How to simplify the sentence as much as possible using neural networks?

Question

Neural networks

How is the neural network here that converts for example:

"Hi, today I heard a hit called {hit} look on the Internet, what is it"
into
"Look on the Internet {hit}"

How can I make such a neural network, preferably through Neuroph (Studio )
Or maybe it can be done without the National Assembly?

Answer 1

This is ambiguous and depends on the problem statement.
1) If without a teacher - read about the extractive summarization of the text (it is the easiest to implement). In a nutshell - you will need to assign a certain rank to each word and leave only the most important words. You can rank in different ways. You can conduct a semantic analysis of the text, for example, using the stanza library, by building a sentence graph, then calculate the weights of words, for example, through textRank, as in the case of extractive summation of texts, and leave the number of words you need.
2) If with a teacher, you need to make and train the network using examples, and you will need a training dataset before and after. Since the task is complex, most likely you will need to do TransferLearning with a generalized model like Bert or GPT. In this case, in theory, a few hundred examples will be enough.