Abstract
This thesis investigates the multifaceted impact of e‑commerce on customer satisfaction and trust, leveraging advanced Natura Language Processing (NLP) techniques and machine learning (ML) models. The research introduces three novel datasets tailored for Aspect‑Based Sentiment Analysis (ABSA) and Aspect Extraction (AE), derived from customer reviews on platforms such as Trustpilot. Key aspects identified include Shipping, Trust, Customer Service, Pricing, and Refund Process. These aspects are systematically analyzed to understand their influence on customer experiences.
State‑of‑the‑art transformer‑based models, such as BERT and RoBERTa, are employed to perform sentiment classification and aspect extraction. The study demonstrates that these models outperform traditional approaches, such as Support Vector Machines (SVM) and Naı̈ve Bayes (NB), by effectively capturing complex sentiment nuances and extracting relevant aspects from unstructured textdata.
In addition, the research integrates Fuzzy‑set Qualitative Comparative Analysis (FsQCA) with Large Language Models (LLMs) to uncover intricate causal relationships between various stages of the customer journey and trust outcomes. This novel combination provides deeper insights into how specific aspects of thee‑commerce experience influence trust.
The findings offer actionable insights for e‑commerce businesses, emphasizing the importance of improving key areas such as shipping, customer service, and refund processes to enhance customer satisfaction and trust. By leveraging advanced NLP techniques and ML models, this research provides a comprehensive framework for analyzing customer feedback and generating managerial insights from unstructured review data.
This thesis contributes to the field by providing a detailed analysis of customer feedback, demonstrating the practical application of cutting‑edge technologies in sentiment analysis (SA), and offering evidence‑based recommendations fore‑commerce businesses to foster customer loyalty and trust.
State‑of‑the‑art transformer‑based models, such as BERT and RoBERTa, are employed to perform sentiment classification and aspect extraction. The study demonstrates that these models outperform traditional approaches, such as Support Vector Machines (SVM) and Naı̈ve Bayes (NB), by effectively capturing complex sentiment nuances and extracting relevant aspects from unstructured textdata.
In addition, the research integrates Fuzzy‑set Qualitative Comparative Analysis (FsQCA) with Large Language Models (LLMs) to uncover intricate causal relationships between various stages of the customer journey and trust outcomes. This novel combination provides deeper insights into how specific aspects of thee‑commerce experience influence trust.
The findings offer actionable insights for e‑commerce businesses, emphasizing the importance of improving key areas such as shipping, customer service, and refund processes to enhance customer satisfaction and trust. By leveraging advanced NLP techniques and ML models, this research provides a comprehensive framework for analyzing customer feedback and generating managerial insights from unstructured review data.
This thesis contributes to the field by providing a detailed analysis of customer feedback, demonstrating the practical application of cutting‑edge technologies in sentiment analysis (SA), and offering evidence‑based recommendations fore‑commerce businesses to foster customer loyalty and trust.
| Original language | English |
|---|---|
| Qualification | Doctor of Philosophy |
| Supervisors/Advisors |
|
| Award date | 10 Jun 2025 |
| Publisher | |
| Print ISBNs | 978-952-12-4504-6 |
| Electronic ISBNs | 978-952-12-4505-3 |
| Publication status | Published - 10 Jun 2025 |
| MoE publication type | G5 Doctoral dissertation (article) |
Keywords
- Machine learnbing
- Aspect-Based Sentiment Analysis
- Customer Experience
- Customer Feedback Analysis
- Large language models