Chapter 6. Chatbots

Applications

fixed set of responses

no dependency among reponses

FAQ bots

understand and track this information throughout the conversation

flow-based bots

Goal-Oriented Dialog

converse with the user about various topics

open-ended bots

Chitchats

Taxonomy

voice to text

Speech recognition (IVR: Google ASR)

Natural language understanding (NLU)

gathers and systematically decides which pieces of information are important or not

Dialog and task manager

generates a response in a human-readable form according to the strategy devised by the dialog manager

Natural language generation

text back to speech

Speech synthesis

A Pipeline for Building Dialog Systems [link]

the aim of a user command

Dialog act or intent

the fixed ontological construct that holds information regarding specific entities related to the intent

Slot or entity

contains both the information about the dialog act as well as state-value pairs

Dialog state

a set of dialog states that also captures previous dialog states as history

Context

Dialog Systems in Detail - pizza example using Dialogflow

CNN

Good at sequential data

RNN

Bert

Dialog Act Classification

a popular sequence labeling technique and are used heavily in information extraction

CRF++ (Conditional random fields)

Identifying Slots - NER(named entity recognition)

Natural Language Understanding (NLU)

Fixed responses

Templates are very useful when the follow-up response is a clarifying question

Use of templates

conditional generative model 

Automatic generation

Response Generation

Deep Dive into Components of a Dialog System

take a sequence as input and output another sequence

generally LSTM based

seq2seq

End-to-End Approach

generated utterances with considering how to respond in order to have a good conversation

goal-oriented dialog and seq2seq-based generation

Deep Reinforcement Learning for Dialogue Generation

Human-in-the-Loop

Other Dialog Pipelines

performs NLU and captures required slots and their values

Context-based conversations

create more training data and provide human feedback

Interactive learning

Data annotation

API integration

Customized models in Rasa

Rasa NLU

Chapter 6. Chatbots