Development of a virtual conversational agent by using Rasa and Unity frameworks

Avtor: Nikolina Simjanoska

Mentor: izr. prof. dr. Matej Rojc univ. dipl. inž. el.
Co-mentor: Daniel Hari mag. tk.
Degree: 1.
Date: june, 2022
DKUM: NIKOLINA SIMJANOSKA

Abstract

The graduate project involves the development of a virtual conversational agent. Here, we present what conversational agents and tools are needed to develop such systems. In the following, we also present the Rasa framework and project development on the Unity platform, which is often used to create interactive 3D content. The virtual conversational agent uses a pre-prepared scenario that includes information about the DSPLAB laboratory, at the Faculty of Electrical Engineering, Computer Science and Informatics, University of Maribor.

Conversational agents

With the development of technology, new communication approaches began to be developed and used. People communicate on a daily basis through instant messaging platforms, such as: Messenger, WhatsApp, SMS, Viber, Instagram, etc. In the context of new man-machine communication methods, a chat solution using virtual conversational agents, also named Chatbot systems, is becoming increasingly useful. Such systems are supported by Artificial Intelligence (AI), which helps to fully automate conversations and make them more and more natural. The chatbot system includes a conversational interface based on natural language processing and machine learning. Virtual agents are available through the various channels we already use on a daily basis. They simulate conversation with the user, are able to understand the text, i.e. speech or both, interact with users and thus simplify and speed up people’s daily activities. They also provide several benefits: 24/7 availability, better data collection capabilities, improved user experience, lower labor costs, etc.

Chatbot systems are software solutions designed to engage with customers in the most natural way possible. The chatbot system simulates and processes human conversation, thus allowing users to interact more naturally even with digital devices, and to communicate as with a real person. They are, therefore, used as a medium for interaction. Tehre are several types of Chatbot based systems: avatars, mobile terminals, customer support, banking systems, chat system, website system, branding system, consumer systems, travel systems, sales systems, sales support systems, education, etc.

Virtual agents (Intelligent Personal Assistants (IPAs)) are online personal assistants who help people with their day-to-day activities, such as managing their email, scheduling meetings, finding locations, providing information globally, and so on. Virtual agents access a variety of information from a variety of sources. IPA understand the user’s intent to provide tailored and meaningful responses. Therefore, they use NLP user interfaces in conversations with the user. The user no longer necessarily clicks on the content itself, but can also use speech. Virtual agents need to have high quality input data, otherwise they cannot provide the best possible answers. Prior to developing a virtual agent, it must be ensured that the datasets are as complete, accurate and up-to-date as possible. Otherwise, users will not have a positive experience with the virtual agent.

Development of a conversational system for the DSPLAB laboratory

Rasa is an open-source platform for the development of artificial intelligence-driven conversational systems. Extends the capabilities of bots and enables machine learning. It is a very powerful framework and nowadays it serves developers all over the world, to build chatbot systems and virtual agents. With its help we can easily create virtual agents and integrate them e.g. with website, Facebook, Whatsapp app, etc. Rasa for development supports Python programming language and Natural Language Understanding (NLU). It provides control over the NLU process, uses machine learning libraries and deep learning frameworks, such as TensorFlow and Keras, and support users creating Python scripts themselves.

Rasa allows to check and test dialogues from end-to-end by reviewing test stories. Testing a model on stories is also the best way to see how our agent will act in certain situations. Test stories allow to provide complete conversations and also to test whether the agent will behave in the desired way for any input from the user. Test stories include already implemented stories from scripts, and a user message.

Unity is a multi-platform game development environment developed by Unity Technologies and first published and released in June 2005 at the Apple Worldwide Developers Conference as a Mac OS X gaming environment. Unity now supports a variety of desktop, mobile, console platforms as well as virtual reality platforms. It is especially popular for developing mobile games for iOS and Android. The platform can be used to create three-dimensional (3D) and two-dimensional (2D) games, as well as interactive simulations and other experiences. The platform is also used in other industries, such as filmmaking, automotive, architecture, engineering, construction, etc. Unity supports various platforms: Mobile platforms (iOS, Android), Desktop platforms (Windows, Mac, Linux), Web platforms (WebGL), Console platforms (PlayStation, Xbox, Nintendo Switch, Stadia), Virtual / augmented reality platforms, PlayStation VR, Apple ARKit, Google Cardboard). Unity supports C# programming language. To write the script and code, we used Microsoft’s Visual Studio environment for the project.

Communication between all modules is performed using the standard REST – Representational State Transfer protocol. REST is based on web standards and uses the HTTP protocol. Each component is a specific resource, and the resource is accessed using standard HTTP methods. HTTP or Hypertext Transfer Protocol is an application layer protocol for transferring hypermedia documents. It is also used to communicate between web browsers and web servers.

Discussion and results

Virtualni pogovorni agent z uporabo ogrodij Rasa in Unity.

Development of a virtual conversational agent by using Rasa and Unity frameworks

Discussion and results

Laboratory of Digital Signal Processing

E-mail

Contact

Address

Phone