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

When an NLG algorithm fails to understand some part of the user’s utterance, instead of using a reject dialogue act (ex: "Please repeat"), it uses targeted clarification questions that reprise elements of the misunderstanding. For example:

Clarification Questions

Santhanam, Sashank and Shaikh, Samira. A Survey of Natural Language Generation Techniques with a Focus on Dialogue Systems - Past, Present and Future Directions. arXiv, https://arxiv.org/abs/1906.00500.

A Survey of Natural Language Generation Techniques with a Focus on Dialogue Systems - Past, Present and Future Directions

Reiter and Dale presented a standard architecture for the developing NLG systems that comprised of six components each performing an important task to generate a coherent output:
- Content determination
- Document structuring
- Lexicalization
- Referring expression generation
- Sentence aggregation
- Linguistic realization

NLG Six-Stage Pipeline Architecture

Applying deep neural networks to Natural Language Processing has helped achieve state-of-the-art performance across different tasks, including the task of language generation due to the capability of neural networks to learn representations with different levels of abstraction. There are four main approaches that have been applied to the task of language generation:
- Language models
- Encoder-decoder architecture
- Memory networks
- Transformer models

Deep Learning approaches for Language Generation

Dialogue Systems or Conversational Agents (CA) are designed with the intention of generating meaningful and coherent responses that are easy to respond to and informative when the system is engaged in a conversation with humans. A good dialogue model incorporated in conversational agents should be able to generate dialogues with high similarity to how humans converse

Open Domain Dialogue Systems

Language Generation is a sub-field of the field of Natural Language Processing (NLP), Artificial Intelligence (AI) and Cognitive Science (CS).

For natural language generation in the dialogue-state model, its task is to generate the text of the response to the user once a dialogue act has been decided. 

University of Michigan - Ann Arbor

Hamilton College

Dialogue systems, or conversational agents, are programs that communicate with users, through text and/or speech, in natural language.

Dialogue Systems

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

Speech and Language Processing (3rd ed. draft) 

1) Chatbots

2) Task-oriented dialogue agents

Types of Dialogue Systems

A Frame-based Dialogue System defines a dialogue frame structure and the formatted involved important components.

Frame-based Dialogue System

Natural Language Generation (NLG)

- Made use of by dialogue system.
- Represents the interactive function of the turn or sentence, combining the idea of speech acts and grounding into a single representation.
- Different types of dialogue systems require labeling different kinds of acts.

Dialogue Act

- Slot-filling, and the simpler tasks of domain and intent classification, are
special cases of the task of supervised semantic parsing,  in which we have a training set that associates each sentence with the correct set of slots, domain, and intent.
- A simple method is to train a sequence model to map from input word representation to slot fillers, domain, and intent.


Slots Filling

- Determines both the current state of the frame (the fillers of each slot), as well as the user’s most recent dialogue act.
- includes more than just the slot-fillers expressed in the current sentence:  the entire state of the frame at this point, summarizing all of the user’s constraints

Dialogue State Tracking

Chatbots and task-oriented dialogue agents are evaluated on different metrics; chatbots are scored on how the user feels when interacting with the program, whereas task-oriented systems are scored based on their ability to complete a task.

Learn Before

Related

Learn After