Applying the lemma detection only to words that have two or more possible parts of speech.

[Vuizur]

In French, when you activate POS detection, words such as participles, for example glaçant, don't get annotated with their English translation. This is because glaçant is a participle, but appears in the declination table for the verb "glacer". So spacy detects a participle, but the Worddumb data has it as a verb, which is why it gets no English translation.
My suggestion would be to perform POS detection only for words that have 2 or more possible parts of speech, and handle clear cases such as glaçant like the multi-word-phrases and ignore the grammar.

PS: I have been using the project quite a lot recently, it really works well. Thanks a lot for the active development!

[xxyzz]

Please try b8b1092

[Vuizur]

Thanks a lot for the very quick commit. Unfortunately it doesn't seem to work for me yet. For example in this public domain book, the word "véreuses" (a few pages in) is not annotated when using POS detection, but it is when turning it off.
pg63979-images-3.zip

[xxyzz]

Fixed in 32e4559 and 9af0c6e

"use pos" option is removed, the code could find all enabled lemmas and their forms even if spaCy returns wrong POS or wrong lemma.

[Vuizur]

Thanks a lot for implementing this! I tested it, and it works great. 👍 I think your solution seems to be really the best.