| 摘 要: 传统机器学习的自然语言处理系统特别依赖人工手动标记的特征，极其耗时且容易出现维度爆炸等难以解决的问题。本文采用基于卷积神经网络(CNN)的深度学习技术来解决这一问题。通过收集校园热点话题进行预处理以及运用Word2vec模型生成词向量后，运用卷积神经网络提取其中的特征并进行情感倾向分类。通过实验数据的比较，基于卷积神经网络(CNN)的情感倾向分类获得了89.76%的准确率，较传统的支持向量机(SVM)提高了7.3%，获得更好的分类性能。本文的研究对高校治理能力和治理体系现代化建设具有积极作用。 |
| 关键词: 自然语言处理；卷积神经网络；情感倾向分析；舆情分析 |
|
中图分类号: TP39
文献标识码: A
|
| 基金项目: 2019年江苏省高校哲学社会科学研究项目(2019SJA2058)；2019年南京理工大学紫金学院校级科研项目(2019ZRKX0401007)；2020年全国高等院校计算机基础教育研究会一般专题类项目(2020-AFCEC-278)；2020年南京理工大学紫金学院教育教学改革与研究重点资助项目(20200103001). |
|
| Application of Emotion Classification Technology based on Deep Learning in University Public Opinion Analysis |
|
HUANG Ping, ZHU Huijuan, CHEN Linlin
|
(Zijin College, Nanjing University of Science and Technology, Nanjing 210000, China)
huangping984@njust.edu.cn; elainezhj@qq.com; chenlinlin606@njust.edu.cn
|
| Abstract: Traditional natural language processing systems for machine learning rely heavily on manually marked features, which are extremely time-consuming and prone to difficult problems like dimensional explosions. This paper proposes to use CNN-based (Convolutional Neural Network) deep learning technology to solve this problem. After hot topics on campus are collected for preprocessing and generating word vectors using word2vec model, CNN is used to extract features and classify emotional tendencies. Through experimental comparison, the emotion tendency classification based on CNN has an accuracy of 89.76%, which is 7.3% higher than that of traditional Support Vector Machine (SVM) and has better classification performance. This research plays a positive role in the modernization of university governance ability and governance system. |
| Keywords: natural language processing; convolutional neural network; emotion tendency analysis; public opinion analysis |
