I realize most NLP algorithms have multiple steps. (e.g. OCR/speech rec > syntax > semantics > response logic > semantic output > natural language output)

Is there any NN model that can perform multiple steps in NLP at once? For example, a single network which accepts audio input and returns a semantic analysis of the given speech, or a single network which accepts text input and returns natural language output?

  • $\begingroup$ Is T5 the sort of thing you're looking for? ai.googleblog.com/2020/02/… $\endgroup$ Jan 22, 2021 at 8:42
  • $\begingroup$ Yes! That's very close to what I had in mind. Do you know of anything similar that is capable of OCR/speech recognition as well? $\endgroup$ Jan 22, 2021 at 16:28


You must log in to answer this question.