Comment by akavel | Hacker News Reader

Comment by akavel | Hacker News Reader

I found the app awesome for quick writing with a stylus on a MS Surface (probably works on other tablets/tablet PCs too). It seems the development has stopped a few years ago, unfortunately; I'm very happy the author still at least makes the app available for download. And thanks to the HTML+SVG format used, the documents created in the app are at least open and future proof!

It would be nice if someone wrote an extension/post-processor applying some OCR to the SVGs, so that plain text could be extracted. I tried to do something like this myself, but failed. What was most annoying was that the SVGs the app writes are actually outlines of the pen strokes, after applying "pen width". As far as I've seen, OCR systems prefer raw strokes as input, ideally with pen pressure information. I don't have enough experience in Machine Learning either to try to build a new model for this use case from scratch.

edit: A sample article draft/experiment I wrote with it:

- as a HTML+SVG: https://akavel.github.io/post/2018-05-31-stylish-elephants/2...

- as a PDF: https://akavel.github.io/post/2018-05-31-stylish-elephants/2...

I'm hoping to add some rough OCR to allow searching in the future. In the meantime, if you turn off pressure sensitivity (Pen menu -> Custom Pen), you'll get raw strokes in the SVG instead of filled paths.

Even inaccurate OCR could be cool if it appeared on separate, correctable pane.

Oh, cool, didn't think of it! On to different projects now, but maybe some day... if you don't do it earlier yourself ;) (Though I actually like the pressure sensitivity feature very much...)

Also, in case you'd come back to this reply at some point in future: one feature I'd also love (I think I even wrote an email to you about it at the time) would be easy pasting of images into the document, and a "knife" tool, which would allow cutting them to parts along drawn lines. (And then moving and processing them separately.) I don't remember what exact use case I had for this, but I remember missing it badly.

Ah, and there was some problem with not being able to easily paste links into the document purely with stylus, i.e. without reaching for keyboard. I used Write once for writing a journal from a trip, with photos, and I think I might have had to edit the raw HTML/SVG to add links from thumbnails to full photos, or something like this.

For me, this would be perfect a perfect use for removing tablature from some guitar sheet music:

https://s3.amazonaws.com/halleonard-pagepreviews/HL_DDS_0000...

...basically be able to slice the above image into various rows, and delete the "numbers" part, but keep + rearrange the "music" part.

Seems like the ability to paste / rotate an image and then add in a "razer-slicer" set perpendicular to page or matching the image rotation: https://www.youtube.com/watch?v=H6SuHTHoRd0&t=1m25s

It's a relatively common operation (even for text, maybe even especially for lined text) giving you the option to write single-space (or double-space) and then magically have extra room to take notes, edit, etc, then later remove those edits.

Actually it'd be very similar to "folding" handwritten documents (in the code-folding sense).

Other note-taking tools generally do it by breaking strokes into lots of little line segments, each with a different width coming from the pressure. It works pretty well, and if the program stores the pressure-location data, the stroke width can be adjusted later. I do this all the time in Squid on Android, and have extracted stroke data from PDFs it exports.

I don’t really like Evernote/too heavy for my needs, but the OCR feature for handwriting/transcribing whiteboard sessions etc, was top notch. It’s cool to search your notes and get results from random whiteboard text, very usable.

The size of that HTML page is 1.4MB...

Don't ever look at the size of handwritten pages from OneNote then.

Reply on news.ycombinator.com