Cross‑Platform Data‑Driven Analysis of Hot Search Topics and Sentiment Using Multi‑Model Voting

Read full text Summarise with VIRAT
Authors: Weichen Wang
Cited by: 0
References: 8

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

This paper investigates hot search texts from Chinese internet platforms to examine the application and methodology of data-driven text research in the new-media context.The research collects over 150,000 hot search entries from three mainstream social platforms, including Weibo, Zhihu, and Toutiao, and constructs a multidimensional analysis framework through data cleaning, classification, processing, and modeling.Specifically, a hot search classification system is developed using a multi-model voting procedure that involves Gemini, Doubao, Kimi, and ChatGPT, resulting in ten final category labels.The classification accuracy of each model is evaluated against manual annotations, from which Gemini emerges as the best-performing classifier.Following classification by Gemini, the study conducts a systematic analysis that combines high-frequency token statistics, SnowNLP-based sentiment analysis, and Ordinary Least Squares regression.The paper provides a methodological reference for datadriven analysis of popular social media texts and emphasizes the integration of multi-model voting and statistical testing to strengthen the reliability and policy relevance of research on hot search topics and sentiment.

Fields of Study & Subjects

Sentiment Analysis and Opinion Mining Computational and Text Analysis Methods Text and Document Classification Technologies

Publication details

Journal / source
Advances in Social Science, Education and Humanities Research/Advances in social science, education and humanities research
Pages
822–833
Publication date
25 March 2026
Language
EN