Hi, I'm glad to hear you to say that. I love Firefox for Android and use it everyday, and rely on addon heavily. So I always ensure all of my addon are Android-compatible if possible. Besides, DeepL is created by lots of professional linguists with lots of super computers, which is not an ordinary developer can do... And this addon is merely using a package "igo.js" (https://github.com/shogo82148/igo-javascript) created by a Japanese "shogo82148" to analyze sentences & get the pronunciations of vocabularies.

1. I've tested 元気 and it's furigana is correct 「げんき」, I guess the described problem occured because you used in Duolingo, which may frequently split a single vocabulary into multiple HTML elements, and this addon process the HTML elements one-by-one. This is a technical limit. 2. Such wrong furigana are maybe sometimes due to the used dictionary or igo.js, which is prepared by another software project, so it's not what I can fix. You can see the GitHub issue for detailed info for this. 3. You should use this addon for reference only, it's impossible to provide 100% correct furigana for kanji. This is technical limit, even Google's or Apple's AI can never do this 100% correctly. You should use an independent dictionary addon or app to ensure the vocabulary you want to know when in doubt.

Yes, I have noticed this bug since 1 year ago, however didn't find the reason and the way to fix it. But I finally found the reason this week, to fix this issue please update to 0.5.2.