1. Some key combinations allow moving forwards and backwards through the layout list. Of course you still need to know the sequence and the current position in that sequence for this to work efficiently.
2. Using gconf you can add some of the layouts to a secondary list. This mean that these layouts will only be available for selection through the keyboard indicator applet – only 2-3 mouse clicks away – which is really cumbersome when you need to switch around a lot.

Unfortunately there doesn’t seem to be a sane way to switch between more than two layouts. In my case I have two primary layouts and a third that I only use occasionally (albeit when I do use it, I switch a lot between it and the other two). My solution involves messing with the XKB configuration. It is not a perfect solution because I had to modify system files which will get reverted at the next upgrade, but at least it doesn’t interfere or override Gnome so hopefully there won’t be any “weird” problems.

Since I was accustomed to switching between my primary layouts with the otherwise useless CapsLock key, I didn’t want to change that behavior. As a result in the new configuration CapsLock still switches between the two primary layouts and to go to the third one Shift-CapsLock is employed. It would have been nice to hit Shift-CapsLock once again to go to the primary I switched from, but unfortunately I didn’t find a way. In the meantime when in the third layout CapsLock will switch to the first and Shift-CapsLock to the second.

These are the things I added:

/usr/share/X11/xkb/symbols/capslock

partial hidden modifier_keys
xkb_symbols "cap3switch" {
  replace key <CAPS> { symbols[Group1]=[ ISO_Next_Group, ISO_Prev_Group ],
                       symbols[Group2]=[ ISO_Prev_Group, ISO_Next_Group ],
                       symbols[Group3]=[ ISO_Next_Group, ISO_Prev_Group ]};
};

/usr/share/X11/xkb/rules/evdev

  grp:caps3switch	=	+capslock(cap3switch)

/usr/share/X11/xkb/rules/evdev.lst

  grp:caps3switch      CapsLock (between first two layouts) and Shift+CapsLock for third

/usr/share/X11/xkb/rules/evdev.xml

<option>
   <configItem>
      <name>grp:caps3switch</name>
      <description>CapsLock (switch between the two first layouts),
                               Shift+CapsLock goes to third layout</description>
   </configItem>
</option>

Then all that needs to be done is go to the keyboard preferences, add the layouts and from the layout options choose the new switching mechanism. Btw besides ISO_Next_Group and ISO_Prev_Group you can also use ISO_First_Group and ISO_Last_Group, that might be more useful if you use 4 active layouts. If you have a better way of accomplishing the same thing please let me know.

I decided to use Xournal since it is really easy to create PDF annotations with it and it saves them separately in XML format. Xournal allows text notes as well as a form of “highlighting”. I am using quotes because it does not really allow you to highlight the text itself, but you just freely paint over it. To capture the information from Xournal I wrote a small python script. The script exports text notes as they are, using the first line as a title. For highlighted regions, it calculates the region’s bounding box, crops the PDF and then converts the remaining to text. I also reserved the black highlighting for exporting “images” from the PDF such as graphs and photos. Each item is followed by a link either to the original PDF at the page it was found, or in the case of the images to the image file. Since I am using LaTeX and BibTeX, I also needed a way to capture the citation information. Personally I save the PDF files using the citation key, e.g. Author2009.pdf refers to the entry with citation key Author2009. To make things easier, the title of each link includes the citation key and the page in the original document.

The following picture shows page 16 of the pgf-plot manual in Xournal with a couple of highlighted regions and a text note. It is followed by a picture of the captured notes in Emacs.

Sample paper annotated in Xournal

Captured org-mode notes. The Figure 1 link was clicked and xpdf opened to show the image file.

I should note that using the bounding box of the highlighted region may lead to copying over redundant text and in any case the note file may need some post-processing to correct the pdf-to-text conversion artefacts. Nevertheless, I believe that the benefit of using Xournal outweighs this drawback. Personally I do have to post-process it a bit anyway, to add tags and copy the notes over to a master file specific to the project I am working on.

Here is a link (check below) to the script in case anyone wants to try it (it uses pdftk, pdfcrop and pdftotext) – by the way I am not a programmer so although nothing nasty is expected don’t actually count on it. To open the pdf links you will need the following snippet in your .emacs file (copied from here) which opens the pdf at the right page.


(require 'org)
(org-add-link-type "pdf" 'my-follow-pdf-link 'identity)
(defun my-follow-pdf-link (link)
"Follow links of the type path/to/file@line-no"
(when (string-match "\\(.+\\)@\\([0-9]+\\)$" link)
(start-process "*xpdf*" nil "xpdf"
(match-string 1 link)
(match-string 2 link))))


One additional feature that could be added is to use OCR to capture text regions in scanned documents, but I doubt that the resolution available in most such documents is sufficient. Anyway I would appreciate any suggestions for improvements and I would like to thank Shrutarshi Basu, whose post was the starting point of this idea.

Update 22 Oct 2009: It turns out that I had more scanned papers than I thought. As a result I added OCR-support in my script using Tesseract. I’ve actually tried it and it works pretty well so far. The new version can be downloaded from here. I also added a couple of options and fixed 1 or 2 things. If you have any comments, suggestions or corrections please let me know, especially if you have a patch ready…

1. Collect data
2. Capture information (excerpts from data, annotations)
3. Organize information for easy processing and retrieval

Unfortunately there seem to exist different scenarios with different requirements and each application I came across was geared towards one of them.

Scenario (A) Student taking notes in class: The data collected in this case are (or supposed to be) handwritten notes using a Tablet PC. No capture step is usually required since – with the possible exception of OCR – no post-processing is required to use the notes to study. Organization can be as simple as putting the files in the appropriate course folder and naming it using the date or keywords. In linux this niche is filled by Xournal and its variants.

Scenario (B) Document review: In this case the “data” is the “information” and is captured inside the document itself so no additional organization is necessary. This corresponds to the “track changes” feature in Openoffice. For PDFs one can use Xournal again although it is not really streamlined for this job. Recent developments in poppler will (hopefully) soon allow full support of annotations in other pdf viewers as well. Okular also features commenting but the last time I tried it, it was very crash-prone. There are a couple of online solutions as well (e.g. A.nnotate)

Scenario (C) General notes and GTD: This is a more general case that encompasses different applications. The information is entered directly into the application and is organized in groups/hierarchically (KeepNote, Tomboy, Basket) as a wiki (Zim, Tiddlywiki) or as a mind map (Freemind, Compendium). A multitude of web-applications are available, some generic and others very specialized. Personally I started using the emacs org-mode and I am not going back.

Naturally, I am not in any of the scenarios above. What I need is to review a large volume of mostly scientific literature, aggregate the information and write my stuff. This complicates things due to the following reasons:

• Almost all documents are in PDF format and the required information is in the form of notes on specific sections, text excerpts and figures/graphs.
• All documents come with citation information. As a result, the notes must interface with my current BibTeX database.
• Notes are to be central, linking back to the original documents and citation information and not the other way around as in many reference management software I found (e.g. Mendeley, Zotero).
• The solution I will use must be compatible with LaTeX.
• For my personal convenience (and in accordance with Unix tradition) I prefer to have as much as possible in simple text files, locally on my PC.

To accommodate my needs, I decided to use multiple tools since I couldn’t find a single one that could do everything – maybe someone can prove me wrong. I’ve started preparing a solution based on Emacs’ org-mode, Xournal and scripting. In my next post I plan to write on capturing org-mode notes using Xournal (as soon as I get it to work adequately).

Update: The script is now ready, check this post.