End of Term! (and how I read a paper…)

It’s been a real pleasure interacting with so many of you that read The Morning Paper. Some of you have told me that you read it in the morning, but others save it for lunchtime, or even to wind down after work. Some forward it around their teams to encourage broader awareness, and many of you save editions away to read later. However you read it, one theme I hear a lot is ‘it’s hard to keep up, I’ve got a backlog saved up to read…’. So here’s your chance! The Morning Paper will be taking a two-week break starting today for the Easter holidays. Make the most of it because normal service will resume on Monday 13th April and there’s lots of good stuff to come: I’m working through some of the papers from ASPLOS ’15 to bring you the conference highlights; we’ll have some more Desert Island Papers guests; and of course the usual mix of important/influential/interesting papers from the CS back-catalog together with the latest research.

During these two weeks I’m going to take some time to go back over some of the papers we’ve covered in recent months to pull out high-level themes and start thinking about my forthcoming keynotes at JAX and NoSQLMatters. I’ve also got a small backlog of books to work through: 😉

Books

I hope you’ve been enjoying The Morning Paper – I love hearing your feedback so if you’ve discovered or (re-discovered) a favourite paper, found something that’s really relevant to your work, or just enjoyed getting a curated mix of papers in your inbox do please let me know: blog comments, twitter, even email, all work! And thank you for all the retweets and favourites of the paper announce tweets too, they all help to keep my motivation high. If you know someone you think would enjoy The Morning Paper, do please help to spread the word!

Finally, this seems a good opportunity to address a couple of frequently asked questions: how long does it take me to review a paper? and, any tips for getting better at reading papers?

How long does it take me to review a paper?

Each edition of The Morning Paper is between 2 and 4 hours work, most commonly somewhere around the 2.5hr mark. My initial read of a paper typically takes 45 minutes or more (depending on paper depth and length), the write-up takes me 1-1.5hrs, and scheduling the content on the blog and mailing list, plus putting together the announce tweets takes another 15 minutes or so. Not counted in this is the period of time I spend between the initial reading and starting on the write-up, when I mull over what I’ve read.

The goal of all this is to end up with something that you can sit down and read in 20 minutes or less. And if the ideas catch your imagination, I always recommend going on to read the original paper (I put the link at the top of each post) – of necessity I have to be selective in the parts I choose to highlight so there’s always a lot more to discover. Even if you don’t go on to read the original though, my goal is to share with you the big ideas from the paper and bring them to your awareness so that you can start to build up a picture of what’s out there over time.

Any tips on reading papers?

There are several ‘how to read a paper’ papers and articles on the internet full of good hints and tips. Here’s the way that I do it:

  • I use Mendeley to manage my paper collection. I started out just using a notebook in Evernote but found that Mendeley’s paper-specific support came in very useful once my collection started to grow. I mostly use the Mendelely desktop application for managing my paper collection, but I use the iPad app for reading.

  • For a one-off paper review, I still think you can’t beat printing it off and marking it up by hand. But for me there is a lot of value in having all of my annotations stored electronically. So the next-best thing that I have discovered so far is to read and mark-up papers using the Mendeley iPad app. There are two key features that I think are important: (a) reading on the iPad, the paper occupies the whole device, there are no other icons or distractions, just you and the paper, and (b) whatever you use, you want good pdf mark-up tools.

  • I read the paper straight through the first-time, and highlight what seem to be key passages as I go through. I also sometimes make notes in the margin (using the note tool) to capture my own thoughts, or related work and ideas that come to mind. Mostly I highlight the paper text though. I highlight the main ideas, supporting details that reinforce key points, and anything I think is important in telling the story of the paper.

  • Then I might quickly go back over the abstract, introduction, related work, and conclusion sections. It’s here that you’ll find the ‘big picture’ of the paper in the author’s own words, as well as material that helps set it in context.

  • At this point I put the paper down. Typically I read a paper in the morning (that’s where the blog title comes from!), and then write it up either the same evening, or a couple of days later.

  • I let the ideas in the paper swirl around in the back of my mind. Sometimes consciously, sometimes unconsciously – but this ‘thinking time’ definitely seems to help me. If I’ve read a paper on the train into work for example, I’ll think about it while walking from the station to the office (about 20 mins) – after which it will be relegated to my subconscious.

  • Later on when I come back to the paper, I read through all of my highlighted sections – often dipping back into other parts of the paper as well, and start to form in my mind the key ideas and story.

  • Now it’s time to write! I frequently use Editorial on my iPad for this (with a bluetooth keyboard of course) as I’m out and about (on the train, in a coffee shop, …). I also like the full-screen distraction-free writing environment it creates.

  • I’ll sketch out the main points I want to cover (section headings, or just placeholder sentences), after which I just start at the beginning and get writing! The process of telling the paper story in my own words really helps me to engage with the material – forcing me to think critically about it, to extract the key points, and to make sure I really understand it. I’ll often find I need to go back over parts of the paper several times as I do this as the understanding I thought I had turns out not to be deep enough. I’m not simply writing down a story that was pre-formed in my mind, the act of writing forces me to engage with the paper in a deeper way and this frequently takes my level of understanding and insight to the next level. Even if you’re not blogging, I definitely recommend making some notes or a short summary for yourself on papers you read because this act itself is part of the understanding process.

  • Finally, the more you do it, the easier it gets! Quickly extracting the essence of a paper and separating the wheat from the chaff is a skill you can develop over time. I’m better at it than I used to be, and I’m sure there are plently of people out there who are better than me too! Plus of course, the more you’ve read in a certain area, the more familiar the foundational ideas and related work becomes, helping you to quickly put ideas in context.

Bear in mind my goal is to capture the big ideas and set the work in context. There’s a whole other level of understanding beyond that I’ve outlined above, in which you engage with the material at a deeper level again by recreating proofs and arguments, and implementing key algorithms etc. This might take you a week or more for a single paper and thus has to be done selectively for the key works that are central to whatever it is you are currently working on.