文本情感分类是利用计算机技术对含有情感特征的数据进行处理，它能够识别和提取文本中的主观信息。由于基于传统方法的文本情感分类对特征信息提取的不够充分，而基于深度学习的文本情感分类方法可以改善这一问题，因此，深度学习成为了文本情感分类的主要研究方法。论文为了提取更多的特征信息，提高分类准确率，提出了非静态的FastText词向量模型和Attention-CNN的融合模型。论文的主要研究内容如下：

（1）提出了一种基于非静态的FastText词向量的方法。该方法通过FastText模型进行词向量训练得到初始的词向量矩阵，并在训练过程中作为参数训练，通过不断进行调整和更新，使得词向量更加适用于该文本情感分类任务，获得更多词语之间的语义和语法关系，从而提高分类准确率。在卷积神经网络进行文本情感分类实验，与静态的FastText、Word2Vec和Doc2vec模型进行相比，非静态的FastText模型在公开的评论数据集上，准确率提高了2.54%、6.22%和4.82%；在酒店评论语料库上，准确率提高了1.84%、6.22%和5.10%。

（2）提出了一种基于Attention-CNN的融合模型，该模型结合了卷积神经网络、门控神经网络和注意力机制，针对卷积神经网络无法捕获长距离依赖关系，提取的特征不够充分的问题，首先通过实验确定了相关超参数的取值，包括 Dropout值、batch-size参数和激活函数等；然后进行了模型对比实验，与CNN、LSTM和GRU相比，基于Attention-CNN的融合模型在公开的评论数据集上，准确率提高了4.70% 、3.66%和2.95%，在酒店评论语料库上，准确率提高了3.61%、2.88%和2.03%。

（3）构建了文本内容情感分类系统。该系统验证了基于非静态FastText词向量和基于Attention-CNN的融合模型的有效性，用Python对该系统进行实现，通过可视化操作可以清晰地展示分类结果的具体细节，在生活中有一定使用价值。

Text sentiment classification is the use lor='red'>of computer technology to process data containing emotional features, which can identify and extract subjective information in text. Because the text sentiment classification based on traditional methods is not sufficient for feature information extraction, and the text sentiment classification method based on deep learning can improve this problem, deep learning has become the main research method lor='red'>of text sentiment classification. In order to extract more feature information and improve classification accuracy, the paper proposes a non-static FastText word vector model and Attention-CNN fusion model. The main research contents lor='red'>of the thesis are as follows:

(1) A method based on non-static FastText word vectors is proposed. This method uses the FastText model to train the word vector to obtain the initial word vector matrix, which is used as parameter training during the training process. Through continuous adjustment and update, the word vector is more suitable for the text emotion classification task, and more words are obtained. The semantic and grammatical relationship lor='red'>of the, so as to improve the classification accuracy. Perform text sentiment classification experiments on convolutional neural networks. Compared with the static FastText, Word2Vec and Doc2vec models, the non-static FastText model has increased accuracy by 2.54%, 6.22% and 4.82% on the public comment data set; On the hotel review corpus, the accuracy rate increased by 1.84%, 6.22% and 5.10%.

(2) A fusion model based on Attention-CNN is proposed. The model combines convolutional neural network, gated neural network and attention mechanism. Aiming at the inability lor='red'>of convolutional neural network to capture long-distance dependencies, the extracted features are not sufficient For the problem, firstly determine the value lor='red'>of related hyperparameters through experiments, including Dropout value, batch-size parameter and activation function, etc; then conduct model comparison experiments, compared with CNN, LSTM and GRU, based on Attention-CNN fusion On the public review data set, the accuracy lor='red'>of the model increased by 4.70%, 3.66%, and 2.95%, and on the hotel review corpus, the accuracy increased by 3.61%, 2.88%, and 2.03%.

(3) Constructed a text content sentiment classification system. The system verifies the effectiveness lor='red'>of the non-static FastText word vector and the Attention-CNN-based fusion model. The system is implemented in Python, through visual operation, the specific details lor='red'>of the classification results can be clearly displayed, which has certain use value in life.

[1] 黄琼霞.基于深度学习的文本情感分析研究[D].福建农林大学,2018.

[2] Neethu M S, Rajasree R. Sentiment analysis in twitter using machine learning techniques[C]//Fourth International Conference on Computing, Communications and NETWORKING Technologies. IEEE, 2014:1-5.

[3] Krizhevsky A , Sutskever I , Hinton G . ImageNet Classification with Deep Convolutional Neural Networks[J]. Advances in neural information processing systems, 2012, 25(2).

