Multilingual translator with intelligent intermediate form

[AndrzejB]

Language to language translator without intermediate form need n*(n+1) labuage models whereas
translator with intermediate form only 2*n
For 100 languages will be 9900 comparing to 200,for 150 languages will be 22350 comparing to 300.
Intermadiate form must not be usual national language but language without ambiguity, without homonyms.
Candidate for intermadiate form can be Universal Networking Language (UNL) Translation from UNL to national language is easy problem, whereas national language to UNL is hard problem.
One of problems are homonyms.
For example "band" has at least tow definitions
1.a musical group
2.a ring
How connect two sentences:
1.Band play rock ballad
2.I wear hair in band
to definitions 1 and 2?
(rock is also homonym)
I hear about algorithm Universal Sentence Encoder to compare two sentences, how it work?
How start with language->UNL translation problem?

[AndrzejB]

Main problem is Word Sense Disambiguation (WSD).
Lesk algorithms works better than I though: https://github.com/alvations/pywsd + Explossion/spaCy.