In real practice, we are unlikely to do data training with every possible attribute in many different worded sentences. Thus, it's common to increase the generality of the training examples.


Method(s) to Improve Generality

Sentence realization determines how to express the response to the users. The goal of the sentence realizer is to generate a sentence like lines 1 and 2 shown below, by training on many such examples of representation/sentence pairs from a large corpus of labeled dialogues.

Hamilton College

The task of natural language generation (NLG) in the information-state architecture is often modeled in two stages, content planning and sentence realization.

There is also a three stage pipeline architecture proposed by Reiter and Dale focusing on the aspects of document planning, sentence planning and surface realization.

NLG Stages in the Information-state Architecture

An on-going but a helpful book resource about NLP
https://web.stanford.edu/~jurafsky/slp3/

Speech and Language Processing (3rd ed. draft) 

Content planning decides what to say in response to the users. It's normally assumed to be done by the dialogue policy, which has chosen the dialogue act to generate, and chosen some attributes (slots and values) that the planner wants to say to the user.

Learn Before

Related

Learn After