ML and NLP algorithm feature complete for MVP

The first 7 months of working on zimu.ai I spent completely on R&D for two core value propositions:

1. Have the ability to extract hard subs from any video
2. Segment Chinese captions into words, produce accurate pinyin romanization, and accurate in-context translations for each word

Number 2 was surprisingly difficult, especially getting in-context word translations to work reliably. It's still not certain that the work I put into this will pay off, but I think so. Chinese is a highly ambiguous language, so trying to reduce that ambiguity, and thus the cognitive load of the learner will be valuable I hope.