Inspiration

Since mobile voice assistants are currently used by many people, the improvisation of this system is needed. Sometimes people get bored with a plain voice assistant or flat chatbot. Also for some cases, most of all voice assistants are too general, which can't provide specific cases with more details.

What it does

SOVA is an intelligent assistant with 3D visuals that can be more entertaining compared to other plain text based or just voice assistants. You can interact with this virtual assistant using your voice and it will respond to you and act like a human being. This application also specifies some cases so it can be more detailed if we want to ask about something. Example : for this prototype version, we have our first case "COVID-19 Assessment". You can ask SOVA to help you on COVID-19 assessment in more detail, and it will try to calculate the result and give you some suggestions after the session.

How I built it

  1. Wit.ai as our intelligent system for Natural Language Processing
  2. Unity as a tool to build the application
  3. iClone to create and modify the 3D Avatar
  4. Facebook SDK for login needed.

Challenges I ran into

Maybe we can say about integrating API from Wit.ai to Unity, we need to do some adjustment regarding the JSON file. Also to make the Natural Language Processing can make sense even more, because we have to keep training it so it can recognize what our intentions. This is the first time we built NLP from scratch so maybe it's not that really good for now, but we are still improving it.

Accomplishments that I'm proud of

Happy to accomplish an Artificial Intelligence project with a visual assistant like this, and this can help people in some ways.

What I learned

  • Learn about Wit.ai itself and learn how Natural Language Processing works.
  • Learn how to integrate Wit.ai and Unity (How they communicate with each other)

What's next for SOVA

  • If possible we will improve this more accurately and add more cases like education, financial and other fields.
  • Improve how the avatar will interact with the user (expression, animation etc)

Built With

Share this project:

Updates