Sunday, 4 August 2013

a new piling system

I write academic papers, and read a lot of other papers. What with hard copy scattered through filing cabinets, and soft copies on a wide variety of websites, it can be hard to keep track of what I’ve read, made notes on, made BiBTeX citation data, and referenced.

Nearly two years ago I had a go at using Mendeley, designed to help with this very task. But for some reason, it never gelled for me, and I gave up using it. Then, when I became an Evernote convert, I fiddled about with using it to store data on the papers: author, title, etc, notes, BibTeX, and links to the papers themselves. But it was clunky work-flow, requiring a lot of hand-effort. Only the note keeping part of it worked well.

Then a colleague raved about Paperpile on his G+ stream. I had a look. It has a similar concept to Mendeley and other such systems, keeping references, supporting tags, good searching, associated notes, and citation output in multiple formats (although I only care about BibTeX). Instead of being a stand-alone application, however, it’s Chrome-based, and uses your Google Drive as a repository to store your papers. It’s currently in Beta, and I thought I’d give it a try. I emailed off for an account, and about a week later I got one.

The first thing I did was import my dusty Mendeley data (with a single button click and a log on), so I started off with a well-populated, if slightly out-of-date, database. I decided to trial it on the paper I had just started writing.

I defined a tag for the new paper. As I was writing, each time I included a reference in the text, I checked if that reference was already in my Paperpile database; if not, I added it. (Since there is usually a fair amount of overlap between bibliographies of papers in a similar domain, this should get faster as the db gets more populated.) Then I tagged it with the paper’s tag, and any other relevant tags.

My current Paperpile db, with 321 entries

To generate the BibTeX, all I needed to do was select the tag, which selected all the references, and then export the whole BibTeX file in one go.

The real benefit showed up when I got comments back on a draft of the paper. One was about a short quotation I had included, and I needed to find the quotation in its original context in order to address the comment. Click the Paperpile tab in Chrome. Type the author’s name into the search box; the paper’s details appear. Click the “view PDF” button. Voila! The paper is on my screen! I can’t think of a faster way to achieve that, without using telepathy.

You can upload in various ways, including PubMed, arXiv, doi, navigating to a website, or just giving it a PDF.  When you upload a PDF directly, the system searches the web to find the right data to populated the various fields. It gets this mostly right, but there have been a couple of odd peculiarities. Still, the system is in Beta. I reported the couple of weird results, and got very fast and helpful responses from the team. (I like Beta testing. I can whinge about software, and feel virtuous about doing so!)

One thing I particularly like is that the uploaded PDFs are stored in my Google Drive in a very sensible way. The files have obvious names: <Author> <Year> - <Title>.pdf. So if Paperpile were to disappear for any reason, I still have all my PDFs in a very usable form. (I’m not expecting it to disappear; I’m just paranoid: I have my Google Drive synched to my hard drive at home, in case Google disappears!)

I haven’t used all the facilities, in particular, I haven’t exploited the Google Docs integration. The Paperpile team report that some people are using it to write their papers collaboratively in Google Docs. That looks potentially useful, but I don’t use this plain format: I write my papers in LaTeX/BibTeX, using TeXnicCenter.

One part of the system I probably won’t use that much is the notes field. It’s perfectly fine, but I like to keep more substantial notes, with formatting, sketches, figures, and the like. So I’ll continue using my Evernote “papers” notebook, but now just for the notes, not the entire paper and other detials. For a paper with notes, I add a link from the Paperile entry to the relevant Evernote note, using the URL field which gives a clickable link (and I add an Evernote tag, to make it clear to me that there is a note). Similarly, I have a lot of already existing book reviews on my website. So when I include a book that I own, I add a link to the review on my site.

In fact, I think that having a variety of tools that each does their own thing well, with links between the relevant parts, works better than trying to shoehorn all capability into every application. So in Evernote, for example, for a meeting note, I don’t copy the agenda from the email notification, I just store a link to it. Evernote Webclipper allows you to clip the whole email into Evernote, but that just seems to cause unnecessary duplication. It’s good practice to keep only one copy of data (except for backups, of course!), otherwise there is the potential for copies to get out of synch. When that agenda is inevitably later updated, and a new copy emailed out, I don’t have to do anything: provided it is sent out with the same subject line, the Gmail link pulls up the entire conversation, and I can select the most recent agenda. (One day, the agenda setters may start using Google Docs, and so everyone will always have the most up to date version – provided they haven't clipped it into Evernote!) With all these web-based tools, the URL can link everything together.

In summary: first impressions of Paperpile are very favourable. I’ve successfully and productively used it for writing a paper, and it being web/Google based means it’s available at work, at home, and elsewhere, without me having to do anything special. I’ll report on progress if and when it gets more deeply integrated into my web-life.

1 comment:

  1. Dear Susan, one month on, how are you getting along with Paperpile? I having been using it for a few months, and am pretty happy with it. One key issue I have with it is its choice of bibtex keys -- I've talked to the developers about this. Have you seen the solution from Papers?: