For a new assistant professor in CS, how to build/manage a publication pipeline

Question

I am a new assistant professor in a research institute, where the number of top-tier publications matter a lot. In short, they simply count the number of tier-one papers.

I am highly advised to work out a good publication pipeline, in terms of how many papers are in progress (e.g., implementation), how many papers are under review, and also how many are in the preliminary stage to investigate the feasibility.

So besides "working hard", what are some tips, comments and advices on start to setting up a "publication pipeline"? To concretize a bit, what I can come up with are:

1. has one or two main research lines such that you can constantly publish your major research output towards some really prestigious conferences in your field, in CS it's like SIGGRAH, OSDI, POPL and so on. But of course, usually preparing such a work takes a very long time; my personally experience is about 1~1.5 year, when I was a Ph.D. student.

2. Simultaneously, become versatile in terms of skillset and therefore can quickly hunt for some (low-hanging) fruit. This can be much shorter, say 3~5 months, can still target on tier-one conference, but maybe not that "prestigious" ones. I am not going to name such conferences to avoid some arguments here...

3. What else?

All I can find is a thread here (https://www.chronicle.com/forums/index.php?topic=38427.0), but the message is a bit unclear to me. Any suggestion or advice would be appreciated a lot. Thanks.

Answer 1

Here two pieces of advice:

• As a mid/long term perspective you should build a large network with bright people. In the beginning talk to as many as possible. Tell them about your ideas and ask them about theirs. This will lead to a lot of collaborative papers. Don't waste too much time with people who are reluctant. Most people will be very open (especially the younger ones).
• Do this also with people who are not working in exactly your field. There might be a lot of low hanging fruits to collect i.e. something that is easy for you to do but not for them or vice versa. There is hardly any type of research field that would not like to get some input from CS (buzzword "data science"). e.g. Experimental biologists here or people working in business or geography etc. This will not lead to first-authorships but possibly get your name on many papers by putting only few days of work in.

Answer 2

So besides "working hard", what are some tips, comments and advices on start to setting up a "publication pipeline"?

As Lordy's answer points out, the key to a regular stream of publications is a healthy network of collaborators: external collaborators but also the students or postdocs that you supervise and consequenly who follow your research agenda. So to some extent a sustainable publication pipeline depends on maintaining a pool of PhD students or postdocs working with you. This usually depends on you getting some funding to pay them, by submitting applications to the appropriate funding bodies in your domain.

So the standard strategy goes like this:

1. Follow the calls in your domain and submit applications regularly in order to ensure a stream of funding for the next years
2. Fund some PhD students and/or postdocs on the grants awarded to you
3. They follow your research agenda, carry out most of the exploratory work under your supervision and you co-author their papers

Answer 3

While collaboration is important, I would caution you to get too hung up on this. You want to avoid being the person that just hangs around in the middle of the publication list in many papers - in my experience people eventually develop a bad taste towards scientists that they perceive to be freeriders on other's top research.

Instead, in my experience the most important key to having a good pipeline, especially if you are in one of the more applied CS fields, is to have a clear research programme. If you have, say, three PhD students, try to make sure that there are synergies between their work. In the ideal case, none of your future students past the first one or two should start from a completely green field - develop a portfolio of prototypes, methods, and data sets that you and your students can build on in the future. In my experience, this drastically cuts down on the time needed to write an A+ paper - if you start from a completely green field, it can easily take a year or two to collect enough material to have a shot, but if most of the scaffolding (knowledge- and technology-wise) is already there, I have seen people churn out excellent papers in surprising small time. This also has the advantage that your different research strands will eventually build up to something larger than individual papers, which ultimately ends up more important in tenure and promotion evaluations than the pure number of papers.