基于IMDb影评的情感分析研究

Sentiment Analysis Research Based on IMDb Movie Reviews

摘要: 情感分析的任务是挖掘文本中蕴含的情感倾向，对文本情感极性进行分类，是自然语言处理领域研究的重要问题。本研究通过构建一个情感分析系统，选取IMDb影评数据集作为训练与测试对象，探索如何从数据收集、文本预处理到模型训练的完整流程。同时，针对多语言情感分析的挑战，研究了翻译对情感分类效果的影响。实验表明，基于传统机器学习模型(如朴素贝叶斯和支持向量机)的情感分析方法在高质量预处理和特征提取下能够获得较好的性能，而翻译对分类结果的影响取决于语言的特性和翻译工具的准确性。

Abstract: The task of sentiment analysis is to mine the emotional tendencies contained in texts and classify the polarity of text emotions, which is an important issue in the field of Natural Language Processing (NLP). This study explores the complete process from data collection and text preprocessing to model training by constructing a sentiment analysis system and selecting the IMDb movie review dataset as the training and testing subject. Simultaneously, addressing the challenges of multilingual sentiment analysis, the study investigates the impact of translation on sentiment classification effectiveness. Experiments show that sentiment analysis methods based on traditional machine learning models (such as Naïve Bayes and Support Vector Machines) can achieve good performance under high-quality preprocessing and feature extraction. The impact of translation on classification results depends on language characteristics and the accuracy of translation tools.