Universal Chatbot

Taking the principles of aiml - but allowing a dynamic and extendable *.js environment

npm install universal_bot

To create a new instance of the bot

var UniversalBot = require("universal_chatbot);

var bot = new UniversalBot();

Passing parameters

Usuallay, you you would pass some parameters, al list of aiml-type files, a strategy and so forth.

File structure and internal logic

index.js is where it all begins. Here you define the bot's psychology, set the parameters and so forth.

The bot may follow a strategy.

aiml files

var params = {};

params.aiml = [
        "./aiml/conversation_start.json",
    ];


var bot = new UniversalBot( params );

bot_identity

You can pass a bot_identity like this:

params.bot_identity = {
    name: "UniversalBot",
    age: 12,   
    hobbys: ["music", "telepathy"]
}

conversation strategy

The bot will follow a conversation strategy which is laid down in a *.json file and may be passed indvidually.

[
    {
        "name": "greeting",
        "weight": 0,
        "in": [],
        "out": [
            {
                "name": "identity",
                "weight": 0.33
            },
            {
                "name": "wheather",
                "weight": 0.33
            },
            {
                "name": "compliment",
                "weight": 0.33
            }
        ]
    },
    {
        "name": "identity",
        "weight": 0,
        "in": [
            {
                "name": "greeting",
                "weight": 1
            }
        ],
        "out": [
            {
                "name": "veracity",
                "weight": 0.33
            },
            {
                "name": "compliment",
                "weight": 0.33
            },
            {
                "name": "age",
                "weight": 0.33
            }
        ]
    },
    etc.

The different communication layers

AIML-Layer

The most fundamental, in a way hard-wird layer is the aiml-level. Here you just input a phrase - the bozt scans the respectives aiml-files and gives an answer

The paraphrase-Layer

This layer resembles the classical eliza-bot logique. Here the bots gets the input - analyses it grammatically - and gives a return.

The speechact.js file refers to the analysis and detection of speechacts. This includes a thougrough syntax analysis.

AIML combined with intentions

AIML

A typical aiml.json file looks like the corresponding .aiml type, with some minor differences. You have the typical wildcards, like * and ^, you have the srai parameter which redirects the phrase. You have also the reponse, stored in the template. It it is just one string, this will be taken as the desire4d input, it it is an array, it will be understood as a random list.

[
    {
        "pattern": "Hello *",
        "srai": null,
        "topic": "greeting",
        "template": [
            "Hi",
            "How is it going"
        ]
    },
    {
        "pattern": "Hi *",
        "srai": null,
        "template": [
            "How is it going, {name}, you old {nickname}?"
        ]
    }
]

The bot also supports aiml conditions.

bot.aiml.input ( { pattern: "What's up?", condition:  { mood: "sad" }  }, bot.session, bot.process_aiml);

The corresponding json looks like this:

    {
        "pattern": "What's up?",
        "condition": true,
        "template": [
            {
            "mood": "sad",
            "template": "I feeld really depressed"    
            },
            {
            "mood": "happy",
            "template": "Life is great"
            }
        ]
    },

Working with intentions

An intention file looks like this:

[
    {
        "name": "price",
        "weight": 0,
        "in": [],
        "out": [
            {
                "name": "registration",
                "weight": 0.2
            },
            {
                "name": "payment_mode",
                "weight": 0.2
            },
            {
                "name": "withdrawal",
                "weight": 0.2
            },
            {
                "name": "plausibility",
                "weight": 0.2
            },
            {
                "name": "info_text",
                "weight": 0.2
            }
        ],
        "aiml": "HOW_MUCH",
        "keywords": [
            "price",
            "how much",
            "pay"
        ],
        "excludes": [
            "NEGATION"
        ]
    }
    etc. 
]

Each node of this consists of ins and outs - that means reference to other nodes. The nodes - taken as a whole - form a Markow chain.

Building such a chain is quite easy, and there some methods in the bot.intentions.intentions object.

    var obj = bot.intentions.intentions;

Here you add some nodes

    obj.addNodes(   ['price',
                    'registration', 
                    'payment_mode',
                    'withdrawal'
                    ]);

This defines a node in detail

    obj.characterize("price", {
        aiml: "HOW_MUCH",
        keywords: ["price", "how much", "pay"],
        excludes: ['NEGATION']
        })

With the aiml parameter you can set an aiml pointer. When the user input contains some of the keywords, the ususal chain will be overriden and the intention process is privileged.

With this function you create a Markow-relation between different nodes.

    self.addRelation("price", ["registration", "payment_mode", "withdrawal"]); 

Bot interaction

bot interaction is quite easy. As a first parameter you pass the input, then the session_id (that will be returned after you have started a conversation, then a callback)

bot.input("Who is the leader of the seminaire", null, function( data, error ) {

    if (data) {
        console.log( data );
    }

});

Syntax detection

The bot has an inbuilt syntax detection which allows to detect simple sentences (including adverbial phrases and relative sentences)

Conversation Strategies