Inspiration

Have you heard of the fact that one in twenty Indians suffers from depression and it makes it worse to know that more than 300 million people worldwide suffer from depression, and in some cases, it gives very bad results for the mankind? Also, some introverted personalities like to get rid of their boredom with things that provide them with some information about the things they love. So what is the solution for this? A human needs to talk to someone so that he can get rid of boredom. But is chatbot a better option? Yes of course! But most of it has a time-consuming typing process that makes the process less interactive and even more boring, they do not recommend and inspire you to meet new people who have similar interests. Or those who do similar things like you. So here we are bringing more interactive AI-powered bot which will overcome this problem.

Why Speech bot

  1. Automated Support For Similar Queries & conversation
  2. Save Human Resources for Qualitative Tasks
  3. Better User Interaction
  4. Speechbot- Easy-to-use
  5. Cost-effective and Time-efficient

What it does

Wit Chat AI is a web-based AI bot. This gives you the option to talk on some topics. After login up through your Facebook / Twitter account, it shows you some interesting topics to interact with WIT Chat AI. You can interact through speech. This makes you feel genuine human interaction on the other side. Also once you click on Connect Friend Card. It will sync your following friends according to your sign in option whether it was Facebook or Twitter which will show you in the recommended friends list. Finally, the user will feel more connected to his friends and will have a list given by the bot, through which he can initiate chat. WIT chat gives you a better option to ask stock related questions in real-time. So it is a complete package as a web-based solution and is superior to some other voice assistants and is free and easy to use.

How we built it

We are using Wit.ai to find the intent of the question. This gives us the confidence of the question asked. Once the user clicks the mic button. It records the voice for a few seconds and sends a voice to the server. Where server-side we are using wit.ai API to fetch the intent of the query. Once we get the real meaning of the question. We are using a MySQL database to give them the proper answer. The answer will come in the form of text and we are using the gTTS library to convert that processed text into speech. Wit Chat provides various cards as an option:

  1. Music
  2. Sports
  3. Stocks.
  4. Movies.
  5. Connect people Also, we wrote an algorithm to recommend similar interest friends to whom he/she can start a discussion. it will continuously look to the bio of users and filter out the friends according to the interest and field.

Challenges we ran into

Overall the project building was a bit challenging in lockdown. The challenges we faced are as follows:

  1. The main challenge we faced while working on this is how users can communicate through speech to the WIT. That was something new for us to divide the work and to come up with good logic.
  2. Due to lockdown, it was a completely remote project we have worked on. We have to take several zoom calls to decide the approach.
  3. Less time was also a pressure, we have to complete our office work and then take a good start after office hours and work over the weekends.
  4. The feature we want to enable lip sync while WIT processes the answer which was a bit difficult for us. We ended up with a good alternative that makes the conversation human friendly.

Accomplishments that we're proud of

A great team that has worked within the stipulated time frame was a good achievement for us. Also, we think that we have actually done a good job in deciding the architecture and flow of the code. This enabled us to make good command of Django to add features like social authentication and JS. This was a very good part of the project and that made us excited to see this project up and running in the end.

What we learned

We can say that this was the first project we worked on together and we were completely new to the secret management part, that way we would manage tokens that we should not uncover the code. We found that configparser, a better way to hide secrets and furthermore we have used our Django skills and competitive programming skills to write a good backend logic. That has ended up as completion of this project.

What's next for WIT CHAT AI

We want to do a little work on the UI part and make a feature in the WIT chat that will hold the intent and send a greeting message to the most priority user and how will we decide the user's priority? Yes, we have to write an algorithm for this. It will answer who is the most appropriate user to whom it can take the initiative to talk. Even WIT will initiate a chat on behalf of the user. Also, we want to deploy it so the people can take advantage of the WIT.

Built With:

· Wit-ai · Django · Django-social-auth · Twitter-friends-API . Facebook login API · Django-rest-framework . JS . HTML5 , Bootstrap3 . gTTS, Pyaudio

Github repo

Built With

Share this project:

Updates

Submission history