VIRTUAL ASSISTANT IZAN
BREVE DESCRIPCIÓN DEL PROYECTO
Izan is a virtual chatbot that aims to assist customers in the purchasing process. It features a graphical interface based on an Avatar residing in a passive house.
Detalles
del proyecto
Nombre: Assisting customers in the purchasing process through conversational AI
Cliente: E3 Windows
Localización: CT Sevilla – Customer in Vélez Málaga
Año de desarrollo: 2022-2023
Descripción narrativa del proyecto
IZAN has been meticulously designed to provide accurate and detailed answers to any question related to our company or its products. The graphical interface of this chatbot has been carefully built using the Unity platform, thus offering an attractive and immersive visual experience for our users. The avatar, integrated into a passive house, not only serves as a unique visual metaphor, but also reflects our commitment to sustainability and energy efficiency.
From a technical point of view, the chatbot is based on RASA’s natural language processing technology, supported by BERT language models to understand and respond in a contextual and accurate manner. In addition, we leverage the potential of OpenAI to further enhance its responsiveness and comprehensiveness, thus providing a more complete user experience.
This project represents the convergence of industry-leading technologies, merging the power of Unity, RASA, BERT and OpenAI to deliver a virtual chatbot that goes beyond simple interactions. We immerse users in a unique graphical environment while providing them with detailed information about E3 Windows and its products.
Alcance de los trabajos de Ghenova
The project integrates various technologies and specialized skills to achieve its operation. We use Unity as a graphic interface, taking advantage of its experience in videogames and visualization (VX). The chatbot, developed with RASA and OpenAI, is complemented by a 3D design of a real passive house provided by a specialized architect and textured by Ghenova Digital. The Avatar, created from scratch at Ghenova Digital and based on the robot from Ex Machina, adds a unique and visually stunning layer to the set.
In the graphic design area, we used 3DMax, while the programming of the interface in Unity was carried out using the C# language. As for the backend and Natural Language Processing (NLP) models, we opted for Python. The API was structured with Flask, and the coordination of the chatbot was entrusted to RASA, a framework specialized in the creation of chatbots.
The different phases of the project were carried out as follows:
– October 2022: Initially, we approached the chatbot project with RASA, incorporating predefined questions and predefined speeches. At the same time, we created the Avatar and developed the Unity interface, which integrated the house, voice recognition, avatar and API connection. All this was implemented locally on a PC.
– May 2023: Improved Izan’s performance and comprehension capabilities through a parser and the application of techniques learned from previous projects.
– November 2023: Implemented Image Recognition (IR) to answer questions about the manual. We brought the chatbot to AWS for greater accessibility and connection to OpenAI to answer random questions. We also adapted the interface to connect to the AWS API.”
In the graphic design area, we used 3DMax, while the programming of the interface in Unity was carried out using the C# language. As for the backend and Natural Language Processing (NLP) models, we opted for Python. The API was structured with Flask, and the coordination of the chatbot was entrusted to RASA, a framework specialized in the creation of chatbots.
The different phases of the project were carried out as follows:
– October 2022: Initially, we approached the chatbot project with RASA, incorporating predefined questions and predefined speeches. At the same time, we created the Avatar and developed the Unity interface, which integrated the house, voice recognition, avatar and API connection. All this was implemented locally on a PC.
– May 2023: Improved Izan’s performance and comprehension capabilities through a parser and the application of techniques learned from previous projects.
– November 2023: Implemented Image Recognition (IR) to answer questions about the manual. We brought the chatbot to AWS for greater accessibility and connection to OpenAI to answer random questions. We also adapted the interface to connect to the AWS API.”
