I'm sure that simulating emotions is important for creating realistic Virtual Humans - a field of study I was interested in. After reading some of The Cognitive Structure of Emotions - also known by computer scientists as the OCC model -, I had contact with the EMA model of emotions and discovered it was implemented in the Soar Cognitive Architecture. 

This project aimed to implement a Soar agent that could survive by itself in a Stardew Valley-like game. The motive at first was learning how to create an intelligent agent capable of switching between many different tasks according to the current context, and later, to implement a model of emotion inspired by EMA, to make the agent's decisions more human-like. This project was also a Scientific Initiation I proposed and my Graduation Thesis.

There is a simple demo of my integration between Soar and Unity in this repo (Soar is in C++ and Unity is in C#). The repo where the project itself is being developed is currently private, but one can check the demo videos listed below.

• Game Dev Video: The player can plant, water, grow and harvest a potato-like plant. It also has food and life bars, and can eat the potato. If hungry, he will lose life points. If he eats the potato and recovers food points, life will also get recovered.

• Soar Agent Video: The integration Soar/Unity is working not only in the sense that the Soar's C++ code is working inside Unity (C#), but also the life cycles of both tools are well integrated.

The agent gets items on the map inserted into its working memory. It detects the seeds and sends a command to move into their position. After food goes below 100% it starts the plant routine, digging the soil for the seeds it has, planting, and watering then for 3 days. It will collect the potatoes, and eat, then after food get below the food recovery the potatoes provide. 

Why: A professor asked me if I wanted to enter a project about Virtual Humans in partnership with the administration course. It was about a chatbot to sell products.

How: First, I created a simple library in Dart (language) to deal with State Machines, which you can check in this repo. After that a Flutter (Dart front-end SDK) interface can be generated based on that state machine lib to generate a chatbot, the code can be found here.

     There was also the intention to turn this into a web service, where a user could create the state machine online and expose it in a URL in the form of a chatbot. The attempt involved me learning and teaching Fast API (Python backend) to some colleagues, and got documented here.

I studied some ML and NLP on Coursera and coded a Twitter sentiment analysis bot by using the sentiment analysis algorithm learned in the NLP course.  

Machine Learning Course: You can check the certification by accessing this link. Also the code I produced by studying it can be found here.

NLP Course: You can check the certification by accessing this link. Also, the chatbot code can be found here.