Smart Chatbot for Discovering Finnish Theses: A Conversational System Using RAG and Theseus Data
Year: 2025
Type: dissertation
Abstract: This thesis presents a conversational chatbot system for discovering Finnish university theses. It is using “Retrieval-Augmented Generation” (RAG) technology. The project was developed during an internship at Nokia Garage Espoo. It integrated into Nokia Stand, a collaborative platform showcasing Nokia and Metropolia University partnerships. The traditional keyword-based search on Theseus.fi, Finland's national thesis repository, requires exact questions and multiple attempts to find relevant content. This thesis addresses these limitations by implementing a RAG-based chatbot. The chatbot understands semantic meaning and enables natural language interaction, interpreting what users want rather than simply matching keywords. The system has two main parts. The first part is a mobile app built with React Native. The second part is a backend built with Node.js that uses RAG technology. The backend collects thesis information from Theseus.fi. It then turns this data into vectors using a model called all-MiniLM-L6-v2. The system stores 45 Nokia-related theses in its memory. When users ask questions, the system finds the right theses and creates answers using ready-made templates. This makes the system very fast. It gives answers in less than 0.5 seconds. All development, implementation, and testing were conducted in a local development environment. Testing showed that semantic search successfully retrieves relevant theses regardless of terminology differences, demonstrating that RAG technology improves academic search experiences compared to traditional keyword matching.
Source: Theseus (Ammattikorkeakoulujen)
Author: Kharel, Suraj
Language: English
Subfield: Information Systems
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W7113120400