基于深度学习的文本分类关键问题研究

发布时间：2022-01-25 15:01

　　文本分类由来已久,近年来,随着人工智能和机器学习的迅速发展,文本分类也出现了很多新方法。随着技术的发展,一方面,文本语料的数据质量和数量发生了巨大的变化,大规模语料的积累为更复杂的模型提供了必要的数据保障。另一方面,计算机的计算性能的提升为大规模语料的计算和分析提供了有力的计算资源保障。随着机器学习和深度学习的推进,深度学习的方法在各个领域都表现出强大的优势。本文将在深度学习的基础上探讨文本分类中的基本研究问题。介绍了不同的深度学习方法,如卷积神经网络（Convolutional Neural Network,CNN）和长短期记忆（Long Short-Term Memory,LSTM）。我们提出了分别利用 CNN 和 LSTM,并利用朴素贝叶斯（Naive Bayes,NB）作为对比方法,,以PyCharm是开发平台,在文本情感分类的公开数据集上做了实验,并对实验结果进行了分析。结果表明,所提出的方法比基准方法取得了更好的效果。 

【文章来源】：华北电力大学(北京)北京市 211工程院校 教育部直属院校

【文章页数】：60 页

【学位级别】：硕士

【文章目录】：
摘要
ABSTRACT
CHAPTER 1 INTRODUCTION
    1.1 Text Classification
        1.1.1 Definition
        1.1.2 Basic Concepts of Text Classification
        1.1.3 Text Classification Processes
        1.1.4 Applications of Text Classification
    1.2 Deep Learning
        1.2.1 Definition
        1.2.2 History of Deep Learning
        1.2.3 Applications of Deep Learning in Text Mining
    1.3 Literature Review
    1.4 Research Motivation
    1.5 Thesis Layout
CHAPTER 2 DEEP LEARNING TECHNIQUES
    2.1 Data Preprocessing
        2.1.1 Stemming
        2.1.2 Word Segmentation
    2.2 Text Representation
        2.2.1 Word Embedding
        2.2.2 One-hot Vector
    2.3 Classification
        2.3.1 Convolutional Neural Network (CNN)
        2.3.2 Convolutional Neural Network (CNN)Applications
        2.3.3 Long Short-Term Memory(LSTM)
        2.3.4 Gated Recurrent Unit (GRU)
        2.3.5 Long Short-Term Memory(LSTM)Applications
    2.4 Evaluation
        2.4.1 Confusion Matrix
        2.4.2 Accuracy
        2.4.3 Precision
        2.4.4 Recall
        2.4.5 False Positive rate(FP), True Negative rate (TN)and False Negative rate(FN)
        2.4.6 F-Measure
CHAPTER 3 CONVOLUTIONAL NEURAL NETWORK(CNN)-BASED TEXT CLASSIFICATION
    3.1 Dataset
    3.2 Baseline Method: Naive Bayes (NB)
        3.2.1 Definition
        3.2.2 Environment
        3.2.3 Experiment
        3.2.4 Results and Analysis
    3.3 Proposed Method 1: Convolutional Neural Network (CNN)
        3.3.1 Definition
        3.3.2 Environment
        3.3.3 Model
        3.3.4 Experiment
        3.3.5 Results and Analysis
CHAPTER 4 LONG SHORT-TERM MEMORY(LSTM)-BASED TEXT CLASSIFICATION
    4.1 Definition
    4.2 Environment
    4.3 Model
    4.4 Experiment
    4.5 Results and Analysis
CHAPTER 5 CONCLUSIONS AND FUTURE WORKS
    5.1 Conclusions
    5.2 Future Works
REFERENCES
APPENDIX A-Some Codes From The Baseline Method: Naive Bayes (NB)
APPENDIX B-Codes From The Convolutional Neural Network (CNN)Model
APPENDIX C-Codes From The Long Short-Term Memory (LSTM) Model
ACKNOWLEGEMENT



本文编号：3608745