Neural Machine Translations
When the analysis is complete, the sentence is "transferred" by a second process with equal or less depth of analysis to the target language. Then a third process called "generation" forms the actual target sentence from this interpretation, that is, creates a sequence of words in the target language.
The idea behind using a triangle is that the higher (deeper) you analyze the original sentence, the easier the carryover phase is.
When the analysis is complete, the sentence is "transferred" by a second process with equal or less depth of analysis to the target language. Then a third process called "generation" forms the actual target sentence from this interpretation, that is, creates a sequence of words in the target language. The idea behind using a triangle is that the higher (deeper) you analyze the original sentence, the easier the carryover phase is. Ultimately, if we could convert the source language to some kind of universal "interlingualism" during this analysis, we wouldn`t need to perform the transfer procedure at all. It would only require an analyzer and generator for each language to be translated into any other language.
This general idea explains the intermediate stages when the machine translates sentences step by step. More importantly, this model describes the nature of the actions during translation. Let`s illustrate how this idea works for three different technologies, using the sentence “The smart mouse plays violin” as an example. , has 17 more meanings in the dictionary as an adjective, for example, "agile" or "dexterous" approx. per.)
Rule-based machine translation
Rule-based machine translation is the oldest approach and covers a wide variety of technologies - https://www.conveythis.com// as an example.
Phrase-based machine translation
Phrase-based machine translation is the simplest and most popular version of statistical machine translation. It is still the main workhorse today and is used in major online translation services.
Technically speaking, phrase-based machine translation does not follow the process formulated by Wokua. Not only that, there is no analysis or generation involved in the process of this type of machine translation, but more importantly, the subordinate clause is not deterministic. This means that the technology can generate several different translations of the same sentence from the same source, and the essence of the approach is to choose the best option.
Search algorithms intuitively prefer to use sequences of words that are the most likely translations of the original given the reordering table. This allows the correct sequence of words in the target language to be generated with high precision.
There is no explicit or implicit linguistic or semantic analysis in this approach. We were offered many options. Some of them are better, others are worse, but as far as we know, the main online translation services use this technology (ref: Wiki).