Classification of E-commerce Reviews Sentiment Tendency Based on ALBERT-SVM

Konferenz: EEI 2022 - 4th International Conference on Electronic Engineering and Informatics
24.06.2022 - 26.06.2022 in Guiyang, China

Tagungsband: EEI 2022

Seiten: 4Sprache: EnglischTyp: PDF

Autoren:
Cheng, Wenli; Wang, Yuying (College of Cyber Security, Changchun University, Changchun, China)
Li, Lina; Wang, Shaoqiang (College of Computer Science and Technology, Changchun University, Changchun, China)

Inhalt:
On the e-commerce platform, customer comments are very important for businesses to improve their competitiveness and service level. This paper proposes an sentiment analysis model ALBERT-SVM for E-commerce review text, which combines the ALBERT pre-trained model and Support Vector Machine (SVM). In this model, the ALBERT pre-trained model is adopted to produce the dynamic feature representation of the e-commerce review text, so that the same word has different word vector representations in different semantic environments, and then the SVM classifier is used to classify the sentiment tendency of e-commerce reviews. Experiments are conducted on the dataset of online_shopping_10_cats. Compared with SVM and ALBERT models, F1 value of this model is increased by 7.1 and 5.2 percentage points, respectively. The experiment results show that our model has better performance than existing models.