A carregar...
Publicação
Improving an Employee Stock Ownership Plan Communication Strategy with GenAI: Development of a GenAI Chatbot for an Employee Stock Ownership Plan in a major aerospace company
| Nome: | Descrição: | Tamanho: | Formato: | |
|---|---|---|---|---|
| 4.13 MB | Adobe PDF |
Autores
Orientador(es)
Resumo(s)
This thesis presents the development of a Retrieval-Augmented Generation (RAG) chatbot to
support internal communication of the Employee Stock Ownership Plan (ESOP) at a major
aerospace company. ESOPs often involve complex documentation and regulations, leading to
difficulties in understanding and low employee engagement. To address this, a Generative AI
chatbot was built using Google Vertex AI, LangChain, and Gemini large language models. The
methodology included cleaning and analysing over 100 ESOP PDF documents, splitting them
into chunks using optimal configurations, embedding them into semantic vectors, and testing
different retrievers and generative models. Evaluation was conducted using the RetrievalAugmented Generation Assessment (RAGAS) framework across three metrics: faithfulness,
answer relevancy, and noise sensitivity. The best-performing configuration combined hybrid
retrieval with the text-embedding-005 model for embedding and gemini-2.0-flash-001 for
generation. In addition to quantitative metrics, a qualitative user study with ten employees
was conducted to assess usability and perceived helpfulness. Results demonstrate the
chatbot’s potential to enhance information accessibility, reduce dependence on HR teams,
and improve overall user experience.
Descrição
Dissertation presented as the partial requirement for obtaining a Master's degree in Data Science and Advanced Analytics, specialization in Data Science
Palavras-chave
Retrieval-Augmented Generation Large Language Models Chatbot Generative AI Employee Stock Ownership Plan SDG 9 - Industry, innovation and infrastructure