What a business wants and which tools can fulfill it?

Different features that you can add to your business later on before we jump on the next slide do you want to elaborate a little bit more as to what what fulfillment means yes sure so first i start with the whatsapp because last i just covered facebook for whatsapp also twilio offers whatsapp sandboxes where you can easily develop uh whatsapp chatbots and even there’s another latest tool called landboard.

Io where you can easily just create a flow of the conversation and you can deploy the bot for whatsapp and for website again we use dialogflow and you can easily uh integrate the dialogflow agent to a website using different tools like communicate or what copy and there are many other tools as well for twitter slack and telegram bots again we use the same thing for bot fulfillment um the main two uh programming languages that are used are uh node.js and python.

Basically you use uh bot fulfillment to add to create a web hook for your chat bot you can even create a chat button on dialogflow with zero coding but if you want to do uh some some extra stuff like uh like for example connecting with any third-party apis or maybe connecting with the database so for that you need to add a fulfillment framework to your bot and you can easily do that as well on dialogflow they do have a fulfillment section where you can easily deploy your code which gets deployed on the google cloud functions for firebase so that makes it easy as well again so this is mainly for the third-party apis like database connections.

So give us an example as to what that would look like and i mean in that case you need somebody who understand a little bit of programming the back-end side yes uh dialogflow does have a really good documentation on this and uh that’s where in fact i learned all of this from so uh so using that you can go through it once you can so they do have uh client libraries for node.js and python and you even have a client library for java too so you can use any of these whichever of these languages are comfortable with and then uh and mainly it is uh used to handle specific intents that would require uh maybe maybe if you want to generate a dynamic response to a particular query.

So for that we need to add the bot fulfillment because we cannot just return a plain text response back so so for example when they when i log in it will the backfill filament will talk to the database and lets the chatbot know that my name is medusa so the part will say it is high manager and then if i’m saving fulfillment yeah so whatever you’re uh so however you’re interacting with the bot all of those messages they keep on getting logged and then we send and then we’ve integrated our database to which it goes and then the data gets stored there right any questions.

Far is everybody confused yet so i know there was a gentleman who was talking about last week about building a platform for mental health um i mean i’ve seen some of those being done here on on the chatbot side um yes um and it’s it’s good because it’s like you’re talking to a friend you got that feeling in it so and yeah again don’t hesitate to stop and ask any questions if you guys have any for the mental health part even i’ve been recently working on a side project.

So it’s a bot called core relief which i’ve been making so basically it aims to help to help people during covet about their about any anxieties that they have about their future etc and it’s more like a friend that people can talk to because sometimes people actually do hesitate to get proper mental health support so with the chatbot they can receive uh such support because because it’s just a bot that they’re talking to right so once you’re done with the development stage as well the next step is to actually deploy it so whatever you’ve made either.

If it’s like um it’s your first product for your business uh do put that prototype on your website or show it to your clients if you’re making making it for your client then just ship it to them uh get them familiar with how they need to uh see the hotel see the customer logging etc and how they can uh basically um and then also based on that keep on getting feedback from the testing and how the client and other users have been interacting with the bot keep on improving and iterating over it.

So that was it for the design develop and deploy stage for any software product or any product literally so is there any question until now to show us a demo do you have one schedule do you have one set up uh you want to okay like so so yeah so what i mean is let’s say if you take dialogflow for example like the difference between the entities the the the intents and and how to set it up a little bit if that’s possible yeah from that so for example that big sense we we can demo it and we’re both familiar with it anything that you think you can think of sure just give me a second i’ll just share my screen once again okay.

So to give a quick update um so the way the bot understands is you need to define certain intents an intent being finding somebody’s name or telling somebody hello or getting somebody’s information so under that intent you can ask multiple questions and those for those questions you can add entities and entities can be name address email so for example if you were to write what is your name that question you need to define what that entity is and it’s tied to that intent so it’s i know it once you do it a few times then you’ll understand it then you collect a few intents.

It become a collection of intents that you can get somebody and you can categorize it as onboarding somebody to a platform of trying to help them with a setup question so the more intense you define overall you’ll have your bot becoming smarter every day right so um let’s see whether she can quickly demo that and this is available for free anybody can log in create an account start fooling around and then you’ll understand the more you do it my advice in most cases for a lot of founders is if you don’t know how to build a website try a bot first and see how that goes because that interaction itself is very intuitive for a user to understand so this is how the dial-up flow interface looks like.

