Institutional Repository of Vilnius University
Verslo galimybių atitikimo sistemos kūrimas ir tyrimas, taikant LLM išvestas vektorines reprezentacijas /
Abstract
dc:descriptionThe small business sector faces significant challenges in finding new business opportunities, partners, and clients. Thesis develops and investigates a business opportunity matching system based on Large Language Model (LLM) derived vector representations (vector embeddings) to fa‑ cilitate this process. Literature review revealed the advantages of vector embeddings and semantic search compared to traditional keyword‑based search methods. The paper proposes a methodolo‑ gy for transforming unstructured business descriptions into structured data using GPT‑4o‑mini model, which are then converted into vector embeddings. Two OpenAI embedding models (text‑ embedding‑3‑small and text‑embedding‑ada‑002) and two re‑ranking methods – Cross‑Encoder and Field‑aware Neural Ranking Model – were experimentally compared. The best results we‑ re achieved using the text‑embedding‑3‑small model with Cross‑Encoder (82.13% accuracy), although the Field‑aware model (78.80% accuracy) was more computationally efficient. The de‑ veloped system effectively identifies business entity matching based on their needs, offerings, industry, and products, thereby helping small businesses find potentially beneficial partnerships.
Degree
thesis:*- Grantor dc:publisher
- Institutional Repository of Vilnius University
- Year dc:date
- 2025
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Kaveckas, Vasaris,
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:ELABAETD239031038&prefLang=en_US