“Humorist Bot: Bringing Computational Humor in a Chat Bot System”
Chat Bots are computer program that mimic conversation with people using artificial intelligence. They can converse in the same manner as human do. Chat Bot is a term that refers to chatting with machine. There are some bots that can talk also but in this review we will discuss only about the chat bots. These Chat Bots works on the AIML (Artificial Intelligence Markup Language). In the review paper we will discuss about how AIML supports chat bots. This review paper deals with humoristic bot that shows capability to have a “sense of humor” is presented.
Chat Bots is a term that refers to chatting with machine. There are some bots that can talk also but in this article we will discuss only about the chat bots and in chat bots we specifically about Humor. “Humor is a part of everyday social interaction between humans”. There are two stages first is humor generation and second is capability to understand humor. Formally Humor is defined as short expression(jokes) or one liner that is a short sentence with comic meaning. While doing conversation user input some jokes or one liner and in correspondence as a reply system generates humorous sentence.
In this review we will discuss about a humoristic system that can deal with the humoristic expression, telling jokes and answering user. Chat-Bot is capable of understanding jokes, listening them and react by changing expression of avatar. For example avatar can smile if it finds users input funny or remains a neutral face when it finds that user’s input is not funny.
To make this available for public it is being deployed on the Yahoo Messenger as Yahoo Instant Messaging Contact.
Chat Bot system follows simple matching pattern which is mostly being written in the language as AIML(Artificial Intelligence Markup Language). One of the popular examples is A.L.I.C.E. (Artificial Intelligence Foundation) is the Artificial Linguistic Internet Computer Entity, which was first implemented by Wallace in 1995. Alice’s knowledge about English conversation patterns is stored in AIML files. It consists of Basic Knowledge of ALICE (25,000 categories)
AIML consists of data objects called AIML objects, which are made up of units called topics and categories. These topics are an optional top-level element, it has a name attribute and a set of categories related to that topic.
Types of AIML or ALICE Categories
There are three types of categories: atomic categories, default categories, and recursive categories. a. Atomic categories: are those with patterns that do not have wildcard symbols, _ and *, e.g.:
<template> Wow, that is cheap.</template>
In the above category, if the user inputs ’10 dollars’, then ALICE answers ‘WOW that is cheap’.
bot is capable of listening jokes and understanding them and according to understanding if the system finds the joke “funny” then it reflects with the facial expression of avatar and in case if the joke is poor avatar changes it’s expression as straight face or anger.
It is also available as Yahoo Messenger instant messaging clientto make it available to the end users. Humoristic Bot architecture is as follows firstly there is a user who interacts with yahoo messenger’s interface and through that it reaches to AIML Category then with the help of computational humor engine it finds out the appropriate result from WorldNet and CMU Dictionary. As shown in the below diagram.
In the given system they focused on the recognition features as it is difficult to handle. There are some important terms related to it such as Alliteration, Antinomy, Adult Slang, Knowledge base, and generation of humoristic sentences. We will look into that in the next session of this paper.
Alliteration is the process in which we successive repetition of similar kind of word(phonetically similar). The module finds alliteration if there is at least three words are common.
Let’s take an example “Batty bought some butter but the butter was so bitter so batty bought some better butter to make the bitter butter better.”If we say these kinds of lines even if it is not funny but Phonetics(due to its sound) makes it funny.
It is another way of producing humor in the sentences. A special module has been developed where it deals with detection of antinomies such as nouns, verbs, adverbs and adjectives. For better understanding let’s take some examples:-
· A day without sunshine is like night
· Artificial intelligence usually beats real stupidity
· Good girls are bad girls that never get caught
These are some sentences where nouns, verbs, adverbs and adjectives are used in such a manner that they produce humor. That is an important part of generating Humor.
C. ADULT SLANG
Adult Slangs play a key factor in the production of humor. It generates comic effects in the sentences. It deals with sexual domain which is mainly used in Adult Slang. Let’s take some examples for the better understanding of this domain:-
· 1000 sperms and you were the fastest
· Sex is like air; it is important unless you aren’t getting any
The Humoristic chat-bot it’s knowledge base consist of ALICE categories which allow entertaining the user. Below example will help in understanding how it is being implemented.
User: What’s your name?
FunnyBot: My name is FunnyBot… and you? What is
Funny Bot: Nice to meet you, Marc!!
<pattern>WHAT IS YOUR NAME</pattern>
<template> My name is <bot name/>..and you? What
is your name? </template>
You can only generate the proper humoristic sentences only you are able to create proper categories of AIML. In AIML there is <random> tag is being used to create a script in such a way when the system got stuck in such a situation where system is unable to match the specified input from the list in such cases system refers to this category and tries to change the topic.
To make the chat bot available for the real time they used instant messaging system that is Yahoo! Messenger and it is freely available there. You can interact with it at any point of time(24x7). To implement it Java classes (JYMSG) are being used library that enables us to interact with it.
With the help of the Yahoo! Messenger we are able to use the avatar that beings different kind of expressions as shown below.
The initial to faces in the image shows the smiling faces and the third one depicts the straight face and last two faces shows the angry face. Clearly in that bot is not happy with your reply or may be the humor that you are trying to create is, you are failed in that.
To verify the product we need some statistics that is shown below we can verify our bots working. We can check our result in two phases as Positive Data set and Negative Data set.
Clearly Adult Slang is the feature that is attracting most of the users. In the first experiment results report 66% of correct recognition and 34% of no recognition. In the second experiment the results are reported to be 81% of correct recognition and 19% of no recognition. That is shown below:
Humorist bot needs to contain features such as Alliteration, Antinomy, Adult Slang, Knowledge base, and generation of humoristic sentences. With the help of these features an effect bot can be created. The more number of scripts will be there the bot will be more effective.
In the future scope of this they are preparing to make conversation with the bot in different languages.
 Humorist Bot: Bringing Computational Humor in a Chat-Bot System
 Erhardt C. Graeff,MIT Media Arts and Sciences at www.erhardtgraeff.com
 Alice Chabot: www.alicebot.org