# How to autocorelate multiple variants of same text into one?

I want to improve quality of translations for open-source projects in Ukrainian language. We have multiple translations from different authors. We can also translate messages using machine translations. Sometimes machine translation is even better than human translation.

Given multiple variants of translation of the same original text, I want to create AI which will be able to "translate" from Ukrainian to Ukrainian, using these multiple variants in parallel as the source, to produce one variant of higher quality.

So, in general, given multiple similar input sequences, the neural network needs to "understand" them, and produce a single output sequence.

$$S_1, S_2, \dots \rightarrow S$$

For a simple example, we may want to train a NN to recognize a sequence of natural numbers: $$1,2,3,4, \dots$$. We give two sequences to NN: $$23,4,24,6,8$$ and $$3,65,5,6,23$$, then trained the NN is expected to produce $$3,4,5,6,7$$.

How to modify an existing neural network to achieve that? Is it possible at all?