How to restore an English word from a Russian phonetic notation?

Question

Computational linguistics

For example, how to understand that "signature", "signache", "signature", "signature", "seinache", "seinacha", "signatures" are all signature? What is the name of such a task? This is for me to google on my own. Or maybe you can suggest a well-known solution, approach?

Answer 1

I will express my opinion. When it comes to pronunciation, accent - i.e. English, American, Indo-English, Chinese-English, etc., and it is precisely the recognition of pronunciation that the task, according to my information, is unsolvable today. There are no reliable systems capable of reliably recognizing spoken language, especially in conditions of erroneous (or non-canonical) pronunciation, which is an accent.
If we are talking about the pronunciation of words by a person who does not know English at all, but reads as it is written ("signature"), then, firstly, it is not clear why this is necessary, and secondly, all the more so, I have not seen such systems (I do not exclude that from -for their senselessness, no one bothers with this).
And in general, to identify "seynacha" and ['sɪgnəʧə] - I think not every person can cope. I think that just such an example cannot be considered anything other than some kind of gibberish of an illiterate (in terms of knowledge of English) person. There can be an infinite number of such "gibberish" variants ("sinatures", "signatures", etc.).