Institutional Repository of Vilnius University
Informacijos paieška papildyto generavimo (RAG) sistema čekių informacijos paieškai ir išgavimui /
Abstract
dc:descriptionThe bachelor’s thesis examines the problem of searching and analyzing information from electronic receipts and develops an information system based on the Retrieval-Augmented Generation (RAG) principle, which allows users to ask questions about their purchases in natural language. Currently, in order to find specific information from electronic receipts, it is often necessary to open each document individually, manually search for the required data, and summarize it. This process is time-consuming, increases the likelihood of human error, and practically prevents the efficient analysis of larger collections of receipts. The problem addressed in this thesis is how to automate receipt processing and enable users to submit natural-language queries so that answers are generated based on the uploaded documents. The aim of the thesis is to develop a RAG system for searching and extracting information contained in electronic receipts. The developed system allows users to enter queries in natural language, and by combining anonymized receipt summaries with generative language models, it provides clear and understandable summaries of the answers. The objectives of the thesis are - to analyze the application domain and similar solutions; to define the functional and non-functional requirements of the system; to analyze and compare the cloud services of AWS, Azure, and Google Cloud, as well as possible OCR solutions; to design the system architecture and data model; to implement the selected solution; and to evaluate its performance. The developed information system allows users to upload receipts in PDF format, automatically recognize, anonymize, and structure them, and subsequently submit natural-language queries and receive summarized answers about expenses, products, time periods, or other relevant information. The system does not use a separate vector or semantic index: the text extracted and processed from receipts is divided into structured summaries, which are stored in a database and, at query time, are directly passed to a large language model as context. This reduces solution complexity and operational costs while maintaining sufficient answer accuracy at the prototype scale.
Degree
thesis:*- Grantor dc:publisher
- Institutional Repository of Vilnius University
- Year dc:date
- 2026
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Grigas, Kristupas,
Rights
dc:rights- Statement dc:rights
-
- info:eu-repo/semantics/openAccess
- Language dc:language
- lit
Identifiers
dc:identifier.*- Repository record dc:identifier
- https://repository.vu.lt/VU:ELABAETD277688886&prefLang=en_US