. 3. You have a hypothesis, and you want to test it. Clearly, it should be an interesting hypothesis, i.e., of interest to your academic/industrial community. • Bottom line: it is important to identify a gap in the literature, a gap worth filling, and ensure that you are filling it. Identifying the problem • Identifying the problem involves doing a lit. survey, and making sure you aren’t reinventing the wheel. • In the initial stages, when you know what problem interests you, it is necessary to read widely. Start with conference papers (they’re usually easy to read) and then move on to the seminal journal papers (sometimes this kind of material can be found in advanced books). But reading the relevant papers is critical in the end! • After you’ve identified a problem, you must work hard What does that mean? What does the hard work involve? 1. You must test your methods extensively on the problem domain. Wherever possible, benchmark it with existing methods. If it’s a new problem/method, make sure that your method generates the optimal solution, or a useful solution if reaching optimality is not feasible. 2. If you are building a new management strategy or proving/disproving a hypothesis, you must collect data, and lots of it. How much is enough? That depends on the field. In our field, mgmt and systems, 10-20 pieces of data don’t a good paper make! Try to collect voluminous amounts of data (e.g., exceeding 100 samples for each hypothesis). Remember if you are in the medical world, even a few thousand samples do not lead to a respectable study. Make sure when you are collecting data that it will be useful later. There are times when we can’t go back to the experiments. Hence a thorough prior understanding of the statistical tests that will be performed later is a must! Writing the paper • Doing the work is only half the work; writing it clearly is the other half. • Make sure you follow a proper format: Abstract, Introduction, Lit. Review, Body of Your Research (Method/Problem/Hypothesis), Numerical Results, and Conclusions. Abstract • The abstract should not be copied from the introduction, i.e., avoid repetition. • Should be of the proper length (author guidelines will tell you the word limit) • The abstract has to highlight the main contribution. • The abstract should be rather general in the beginning (introducing the broad field in one/two sentences). • Should provide an overview of the paper Introduction • Needs to be an expanded version of the abstract • Some authors also include parts of the lit. review in the introduction. • Remember you aren’t writing a murder mystery! Clearly tell the author, without providing all the details but in broad strokes, what is coming in the paper. • In the last subsection of the intro., you can provide a roadmap to the rest of the paper (explaining what each section of the remainder of the paper contains). Lit. Review • • • • • • Make sure you begin by citing some main references, e.g., seminal papers or textbooks that contain the fundamental theory for your specific topic. Then, introduce the relevant papers using a specific style of arrangement: either chronological (safest, so you don’t offend a potential reviewer who believes his/her research is “more important” but cited later for no obvious reason!) or based on areas within the area you are working on. Your lit. review should pave the way towards showing what the gap in the literature is! This is the most important thing about how to organize the lit review. Don’t forget to include the most highly cited papers in that area if they are relevant. Remember with a high probability, one of the people you cite will be your reviewers (who are almost always blind to you, though you may or may not be blind to them). However, don’t cite someone whose work is not related just because you hope to have this person review your paper. That can backfire because the editor is very likely an expert in the field and can see right through that. Lit. review can cause rejection! • Papers are regularly rejected because they failed to cite a closely related work. In fact, if you know of some work that is closely related, you must explicitly cite it and explain how your work differs from it. • I am a regular reviewer for quite a few journals in my field, and I see that inexperienced authors oftentimes fail to describe how their work differs from existing work and what gap it is that they are filling. This is grounds for rejection, or a major revision, because the reviewer is confused about what the contribution is. Remember not all reviewers are people working in the closely related area, nor are they going to necessarily read the closely related papers. So if they are in doubt, they may reject. Furthermore, if you don’t clarify that doubt, they will most certainly see that as a weakness in the paper. • Citing conference proceedings as seminal work is usually a poor idea! Most seminal work eventually makes it to journals, and so take the effort to find it. (There are some exceptions to this rule, e.g., Widrow-Hoff rule paper or Littlewood’s rule paper from my field.) Body • This is the most important section in your paper. • Make sure you present the notation properly. Don’t use a symbol before defining it. That can be aggravating to the reviewer. • You can introduce the notation as you go along; a glossary with bulleted items at the very beginning is also fine (but I personally don’t like it, because at that stage the reader is not familiar with most terms). • Your algorithm/method/problem domain/hypothesis must be clearly defined. Do not assume the reader knows it already; the reader is not trying to solve a murder mystery! • A