Encoder Decoder model for parameter extraction from text input

Question

I have an input as text from which I want to extract parameters as given in example below.

Input:

"client need to pay penalty of 10%  of amount  if there is delay in project for more than 3 months"

and output:

penalty = 10% and delay = 3

assuming there are N number of such parameters.

Here I have thought of using encoder and decoder model . Where I use RNN as encoder for text input Now I wonder what would be decoder architecture that will output N parameters and their values. what is alternative architecture to solve this problem.

Thanks in advance

Answer 1

Some popular methods for POS-tagging are Hidden Markov Model (HMM) and Conditional Random Field (CRF) book.

But you could use any recurrent networks (RNN, LSTM, bi-LSTM), CNNs or Transformers, to process your sequence of embeddings.

One could also use BERT (or other pre-trained models) to perform Name Entity Recognition blog.