Open Chechgm opened 2 years ago
Thanks for the suggestion! Here are some early thoughts:
I like the general idea of having an ebook export for arXiv. We have an early variant of that possible with latexml, which has experimental support for emitting the epub format.
Kindle does quite well with that, especially if they update their MathML support some time this year.
That said, making sure we generate ebooks of high quality would take some focused development and QA testing, so it is not something to embark on lightly. At present it also implies converting the entire article set twice, which we should optimize away - we should be able to emit both HTML and epub during the same conversion pass.
Finally, and most importantly, I would like such advanced features to get a blessing from arXiv, and ideally be initiated by/coordinated with their team. So I will definitely check with them before starting anything ebook-related.
To add an update to the same I was able to achieve some pretty good results through the following process
Most e-readers like the one I possess doesnt support rendering mathematical functions, so all mathfunctions should be converted to images for successfull viewing in epub
Open the respective ar5iv link in your browser and save a local copy of the same. In my case I took the famous "Attention is All you need" paper. Then I performed the following commands
pandoc -s -f html \[1706.03762\]\ Attention\ Is\ All\ You\ Need.htm -o attention_gladtex.htex
gladtex -f 20 -d \[1706.03762\]\ Attention\ Is\ All\ You\ Need_files/ --png attention_gladtex.htex
pandoc -f html -t epub2 attention_gladtex.html -o attention_gladtex.epub
In this process some minor styling properties like page border are lost, but the results are satisfactory enough I can now happily read research papers from my kindle
Github didnt allow uploading a epub file , the below zip file contains the epub generated with the above steps attention_gladtex.zip
Most e-readers like the one I possess doesnt support rendering mathematical functions, so all mathfunctions should be converted to images for successfull viewing in epub
Another constructive reason for me to wait a bit more on this issue is that there is now an expectation for this state of affairs to change soon.
MathML is a part of EPUB 3. Since MathML support has just recently returned to Chrome 109, I would expect that to cascade into all Blink-based EPUB readers in the near future.
Thanks for the fantastic tool! My dream is to read scientific papers on my kindle, but most of the tools out there render papers very poorly. Yours is working well most of the time. Please consider having an export to some e-reader format.