Problem Description

As globalization continues to make the world more multilingual, the demand for language support has increased. However, it is impractical, for example in customer service centers, to meet this need effectively. Concurrently, conversational agents have become more sophisticated and commercially available, yet they often struggle to engage meaningfully and offer appropriate translation assistance.

The ability to manage multilingual interactions through conversational agents is essential. This involves not only understanding and speaking different languages but also determining the optimal moments to utilize these translation capabilities. To address these challenges, a system for a conversational agent hast to be implemented that leverages a Large Language Model (LLM) to manage multilingual interactions. The system should aim to improve the overall user experience.

Goal of Thesis

The primary goal of this thesis is to design and implement a multilingual dialog system for a conversational agent to manage multilingual conversations. The specific objectives are as follows:

• Conduct a comprehensive literature review on translating dialogs in conversational agents
• Design a conceptual dialog system utilizing a Large Language Model (LLM).
• Preferably integrate the LLM into a conversational agent.
• Evaluate the designed system:
  • For a Bachelor's thesis: Focus on qualitative evaluation, including usability testing and participant feedback.
  • For a Master's thesis: Conduct both qualitative and quantitative evaluations through experiments, assessing the system's impact on user experience.

Requirements

• Experience with Generative AI models and their integration into applications.
• Basic skills in programming in Kotlin, Java or C#
• Understanding of the Design Science Research methodology and ability to conduct user evaluations.
• Strong time management and communication skills, and proficiency in English.