NIU ZHAOHANG
Siti Soraya Abdul Rahman
Mr. Qin Heng
This study effectively utilized the
Fast-LCF-ATEPC model for Aspect-Based Sentiment Analysis of Alibaba user
reviews. Achieving high accuracy in Aspect Polarity Classification and Aspect
Term Extraction, the model showed an Aspect Polarity Classification accuracy of
91.06%, an APC F1 score of 91.07%, and an Aspect Term Extraction F1 score of
83.09%. This robust model was developed through comprehensive data handling,
including collection, preprocessing and postprocessing, data annotation, training,
and evaluation, along with a comparative analysis with state-of-the-art models
like Bert-Base. The findings reveal a mixed user experience on Alibaba,
highlighting both positives in app usability and pricing, and concerns in app
functionality and shipping services. Analysing user-concerned aspects can
enhance Alibaba online shopping platform's user satisfaction and its standing
as a leading e-commerce platform.
Keywords: Aspect-Based Sentiment Analysis, Alibaba
E-commerce Platform, Fast-LCF-ATEPC, Aspect Polarity Classification, Aspect
Sentiment Extraction.