[4] Graves A, Mohamed A R, Hinton G. Speech recognition with deep recurrent neural networks[J].2013, 38(2003):6645-6649.

[5] Yu H, Hatzivassiloglou V. Towards Answering Opinion Questions: Separating Facts from Opinions and Identifying the Polarity of Opinion Sentences[C]// Proceedings of the 2003 Conference on Empirical Methods in Natural Language Processing. Association for Compu-tational Linguistics, 2003: 129- 136.

[6] Kobayashi N, Inui K, Matsumoto Y, et al. Collecting Evaluative Expressions for Opinion

Extraction[J]. Journal of Natural Language Processing, 2004, 12: 596-605.

[7] T'Sou K Y , K wong O Y , Wong W L , et al. Sentiment and Content Analysis of Chinese News Coverage[J]. International Journal of Computer Processing Of Languages, 2005, 18(2):171-183.

[8] Turney P D, Littman M L. Measuring praise and criticism: Inference of semantic orientation from association [J]. ACM Transactions on Information Systems, 2003, 21(4):315-346.

[9] Fiaidhi, J,Mohammed, O.,Mohammed, S.,Fong, S,Tai hoon Kim. Opinion mining over twitterspace: Classifying tweets programmatically using the R approach[P]. Digital Information Management (ICDIM), 2012 Seventh International Conference on,2012.

[10] Kamps J, Marx M, Mokken R J. Using WordNet to measure semantic orientation of adjectives[C]// Proceedings of the Conference on Language Resources and Evaluation, 2004:115-1118.

[11] 徐琳宏,林鸿飞,赵晶.情感语料库的构建和分析[J].中文信息学报,2008(01):116-122.

[12] 柳位平,朱艳辉,栗春亮,向华政,文志强.中文基础情感词词典构建方法研究[J].计算机应用,2009,29(10):2875-2877.

[13] 杨超,冯时,王大玲,杨楠,于戈.基于情感词典扩展技术的网络舆情倾向性分析[J].小型微型计算机系统,2010,31(04):691-695.

[14] 孙本旺,田芳,藏文.情感词典的构建及微博情感计算研究[J]. 计算机技术与发展，2018,28( 11) : 218-222．

[15] 王志涛,於志文,郭斌.基于词典和规则集的中文微博情感分析[J]. 计算机工程与应用,2015,51( 8) : 218-225．

[16] Pang T B, Pang B, Lee L. Thumbs up Sentiment Classification using Machine Learning[J]. Proceedings of Emnlp, 2002:79-86.

[17] Kim S M, Hovy E. Automatic identification of pro and con reasons in online reviews[C]// Coling/acl on Main Conference Poster Sessions. Association for Computational Linguistics, 2006:483-490.

[18] Wang S, Manning C D. Baselines and bigrams: simple, good sentiment and topic classification[C]//Meeting of the Association for Computational Linguistics: Short Papers. Association for Computational Linguistics, 2012:90-94.

[19] Agarwal B, Mittal N. Machine Learning Approach for Sentiment Analysis [M]//Prominent Feature Extraction for Sentiment Analysis. Springer International Publishing, 2016.

[20] Liu S，Li F，Li F，et al． Adaptive co-training SVM for sentiment classification on tweets[C]//Proceedings of the 22nd ACM international conference on Conference on information ＆ knowledge management. ACM,2013．

[21] 张莹. 在线新闻评论的情感分析研究[D]. 南开大学, 2013.

[22] 杨经,林世平.基于SVM的文本词句情感分析[J].计算机应用与软件,2011, 28(09): 225-228.

[23] 李婷婷,姬东鸿.基于SVM和CRF多特征组合的微博情感分析[J].计算机应用研究, 2015, 32(4):978-981.

[24] 徐军,丁宇新,王晓龙.使用机器学习方法进行新闻的情感自动分类[J].中文信息学报,2007, 21(6): 95-100.

[25] Hinton G E, Osindero S, Teh Y W. A Fast Learning Algorithm for Deep Belief Nets [J]. Neural Computation, 2014, 18(7):1527-1554.

[26] Kim Y . Convolutional Neural Networks for Sentence Classification[J]. Eprint Arxiv, 2014.

[27] Ye Zhang, Byron C. Wallace A. Sensitivity analysis of (and Practitioners’ Guide to) convolutional neural networks for sentence classification[J]. Computer Science, 2015:253-263.

[28] Jordan, M.I. A parallel distributed processing approach [J]. Advances in psychology, 1997,121: 471-495.

