Enhancing Trust Factor Identification in E-Commerce: The Role of Text Segmentation and Factor Extraction with Transformer Models
Bandar F. Alkhalil, Yu Zhuang, Khalid T. Mursi, Ahmad O. Aseeri · 2025
Online reviews are increasingly crucial as a data source for companies to understand consumer opinion. This study evaluates the impact of text segmentation on the accuracy of pre-trained transformer models such as GPT-3.5, BART, and BERT in analyzing consumer trust factors from online reviews in E-commerce. Our methodology involved manually labeling a dataset comprising 1,499 unsegmented and 4,528 segmented reviews. This process allowed us to compare the effectiveness of the employed models and assess how text segmentation influences the extraction of trust factors. The results demonstrate a substantial $\mathbf{1 2. 1 9 \%}$ increase for GPT-3.5 in model accuracy with segmentation, enhancing the models’ ability to discern nuanced textual elements. Key factors such as product quality and satisfaction with Size/Fit were identified as crucial factors in influencing consumer trust. This study demonstrates how segmentation can be utilized to comprehend consumer opinions in the E-commerce market. It indicates that text segmentation improves NLP applications in digital marketplaces by identifying factors that impact consumer trust, laying the groundwork for future advancements.