Prof. Anna's Blog

This question is a real evergreen and seems really hard to answer (it is hard). I typically say: aim high in quality, do not overdo it in quantity, and do not waste anything by not publishing it. I am lying sick in a hotel room in Austria and I am bored. So, I decided to explore my own citations from the last 7 years.

Here is what I did: I listed all peer-reviewed publications where I am a co-author, and noted for each paper: number of current citations, year of publication, kind of paper (workshop/poster/demo/phd forum; small conference / small journal; top conference / top journal; book chapter; technical report or Arxiv), off-site (journals, TR, etc.) or on-site (workshop, conference, demo, etc.) whether this was a cooperative effort from outside my group, and whether it was a survey or not. I first calculated the number of citations per year to normalise for time. Then, I calculated the correlation factors between each of the factors described above and the number of citations per year.

Here is what came out. First, the correlation is clearly positive for top conferences / top journals (0.3) and even more for surveys (0.5). There is also a positive correlation between cooperative efforts and citations per year. For the other factors, the correlation factors are towards zero or even negative. What does this mean? It actually confirms what I also said before, but now it is scientifically founded 🙂

1. Focus on top conferences and top journals. It takes some more time to go through the review rounds, the review process takes time, and you might need to re-submit several times before finding a good “home” for your paper. But it is worth it; keep going and do not give up! You might also need to work on your scientific writing, on your experiments and methods.
2. A survey is a great way to start your publication record and publish results from literature studies. Do it!
3. Do cooperate with people outside the lab. This is a proven method to increase your visibility, your network and your citation record. This lies simply in the fact that such papers have more authors from more labs, and all these people will take care that the paper is cited and propagated.
4. Posters/demos/small conferences are a great way to present your work, get feedback and increase your network. Here, the focus is NOT on citations. However, growing your network and discussing your ongoing work is VERY important, not only for young researchers.
5. Now, I need to cope with the results from non-top journals. It seems like those are not really citation runners. However, some universities require their PhD candidates to have at least X papers in journals, which is not easy to achieve in the short time with top journals. Thus, you can try to diversify: one-two top journals and the rest in conferences and non-top journals. If this requirement is not important for you, then focus on top journals and top conferences only (but you can submit also demos / posters to those top conferences, the network and the colleagues are important, not how long the paper is!)

I hope this helps somehow to build a proper publication strategy. In the beginning, focus on a survey and posters / demos /workshops / conferences to get feedback asap. Then, go for top journal publications. This will require also some time and task management. Think about how to schedule your work. You should identify several bigger tasks (2-3) you need to address for your PhD. Then, you should focus on the first and follow the publication strategy above. Once the work is in the review loop, start with the second task, and so on. They need to overlap while you are waiting for reviews and juggling with different papers in different (top) journals. Good luck!

I get this question all the time – Prof, how should I start? Prof, how do I write an article?

Well, the short answer is: simply start writing it, word by word, sentence by sentence. Ok, I know, it is not that simple 🙂

There are many ways of writing and it depends greatly also on the writing process which your supervisor has set in place (I do assume here that most readers are beginning Ph.D. students. If you are a postdoc and you still do not know how to write articles – well, you are in big trouble…). I personally like to be mid-involved in writing articles. I almost never contribute real text or graphs to the articles of my students, but I do give detailed comments early on. This is the process in my group, which seems to work quite nicely:

1. Define a table of contents. Do not limit yourself to section titles, but put 1-2 sentences in each section to say what you plan to put there. Be as concrete as possible. For example:
   

   “Section 4: Experimental results.
   4.1. Experimental setup
   Put here the complete description of our setup, incl. hardware, software, and a picture of the hardware.
   4.2. Metrics
   List and explain our metrics – throughput, delay, energy consumption.
   4.3. Results
   Here they come. Probably 3 graphs, one for each metric. Discuss each of them in detail.
   4.4. Discussion
   Go back to our motivation and overall problem statement, discuss whether we solved the problem and which disadvantages and open issues we see.”

   See, the point is to plan everything in great detail. If instead you simply say: “Section 4: Experimental results.”, your supervisor cannot judge whether this is enough or you are missing something important. For example, I could now complain that the metrics are too few or wrong, or more graphs are needed.

2. Start writing the individual sections down. Make a first complete draft of the complete manuscript, before giving it back to your supervisor for comments. However, it often happens that while you are writing, you realize you miss something in your overall structure or that the structure needs to change. Explain the issue to your supervisor, make a suggestion of how to handle it, and discuss it. But do not send the non-ready manuscript! This could only lead to confusion and misunderstandings because you did not check it yet.
3. Now you have a draft. What is a draft? It is not a ready paper – some individual graphs can be still missing (for a good reason only! For example, waiting for the experiment to finish), and there might be grammar errors or formatting errors. However, try to do your best in terms of wording and explanations. Give your supervisor enough time to read it carefully and to give you feedback. If you only give her 6 hours before your meeting, she will be angry and/or have not read your paper at all. Enough time means several days.
4. Your supervisor will have comments, some of them deep-going. She might suggest even new experiments, new sections, new structures. You can mostly prevent this with a well-prepared and detailed table of contents (see point 1). But you can never fully prevent it and you must be open to these suggestions. You can openly discuss problems connected with such changes – how much more time you would need, how to implement the changes, etc. But try not to take the position: “I do not want to do any more work, this is enough!” This would not be very good-researcher-like 🙂
5. Once you agree on the contents of the paper, it is time to polish: check the grammar, check the references for completeness, check the format, check that all acronyms have been introduced properly, etc.

That’s it. I guess the most critical question is now: “But can I submit also smaller parts to my supervisor to comment on, just to make sure that I am heading in the right direction?”. This is a tough question. Of course, you can try, and your supervisor will always give you feedback – but how useful will it be really? Unready chunks of text tend to be isolated from important context, to have bad wording, and to lack necessary details. The comments will be like “Use a spell-checker”, “Work this out”, or “I do not understand this part”.

It would be better to train your writing skills before you start your first paper. If you do not feel confident in your writing skills, agree with your supervisor on a plan: you could write small summaries of papers, you could write short proposals about experiments, etc. Keep those short (1-2 pages) and always do your best before submitting them to your supervisor.

In fact, I really enjoy writing myself (this blog is a proof of this!). I always see it as THE opportunity to describe what great research we have done, to present it nicely and in an understandable way, to “show off”. I like fine-tuning graphs and figures and text. This is my representation to the outside world. I need to be proud of it. I never see this as a waste of time but as part of the process. Try to adopt this attitude and it will make writing much easier and even fun!

Happy writing!