NAV Navbar

Testing something out

Triniti.ai allows you to create applications for users to be able to converse with triniti engine. You can create varied ranges of chatbot such as a pizzabot, a carbot, a phonebot etc. To learn how to create a chatbot from scratch for your company’s online services, click here


Overview

Create intelligent bots without coding - A how to guide

This tutorial shows you how to build a simple triniti.ai workspace, walking you through the most important features of triniti. You'll learn how to:

* Integrate with the Google Assistant, which lets you deploy your triniti.ai agent as actions that users can invoke through the Assistant.

How to create a triniti.ai account

This page describes how to create and log in to a triniti account.

Create a triniti.ai account

Login to the triniti.ai account

You can directly login to the triniti portal if you are already a user. If not, login to the portal by following these steps.

You have now successfully logged into the triniti portal. Refer Extract parameters


Create your first triniti.ai workspace

This page describes how to create and try out your first triniti workspace.

Note: Before you start, make sure you've created a Triniti.ai account.

Natural Language Understanding Workspaces are best described as NLU (Natural Language Understanding) modules. These can be included in your app, product, or service and transform natural user requests into actionable data.


Extract parameters


Concepts

Intents:

Probably all kind of sentences should have an intent. Intent describes an action that should be done. They can be called as the goal or purpose of what the user says to the bot.

For example,

How to create intents:

Naming intents is for the bot to understand that the particular user’s utterance’s “intent” is to order.This will enable the bot to convert this piece of information into acctionable data.


User says:

Different variations of how the same question can be asked by a user. The more variation is added, the better the agent can understand. To order a pizza (Intent-order), a user can say

I want to order a pizza Margarita with extra cheese. Can I have a small Margarita, please? A pizza Margarita, thanks. I want a Margarita, medium. I would like a Margarita. I am in a mood to have a margarita with extra cheese Give me a margarita please Your margarita tastes good. I would like to have one of that now. Margarita large, please. One margarita. I would like to eat a margarita now with extra cheese. I am hungry.I need some Margarita I would love to have a margarita I am in for a margarita Can you bring me one margarita with extra cheese thin crust Ordering one cheese burst margarita.

These sentences have to be fed in the triniti engine under the console option Intents>Add new intent>user says with entities inside each of these utterances, if any.


Entities:

Entities are a general term which will have many specific values to it. An entity represents a term or object in user says, that provides clarification or specific context for a particular intent. By recognizing the entities that are mentioned in user says, the Triniti service can identify the required parameters to fulfil the action.

Entity - representation

Entities have to be mentioned within curved brackets {}. Keep in mind while naming entities you can use letters, numbers, underscore and dashes but not spaces.

Example:

‘I would like to eat a margarita now with extra cheddar cheese’.

Here, the specific values for the intent "order" are margarita and extra cheddar cheese.These are henceforth called values for the entities {Pizza-name} and {Extra-Topping}

Entities and values corresponding to a pizzabot will include,

Entity 1 = Pizza-Name Entity 2 = Extra-toppings Entity 3 = Sauce
Margarita Feta Pesto
Buffalo Chicken Pizza Goat cheese Bechamel
BBQ Bacon Cheeseburger Pizza Mozzarella Pumpkin Pizza Sauce
Meat Lovers Pizza cheddar Thai Chili Sauce
Premium Garden Veggie Pizza Caramelized onion Curry Sauce
Pepperoni Pizza Pepperoni Tandoori
Italian Meatball Pizza Bacon Alfredo Sauce
Chicken-Bacon Parmesan Supreme Pizza Buffalo chicken Sun-Dried Tomato Sauce
Bacon Spinach Alfredo Pizza corn Wasabi Sauce
Mozzarella Crust Pizza Pineapple Thai Peanut Sauce
Supreme Pizza Jalapeno Zaatar
Entity 4 = Crust-Thickness Entity 5 = Size
Thin Crust Medium - 10" / 25 cm
Authentic Wood Fired Crust Italian - 12" / 30 cm
Thick Crust Large - 14" / 35 cm
Stuffed crust Family XXL - 16" / 40cm

Hence, while typing out user utterances under intents, you should mention the entities along with them.

‘I would like to eat a {Pizza-Name}now with {extra-toppings}’ is the correct input .

Each entity can have one to many values. Once you add an entity, it will get saved under use-cases.The ADD+ icon appears as soon as you open the curved bracket.

One most important thing is to note how you frame the user utterances for a particular intent. For a given intent you should specify as many user utterances as you can with all the entities in it.

Types of Entities

Entities in triniti.ai are of three types

Dictionary based entity.

Entities that are there in a dictionary. It mainly contains text such as the person(Adult or infant), names-of-restaurants, names-of-airlines, size-of-pizza etc.You need to add value to the entity once it is created. Values can include only a finite set.

Pre-Built entity.

For your convenience, all the infinite set values such as numbers,date,time and money and finite set values such as any city,state or country in the world will be pre-built by us. These are called pre-built entities. You just have to use them in your utterances. To view and download pre-built entities, click here.

Regular expression based entity.

This will contain all the detection logics where identification can be done using patterns or regular expressions. For example, email, phone_number, PNR, Pan card number etc.

Prompts

A prompt is used whenever a bot needs input from the user .

For example, for a pizzabot, Answer: Hey, welcome to triniti’s ‘kenpizzapoint’. I am triniti. May I take your order? Click on the link below to get to the menu card.

