How to Model pattern matching using ML - various input predict output

The Blackbox shown provides outputs based on input. Outputs are sometime string concatenation of parts or whole input sometime they are multiplication of two input. I want to create ML model which can predict output1 , ouput2 ... output5 based on provided input1, input2.. input5

Basically I am trying to reverse engineer Black box. Diagrams below show sample data and High level view.. The Blackbox could vary.

What kind of Models can I use for the string pattern ( LSTM, RNN .. ?) High level view

Sample data sample data