Similar projects

[AntonOfTheWoods]

I am just implementing initial epub support for the open source project I am developing for my PhD - transcrob.es - and I came across this interesting project! It does a number of similar things, though the engine stuff (translation, lemmatisation/word segmentation, etc.) is all done server-side, and the initial focus was on web pages. I am using CoreNLP and Bing Translator rather than Stanza and Amazon, but I have also developed an Anki-compatible sync server - djankiserv that integrates into the system. I have started implementing epub.js support, though was thinking more along the lines of a progressive web app rather than electron, as for me the client applications need to be available to everyone everywhere, and that definitely means on mobile. I personally much prefer reading on my tablet. So all intelligence is done on the server - it also means that user data is collected on the server, and that means you can make some really interesting predictions on the highest value things to learn for individual learners, and add lots of other stuff that needs data.

Let me know if you think there is any way we can cooperate!

[mcthulhu]

Thanks, I'll think about it and get back to you. I think I've looked at that web site before, so I must have come across a reference to it before.

[ParaplegicRacehorse]

{no intention to hijack discussion. apologies. I think, however, that at least some of these comments apply to both projects}

I am ... thinking more along the lines of a progressive web app rather than electron, as for me the client applications need to be available to everyone everywhere, and that definitely means on mobile. I personally much prefer reading on my tablet.

Wow! Awesome! Can I suggest a WebAssembly client rather than a progressive web app? This would mean users don't need to get set up hosting or paying a service. LWT was awesome, but I think the turn-off for many was the setting up of a server with PHP, database, etc.

Also WebAssembly is WAY faster, and more RAM friendly, than Electron; and can be compiled for mobile OSes.

Reliance on an internet service is problematic for people, like me, who are required to spend weeks and sometimes months unplugged due to working conditions (no internet at sea!) While these people are clearly outside the target audience, it would be wonderful to have some means optionally translating on-device, even if results are less good.

So all intelligence is done on the server - it also means that user data is collected on the server, and that means you can make some really interesting predictions on the highest value things to learn for individual learners, and add lots of other stuff that needs data.

A pox on Bing/Amazon/Google/Baidu/Tencent! Can we optionally add our own DeepL API key? ( going to open a new Issue for this... )