This is how the dialog interface looks like so we’ve got so on the left panel we’ve got intents intent is um intent is basically you can think of intent to be a collection of different topics so actually this is a bot which can talk about different daily life topics so you can see a lot of daily life topics so so for that like we’ve added different intents to it that can handle like for example like it can talk about like like in the alien intent we’ve got different training phrases uh training phrases are basically the potential phrases that a user could use to uh chat about that particular topic like let me just open one of the intent so over here so actually this is just a demo uh agent so we’ve got a few training phrases so which is like can we talk about aliens etc.

Then we’ve got some responses filled up there so whenever you’re making a what you should try to add at least uh 10 to 15 at least 10 to 15 training phrases in one of in every single intent because that’s how your bot will be better at answering and so this is it for the intent so for every single topic for every single uh use case that we need to handle we create a separate intent.

This is the create intent button and then we’ve got entities entities are basically the specific keywords that we’re looking for in a sentence that a user is saying like we’ve made a few uh entities like you can see negative emotions let me just open that so so like we’ve got a few few words and there’s for the entities and for every single word you can define synonyms like the different words that the user could use to say that same thing so entities are just some specific keywords that you’re trying to analyze and understand from um from a from a query.

So if a user uses any of the words like okay i’m feeling really sad or whatever then you can understand that it’s a negative emotion so this is it for the entities and then this is the fulfillment section where you can uh create your own you can even create your own web hook you can host it uh on helico at anywhere and you can even use the inline editor which is powered by google cloud functions so with the inline editor you can just deploy the code over here itself.

This is a fulfillment code uh written in node.js which works along with um so basically we’re using the duckduckgo api over here to search for any other so if we so so basically how this particular agent is working is that if a user has any query about a topic that is not there already in our agent so that’s when it goes on to the fulfillment and like for example if the user wants to know about covet we haven’t changed the bot with kovid so for that for those sort of queries we’re calling the uh duckduckgo api which will then return a response from wikipedia about what code is and then this is the training section where you can see the logs.

Then you can just uh match it to the correct intent etc you can even see the history about how people have been interacting with the bot and there’s also an analytics tab so there has been no interaction with this actually in the past this particular agent in the past few days so we don’t have much of the analytics right now because the more people use it that’s when the analytics show and this is just a demo bot so this is how the dialogflow platform is actually like.

It’s pretty easy to use and get started with there’s also an option for small talk so small so if you enable small talk you can easily handle um basic queries like hi how are you or emotions hello goodbye thank you so you can just enable that with the tap of a button oh sorry okay this this boy so now it’s asking me how is a day going good nice to hear that tell me about the hobbies i love to paint okay there is so that’s good and now the interesting thing there is if you say hello.

Again say hello yeah um so you can see at the bar detects the context is the mood what you’re doing right so you can define all this thing and things and there’s a few things that are predefined already that you can use as examples that you can play around with um i was just selling out the context so contexts are basically used in dialogflow to create a contextual conversation like for example if you want to talk about some particular topics so you you’re going to have follow-up questions on that as well right so if you want to connect one conversation with another.

So for that you actually use context so that like for example if i’m just asking about the user’s hobby so if i say painting then it’s then it can ask me further more about painting that uh what do you like to paint what was the last painting that you made etc so for that i’m going to use context to make the flow a little bit contextual and make it sound a little more human like okay notations are for dialogflow it is free to use but there’s only a limited number of intents you can integrate i believe was 250 i don’t know whether the number went up yeah.

So yeah this is a free tool to use you can integrate with anything you want um and you can build up to 250 intense i think there’s a limit as to how much api calls you can do as well as certain premium built in pre-built agents that you can add so but it’s good place to start off and then as you start using it then you can implement it for the api calls you can even increase that like if you switch on to a firebase blaze plan you can easily integrate several third party api.

So with that plan uh that don’t get solved and yes the issue is that you cannot create more than 250 intense that’s the limit and i think 200-ish intense should be more than enough i mean the maximum intense that we’ve ever added would be around 198 to 200 so so i mean it’s i mean it’s pretty much fine until going fine until now but of course to scale the chatbots at a greater rate uh in the future we will be using separate uh using different tools etc and frameworks which will be and create our own infrastructure basically for nlp but that’s something for the future as i said and to get started dialogflow is very easy to use i mean anyone can just get started with it you don’t even have to i mean it’s not important to add the fulfillment code.

Leave a Comment