What are the main datasets that can be used to train ChatBots ?

Here we include a list of dialogue datasets that can be used to train chatbots:

Table of datasets for chatbot development

CategoryName	Context	Structure	License
Schema Guided Conversation
Schema-Guided Dialogue (SGD)	Conversations with system spanning 20 domains, ranging from banks and events to media, calendar, travel, and weather	Wide range of annotations including system and user actions and user states	CC BY SA 4.0
DialoGLUE	A collection of data from various sources with labelled intent	Includes Banking77, HWU64, CLINC150, Restaurant8K, SGD, TOP, MultiWOZ2.1	Various Licenses including CC-BY, CC-BY-SA, MIT
MultiWoZ	Collection of single and multi-domain conversations	Annotated with goal and belief state	Apache License 2.0
M2M	Set of conversations between an agent and a simulated user for booking a restaurant ticket and buying movie tickets	annotated with dialogue state and actions	Not mentioned.
Frames Dataset	Conversations about vacation packages including flight and hotel.		Not mentioned.
WoZ	Dialogues with state tracking on restaurant search domain	Annotated with system states	Not mentioned.
Dialogue State Tracking Challenge (DSTC2&3)	Conversations in 2 domains of restaurant information and tourist information.	User dialog-act semantics and dialog states are annotated.	Not mentioned.
Intent Datasets
BANKING77	Fine grained set of intents in banking domain	annotated with intents	CC-BY-4.0
CLINC 150	Popular personal assistant queries	annotated with intents	CC-BY-SA 3.0
HWU64	Popular personal assistant queries	annotated with intents	CC-BY-SA 3.0
Others
DailyDialog	Daily Life Conversations including ordinary life, school life, tourism, attitude & emotion, relationship, health, work, politics, culture & education and finance.	13K of conversations with average 8 turns and annotation for new ideas and emotions	Research Only. No Commercial Use.
└ DailyDialog++	Divided into shorter conversations and added alternative responces for last utterance	Includes adverserial irrelevant responses	Not mentioned
Persona-Chat	Chit chat based on profiles	Includes profiles and conversations	CC BY 4.0
└ ConvAI2	Rephrasing of statements in Persona-Chat to avoid obvious overlap	Includes profiles and conversations	Not mentioned.
DialogRE	Transcripts of Comedy Series Friends	Annotated for 36 possible relation types	Research Only. No Commercial Use.
Dialog System Technology Challenges (DSTC)7 Task1	Selected conversations from the Ubuntu Internet Relay Chat (IRC) channel and an Advising dataset from the University of Michigan		Not mentioned.
└ Dialog System Technology Challenges (DSTC)8 Task2	Added complexity to DSTC7 Task1	Including conversations with more than 2 participants and multiple simultaneous conversations and predicting whether the dialogus has solved the problem yet.	Not mentioned.
Dialog System Technology Challenges (DSTC)7 Task2	Conversational data from Reddit	Includes the conversation and the relevant facts on the topic of the conversation	Downloaded via script.
Taskmaster-1	Conversations in one of the following 6 domains: ordering pizza, creating auto repair appointments, setting up ride service, ordering movie tickets, ordering coffee drinks and making restaurant reservations		CC BY SA 4.0
└ Taskmaster-2	Conversations in one of the following 7 domains: restaurants, food ordering, movies, hotels, flights, music, and sports		CC BY SA 4.0
Curiosity	Dialogs about geographic topics like geopolitical entities and locations	Includes fine-grained knowledge groundings and dialog-act annotations.	CC BY NC 4.0
Coached Conversational Preference Elicitation	Conversations on movie preferences	Annotated with entity mentions, preferences expressed about entities.	CC BY SA 4.0
Twitter Conversations Dataset	Conversations between users and customer care agents on twitter in 25 organizations	Includes the URL to a document including additional information on user request	Apache 2.0
Action-Based Conversations Dataset (ABCD)	Human conversations constrained by policies and sequence of actions	Annotated with user and order details	MIT License
Key-Value Retrieval Networks (KVRET)	Knowledge grounded conversations spanning 3 distinct tasks: calendar scheduling, weather information retrieval, and point-of-interest navigation		Not mentioned.
EmotionLines	Human conversations	Labelled with one of 8 emotions:anger, disgust, fear, happiness, sadness, surprise, neutral, and non-neutral	Not mentioned.
Dyadic Dialogues Relationships (DDRel)	Conversations crawled from movie scripts on IMSDb	Annotated with one of 13 predefined interpersonal relationships.	Not mentioned.