picture is worth a thousand words only if it isn’t confusing! I have seen pictures that make no sense; if you have a co-author, ask his/her opinion about it. Pictures should be convey something enlightening to the reader at first sight, but what they depict must be also explained within the text (for instance, Fig. 1 represents so-and-so). • Organize this part into multiple sections/sub-sections as needed. Some comments on writing • Organize your material into short paragraphs. But how short is acceptable? • One or two sentence paragraphs (paras.) are usually rare in technical papers. They are used in newspapers. • However, very long paragraphs can make the paper boring to read. If you have a long para. split it into multiple paras. Note that each para. should be limited to one idea. • Don’t start a sentence with a symbol if it is in the lowercase. “α denotes so-and-so” does not read well. Instead say: “The symbol α denotes so-and-so.” Numerical results • Provide ALL the inputs needed for perform the experiments you performed. • Preferably use tables to show the inputs. • Show all the relevant outputs that form the crux of your research. Use tables and also graphs if possible. • Write this section clearly, because in some sense this is examined very closely. It is usually tied to your major conclusions and the gap you are trying to fill. Revisions • When you hear back from the editor, they will tell you one of the four things: 1. Accept as is (rarely, but this is the dream scenario) 2. Minor revision (sometimes) 3. Major revision (usually) 4. Reject (you don’t want this) Major revision is nothing to get too worried about, but it means you will probably need significant work, e.g., sometimes re-running your experiments and /or making changes to your model. Sometimes, it is just a matter of writing it better. For 2 and 3, you must submit a revised paper. It has to be accompanied by a detailed report showing how you’ve revised the paper. Address every comment thoughtfully and in detail. You don’t have to agree with every comment, but you must explain (respectfully) why you disagree. Reasonable reviewers usually buy your argument provided they think it is valid. The onus is on you to change their mind. If your paper is rejected, please don’t send a nasty email to the editor complaining! Unless you feel that the reviewers are being unethical or extremely unreasonable! Dealing with a rejection or harsh comments in a major revision • Failure (i.e., rejection) is a part of the game. What is important is to learn from it, i.e., carefully read the comments provided. About 80% of the time, I have gotten very useful comments in rejections that led to significantly improved research (that got accepted in a lower-tier journal with some additional effort). 20% of the time, I have gotten twosentence rejections (after a whole year – in the OR community), which have to treated as plain bad luck (yes, there is evil out there ), and one has to move on! • When depression or panic sets in, you forget how to be creative. Remember: creativity is at the heart of what we do; if we lose that, we lose everything! If you think your research is worthwhile (and you would not have started working on it otherwise in the first place!), with additional work on that paper (sometimes without it), you should be able to get that paper published – most likely in a lower-tier journal. Dealing with rejection (contd.) • Develop a thick skin Easier said than done, of course, but there is no other way to put it. One of my most cited papers in IIE Transactions (which presented an algorithm that has been used by others) was rejected at a top journal, and the reviewers there (unanimously) thought that there was nothing useful in my paper Another time, a reviewer said: “This paper should not be published in ANY journal.” No human is qualified to say that Dealing with a rejection (contd.) • Don’t believe everything the reviewers say, but be objective with your self-evaluation: learn to reject the unfair criticism and accept the fair criticism. • Finally: some human relationships don’t work. If we refuse to accept it, we waste time. Similarly some papers have to be buried because the research is not publishable (i.e., they go into a drawer possibly never to come out). Hopefully, something like that will not happen when you are working on your dissertation, because your advisor (who should have the experience) will be able to foresee such an event and prevent it from happening. Some Final Thoughts • • • • • • • Don’t select a conference because of its location; rather go to a conference (even if it is Cleveland, OH and not Paris, France!) if it is the main conference in your field. Chances are that you will be able to hear some of the leading authors in your field speak, which is very valuable because you may get insights from their talk that are not visible from reading their published papers. Read the seminal papers in your area (chances are they are old and not available online) even if they are not directly related to your research and even if that entails actually going to the library. Identify the best journals in your field, and try to identify the characteristics of the most cited papers from there. Aim your research at the best journal in your field (not the lowest one with a very high acceptance rate, e.g., one where you pay; all papers in the journal are opensource etc.). Even if your work doesn’t end up in the best journal, try and do good research that will get cited as time passes! Your advisor should be able to guide you to a suitable journal (this is not something you usually do as a grad. student). Finally: May you be showered with success!