I want to implement a model that improves itself with the passage of time.
My main task is to build a machine translator (from English to Urdu).. The problem I am facing is that I have very little dataset available to train. Even if I create a corpus still there is a possibility of that corpus having poor translations due to outdated word choice for my native language.
I was thinking to create a model which predicts output and user tells whether it is correct or not. Or maybe suggests a better translation.
Now I have two options.
Take that input from end user, append it to my dataset and retrain the model. (I don't know whether it is even possible or not at production level).
Second is to reinforce that data into previous system. So far I only came to know about Online learning or Reinforcement learning (Q-learning, as my data is very small and even if user is training still not going to be in millions of sentences)
Am I on the right track, and how can I progress with either of these two options? Is there any prebuilt solution similar to this?