Adding context to publications: a small experiment.

I’m making a small experiment with my publications – or at least the version that I host here.

It started like this: I think that PDF kinda suck as a science communication medium, so I wanted to add an HTML version of my papers. But after converting some of my LaTeX sources to HTML (thanks to pandoc), and using a little bit of my limited CSS talent, I got something that was OK, but felt underexploited.

Because now, I have the opportunity to add stuff around the published text. I don’t think modifying the text itself would be right, but there are many cases where, due to the length constraints or lack of knowledge at the time, things that I wrote are not quite as clear as I would like, or miss some useful informations. Or where I would simply do things a bit differently know, or have got more recent results that strengthen or weaken some of the conclusions.

In the current publication model, it’s not very easy to add such context. You can cite the old publication in your newer ones and comment from there, of course, but if someone finds the old article first, there is really nothing that make it clear what may or may not have changed since publication. But here, on my own website, I can do what I want !

The experiment

I found a nice post explaining how to easily create “sidenotes” using only CSS, which is great: I don’t want to start messing with Javascript unless absolutely necessary. I was also inspired by (aka: stole from) the work of Arthur Perret, who has some great posts on publishing to multiple formats from text source, and uses similar sidenotes through his blog and even in the web version of his PhD dissertation (all in French).

Sidenotes seem like a really good way to add contextual information. It’s easy to make it absolutely clear that it isn’t part of the original text itself (for instance, with timestamps on each note) and that it doesn’t represent the opinion of all original authors (by signing each note). And it doesn’t distract too much from reading the text, like footnotes would.

The results

I started from the first article I published during my PhD, on artifact segmentation in digital pathology. I simply re-read it and started annotating. I quickly found things that I thought were interesting to add: links to later publications, clarifications, links to relevant source code, etc.

Moving from one publication to the next, I got more convinced that this was a good idea. It makes it a lot easier, for instance, to follow the links between the first paper, the first SNOW paper from ISBI, and the follow-up report that expands on both.

In my early articles, I also didn’t have the habit yet of creating a snapshot repository with code for replicating the results. So I could also point to later versions of the code in some of my repositories, which in some case may make it easier for some to understand exactly what I was doing.

I’ve now gone through all of the publications where I’m first author, up to my latest ESANN 2025 paper.

Conclusions

I really like this format. Not only because I can correct things that I think weren’t quite right when we published the paper in the first place, but also because it allows me, in a way, to personalize a lot more the way I present that research. My publications are obviously not just mine, even when I’m the first author. But here, I can add things in my own words, without pressure, and I think it brings some life into what is often a fairly dry process.