Whenever you need input from a user, you can send a prompt, wait for the user to respond with input, and then process the input and send a response to the user.

How to add prompts

As you add entities to an intent, it will keep them saved below use cases. Click on the entit one by one and mention their respective prompts and its user says.

Let's go back to our pizzabot example.

When the user says, ‘I would like to eat a margarita now with extra cheddar cheese’, 'Extra large'

Now give a minimum of 20 utterances/user says for each prompt.This is just for cross checking. Give utterances with the missing entities.

Invalid prompt

If your bot does find an entity, but it’s not present in the list of values of your entity type, the invalid prompt is reached.This bot only serves pizza, so if a user says “Hi, I want to order a hamburger”, the bot would know hamburger is not a valid entity.


Small talk:

You may configure to define responses for little chitchats that the user might want to have.

How to train the bot for small talk. * Click Small talk in the left console * Click on add new catagory . * Type invocation(say). * Press enter. * Click on the catagory invocation and add its relevant question and answers. * Click save .

You can repeat this process for as many catagories as you want. Type all the possible user asks in question and bot’s answer in answer.

Here are some examples for small talk categories with examples

Anything else

#####Example1: Question: How are you? Answer: Wonderful as always. Thanks for asking.

#####Example2: Question: You're so sweet. Answer: Thanks! The feeling is mutual.

End conversation

#####Example: Question: Thank you. Good bye. Answer: Thank you for choosing us. We hope you enjoy your meal. See you again. Take care. :)

Small talk is used to provide responses to casual conversation. This feature can greatly improve a user's experience when talking to your user.


Project types

Custom Workspace

Before entering into this section, please look into concepts and have a thorough understanding of intents, entities, small talks, prompts and user says

The Triniti console

When you click on the custom workspace that you created, you should see the Triniti console and the menu panel on the left.

For a custom workspace, you will see custom project type console. Refer to the right side diagram.

How to develop a Custom workspace


FAQ workspace

The Triniti console

When you click on the FAQ workspace that you created, you should see the Triniti console and the menu panel on the left.

For a FAQ workspace, you will see FAQ project type console. Refer to the right side diagram.

FAQ knowledge bots may simply aim to answer frequently asked questions. Questions having static response comes under FAQ workspace. FAQ bot has the built-in ability to scrape questions and answers from an existing FAQ site, plus it also allows you to manually configure your own custom list of questions and answers.This is different from custom workspace in a way that it does not convert user requests into actionable data.It simply gives the data.Hence, intent and entities are not a part of Faq workspace.

FAQ workspace sample question and ansers,if it were a pizzabot

Example questions Example answers
When will your shop open? The shop opens at 7:30 in the morning. Shop is closed on saturdays.
Do you accept Paypal? Customers will need to choose PayPal as a payment option in Kenny’s pizza point's app, which only allows customers to use PayPal or Kenny’s pizza point gift cards to make purchases. Kenny’s pizza point does not currently accept BitPay
Is Kenny’s pizza point delivery free? Kenny’s pizza point Offers 'Free' Delivery (But There's A Catch).Kenny’s pizza point has announced it will no longer charge a delivery fee for online orders, which can save customers up to $8.
How many slices are on a large Kenny’s pizza point pizza? 8 slices. Medium and large pizzas from Kenny’s pizza point both have 8 slices, but the large slices are obviously bigger.Kenny’s pizza point has awesome offers going on, did you know?
How big are Kenny’s pizza point pizzas? A medium pizza is 12 inches in diameter, or about 113 square inches. A large pizza is 14 inches in diameter, or about 154 square inches. A PANalicious (rectangular) pizza is 8.5 x 12 inches, or about 102 square inches.

How to develop a FAQ workspace

You can also export or import the products along with the question sets using the two small square icons to the top right.

Generate Variants

We provide additional tools to help you create content better. ‘Generate variants’ which appears when you click on ‘Add question set’ is incredibly helpful in a way that, it can be used to produce different ways of forming a single question.This helps you save a lot of time and also lets you create a more user-friendly bot. It is also capable of generating the question by itself if you provide just the answer.


Deploy and Launch


Customize your app

You might want to customize the way your chatbot looks on the webpage and the exact brand colour scheme is important to you. For this, choose Customize on the left sidebar. Customize lets you choose your template, create your title name and welcome message, choose primary colour and finally your logo. Your nice looking bot is now ready to function. (The web SDK source code is available in github if you wish to customize templates.)


Make Public

Making it public is a very simple thing to do.Click Publish on the left hand side bar. Note that every deploy creates a snapshot of the existing workspace which can be reverted to at any time. You can choose a specific snapshot and make it live/test it within the workspace. Make public lets you access your bot beyond Triniti.ai.You can also test different versions after going live.


Enable Channels

The channels include many popular services, such as Facebook Messenger, Slack, Skype,Web HDK and several others. It is very easy to assign a bot to a channel. Click Channels option on the left hand side bar and click the channel through which you want to make your bot go live. WebSDK, facebook messenger and telegram are pre-configured for you in triniti.ai.

Follow these instructions to connect your workspaces to your web SDK, facebook messenger and telegram bot.

Telegram

WebSDK

This is a simple web chat. Share the URL to your webpage. Note that it is just a demo and not a real chat.

Facebook messenger

You have now successfully connected your bot to channels.