[29] Hochreiter S，Schmidhuber Jürgen． Long short-term memory［J］．Neural Computation，1997，9( 8) : 1735-1780.

[30] Miyato T, Dai A. Good fellow I J , et al.Adversarial training methods for semi-supervised text classification[C]//Proceedings of the International Conference on Learning Representations, 2017.

[31] Miyato T, MaedaS,I shiiS, et al. Virtual adversarial training: are gularization method for supervised and semi-supervised learning[J].IEEE Transactions on Pattern Analysis and Machine Intelligence,2019,41(8):1979-1993.

[32] Graves A, Schmidhuber J. Framewise phoneme classification with bidirectional LSTM and other neural network architectures[J]. Neural Networks,2015,18(5):602-610.

[33] Kyunghyun Cho, Bart van Merrienboer, DzmitryBahdanau, YoshuaBengio. On the Properties of Neural Machine Translation: Encoder-Decoder Approaches[J].Computer Science,2014, 10.3115/v1/W14-4012.

[34] Chen X, Qiu X, Zhu C, et al. Gated recursive neural network for Chinese word segmentation[C]//Proceedings of the ACL,2015.

[35] Tang Qinting, li Jian, Chen Jiayu, Lu Hengtong, Du Yu, Yang Kehan. Full Attention-Based Bi-GRU Neural Network for News Text Classification[C]//2019 IEEE 5th International Conference on Computer and Communication, 2019.

[36] Chen Y, Ouyang Y, Li W, et al. Using deep belief nets for Chinese named entitycategorization[C]. Named Entities Workshop, 2010: 102-109.

[37] 刘龙飞,杨亮,张绍武,等.基于卷积神经网络的微博情感倾向性分析[J].中文信息学报, 2015, 29(6):159-165.

[38] 冯兴杰,张志伟,史金钏.基于卷积神经网络和注意力模型的文本情感分析[J].计算机应用研究, 2018(5).

[39] 何炎祥,孙松涛,牛菲菲,等.用于微博情感分析的一种情感语义增强的深度学习模型[J].计算机学报, 2017, 40(4):773-790.

[40] 李科.基于多元特征融合和LSTM神经网络的中文评论情感分析[D].太原理工大学,2017.

[41] 吴小华,陈莉,魏甜甜,范婷婷.基于Self-Attention和Bi-LSTM的中文短文本情感分析[J].中文信息学报,2019,33(6): 100-107.

[42] 杨东,王移芝.基于 Attention-based C-GRU神经网络的文本分类[J]计算机与现代化 2018 (2):96-100.

[43] 孙明敏.基于GRU-Attention的中文文本分类[J]现代信息科技2019,3(3):10-12.

[44] 线岩团,相艳,余正涛,文永华,王红斌,张亚飞.用于文本分类的均值原型网络[J]中文信息学报. 2020,34(6):74-80.

[45] Mikolov T，Karafiat M，Burget L，et al．Recurrent eural network based language model［C］//Conference of the International Speech Communication Association，2010: 1045 -1048.

[46] Quoc Le, Tomas Mikolov.Distributed Representations of Sentences and Documents[C] // Proceedings of the 31st International Conference on Machine Learning, Beijing, China，2014.

[47] 汪岿,刘柏嵩.文本分类研究综述[J].数据通信,2019(03):37-47.

[48] Jordan, M.I. A parallel distributed processing approach [J]. Advances in psychology, 1997,121: 471-495.

[49] Schmidhuber J. Deep learning in neural networks: An overview[J]. Neural networks, 2015,61: 85-117.

[50] Hochreiter S, Schmidhuber J. Long Short-Term Memory[J]. Neural Computation, 1997, 9(8):1735-1780..

[51] Tao L , Yu Z . Training RNNs as Fast as CNNs[J]. 2017.

[52] Kokkinos ,Potamianos A.Structural Attention Neural Networks for improved sentiment analysis[J]. 2017:586-591.

[53] Vaswani A, Shazeer N, Parmar N, et al. Attention is all you need[C]//Advances in neural information processing systems. 2017.

[54] Hinton, GeoffreyE, Srivastava, Nitish, Krizhevsky, Alex, Sutskever, Ilya, Salakutdinov, Ruslan R Improving neural networksby preventing co-adaptation of feature detectors[J] Computer Science 2012.

[55] Duchi J , Hazan E , Singer Y , et al. dearly. Adaptive Subgradient Methods Adaptive

Subgradient Methods for Online Learning and Stochastic Optimization[J]. Journal of Machine Learning Research, 2011, 12(7):257-269.