WHAT IS A CHATBOT?

“A chatbot is a computer program which conducts a conversation via auditory or textual methods.”

https://en.wikipedia.org/wiki/Chatbot

TURING TEST

• “I propose to consider the question, 'Can machines think?'”
• “Are there imaginable digital computers which would do well in the imitation game?”
• "Computing Machinery and Intelligence"

Alan Turing, 1950, Mind.

ELIZA, 1966

• Created at the MIT Artificial Intelligence Laboratory by Joseph Weizenbaum.
• Eliza simulated conversation using pattern matching and substitution.
• First to pass the Turing Test?

PARRY, 1972

• Created by psychiatrist Kenneth Colby, Stanford University
• PARRY attempted to simulate a person with paranoid schizophrenia - "ELIZA with attitude".

​

LOEBNER PRIZE

• Annual competition in artificial intelligence that awards prizes to the chatterbot considered by the judges to be the most human-like.
• Launched in 1990 by Hugh Loebner.

LOEBNER PRIZE

• A human judge simultaneously �holds textual conversations with a computer �program and a human being via computer. �Based upon the responses, �the judge must decide which is which.
• Criticized because it does not aim at understanding or intelligence but resort to basic ELIZA style tricks.

A.L.I.C.E., 1995

• Artificial Linguistic Internet Computer Entity
• By Richard Wallace
• Inspired by Eliza, also uses pattern matching and substitution
• AIML (Artificial Intelligence Markup Language)
• Inspiration for Spike Jonze Her

​

AIML: ARTIFICIAL INTELLIGENCE MARKUP LANGUAGE

http://www.alicebot.org/aiml.html

​

<category>� <pattern>WHAT ARE YOU</pattern>� <template>� <think><set name="topic">Me</set></think> � I am the latest result in artificial intelligence,� which can reproduce the capabilities of the human brain� with greater speed and accuracy.� </template>�</category>

RETRIEVAL MODEL

• Easier
• Uses a repository of predefined responses and some kind of heuristic to pick an appropriate response based on the input and context.
• The heuristic could be as simple as a rule-based expression match, or as complex as an ensemble of Machine Learning classifiers.
• These systems don’t generate any new text, they just pick a response from a fixed set.

​

http://www.wildml.com/2016/04/deep-learning-for-chatbots-part-1-introduction/

GENERATIVE MODEL

​

• Harder
• Generates new responses from scratch.
• Typically based on Machine Translation techniques, but instead of translating from one language to another, we “translate” from an input to an output (response).

​

http://www.wildml.com/2016/04/deep-learning-for-chatbots-part-1-introduction/

OTHER CONSIDERATIONS

“Short” vs. “Long” conversations

​

Short-Text Conversations:

• Easier
• The goal is to create a single response to a single input.
• Example: A specific question from a user and reply with an appropriate answer.

​

Long Conversations:

• Harder
• Go through multiple turns and need to keep track of what has been said.
• Customer support conversations are typically long conversational threads with multiple questions.

http://www.wildml.com/2016/04/deep-learning-for-chatbots-part-1-introduction/

​

OTHER CONSIDERATIONS

“Open Domain” vs. “Closed Domain” conversations

​

​

​

Open Domain Conversations(Harder):

• The user can take the conversation anywhere. There isn’t necessarily a well-defined goal or intention.
• The infinite number of topics and that a certain amount of world knowledge is required makes this a hard problem.
• Examples: Conversations on social media sites like Twitter and Reddit.

​

Closed Domain Conversations(Easier):

• The space of possible inputs and outputs is limited because the system is trying to achieve a very specific goal
• They just need to fulfill their specific task as efficiently as possible.
• Examples: Technical Customer Support or Shopping Assistants

http://www.wildml.com/2016/04/deep-learning-for-chatbots-part-1-introduction/

​

A PERSONA-BASED NEURAL CONVERSATION MODEL

https://arxiv.org/pdf/1603.06155.pdf

​

RIVESCRIPT

• Simple scripting language for chatbots with a friendly, easy to learn syntax.
• Exposes a simple plain text scripting language that's easy to learn and begin writing in quickly.
• Handful of simple rules that can be combined in powerful ways to build a chatbot personality.

​

https://www.rivescript.com/

​

RIVESCRIPT

• Takes a "Unix-like" approach to its development: the core library is small and self-contained and it does one thing very well—takes human input and gives an intelligent response.
• Released under the most liberal open source license available, MIT.
• Interfaces are available for Go, Java, JavaScript, Perl and Python.

​

https://www.rivescript.com/

​

​