Related content in the field of sentiment analysis, informal review, for reference only. Some links may fail in the wechat article, please click to read the original article.

(blog.csdn.net/DoJintian/a…

Introduction to the

Sentiment analysis or opinion mining is a computational study of people’s opinions, feelings, emotions, evaluations and attitudes towards products, services, organizations, individuals, issues, events, topics and their attributes.

Existing research has produced a large number of techniques, both supervised and unsupervised, that can be used for emotion analysis of multiple tasks. In supervised methods, early papers used all supervised machine learning methods (such as support vector machines, maximum entropy, naive Bayes, etc.) and feature combinations. Unsupervised methods include different methods using emotional dictionaries, grammatical analysis, and syntactic patterns. There are several review books and papers covering a wide range of early methods and applications.

About a decade ago, deep learning emerged as a powerful machine learning technology, producing current best results in many applications, including computer vision, speech recognition, NLP, and more. The application of deep learning to emotion analysis is also becoming popular.

Three granularities of sentiment analysis

  • Document level: Document level sentiment classification refers to marking the overall emotional orientation/polarity of opinion documents, that is, determining whether the document as a whole conveys positive or negative views. Thus, this is a binary categorization task that can also be formalized as a regression task, such as rating documents on a scale of 1 to 5 stars. Some researchers also see it as a five-category task.
  • Sentence level: Sentence level emotion classification is used to calibrate the expressed emotion in a single sentence. As discussed earlier, the emotion of a sentence can be inferred using subjective classification, which classifies sentences as subjective or objective, and polar classification, which determines that subjective sentences represent negative or positive emotions. In the existing deep learning models, sentence emotion classification usually forms a joint three-category classification problem, that is, the sentence is predicted to be positive, neutral or negative.
  • Aspect level: Also known as topic granularity, each phrase represents a topic. Unlike document-level and statement-level sentiment classifications, aspect level sentiment classifications consider both sentiment information and topic information (emotions generally have a topic). Given a sentence and topic feature, aspect level sentiment classification can infer the emotional polarity/tendency of the sentence at the topic feature. For example, in the sentence “The screen is very clear but the battery life is too short.” If the theme feature is “screen”, the emotion is positive; if the theme feature is “battery life”, The emotion is negative.

Deep learning model

  1. Document/sentence granularity: CNN text classification proposed by Kim et al. (2013) has become one of the important baseline of sentence-level emotion classification tasks; \

  2. Document/sentence granularity: The basic LSTM model and the pooling strategy constitute the classification model, which is usually used for sentiment analysis.

  3. Phrase granularity: Tang et al. (2015) used two different RNN networks to combine text and topic for sentiment analysis; \

  4. Phrase granularity: Tang et al. (2016) combined memory-network to solve the target-dedependent problem, where target is understood as the aspect mentioned above; \

  5. Phrase granularity: Chen et al. (2017) used the compound mechanism of location-weighted memory and stacked attention respectively to model the interactive relationship between target words and texts to solve the phrase-level emotion classification problem. \

  6. Phrasal granularity: Schmitt1 et al. (2018) combined aspect and polarity in training for categorizing tasks, resulting in models of emotion analysis;

  7. Now popular models: large-scale corpus pretraining (word vector /Elmo/GPT/Bert) + deep learning classifier (LSTM/CNN/Transformer). See AI Challenger 2018: Finer-grained User Comment Sentiment Classification Champion Summary for a good example.

The relevant data

Emotional dictionary

  • Part-of-speech dictionary 1 part-of-speech dictionary 2
  • Ontology database of Chinese Emotion Words of Dalian University of Technology
  • Li Jun Chinese commendatory and derogatory thesaurus of Tsinghua University
  • The Emotional Dictionary of CnKI

Affective data set

  • 15 Free Sentiment Analysis Datasets for Machine Learning
  • Dianping fine-grained user comment sentiment data set
  • Automobile industry user viewpoint theme and emotion recognition
  • E-commerce comment on emotional data
  • Hotel review corpus
  • Semeval-2014 Task 4 dataset
  • Citysearch Corpus restaurant review data
  • NLPCC2014 Assessment Task 2_ Deep Learning based emotion Classification
  • NLPCC2013 Assessment Task _ Opinion Factor Extraction of Chinese microblog
  • NLPCC2013 Assessment Task _ Emotion Recognition on Chinese microblog
  • NLPCC2013 Assessment Task _ Cross-domain affective classification
  • NLPCC2012 Evaluation Task _ Sentiment Analysis for Chinese Microblog
  • Cornell University Film Review data set

Other resources

  • [Sentiment Analysis with LSTMs in Tensorflow](github.com/adeshpande3…)
  • Sentiment analysis on tweets using Naive Bayes, SVM, CNN, LSTM, etc.
  • Chinese Shopping Reviews sentiment analysis
  • AI Challenger 2018: Fine-grained User Review Sentiment Category champion summary of ideas

The literature

  • Review of Text Sentiment Analysis (Tencent Semantic Team)
  • Deep learning for sentiment analysis: A survey
  • Sentiment analysis resources
  • Tang D, Qin B, Aspect level sentiment Classification based on Deep memory Network [J]. ArXiv Preprint arXiv: 165.08900, 2016
  • Convolutional Neural Networks for Sentence Classification [J]. ArXiv PrePrint arXiv:1408.5882, 2014.
  • Blog.csdn.net/DoJintian/a…

Please follow and share ↓↓↓\

Machine learning beginners \

QQ group: 727137612

(Note: there are 7 QQ groups on this site, those who have joined any of them do not need to add more)

ID of knowledge planet of this site: 92416895

Past wonderful review \

  • Conscience Recommendation: Introduction to machine learning and learning recommendations (2018 edition) \

  • Github Image download by Dr. Hoi Kwong (Machine learning and Deep Learning resources)

  • Printable version of Machine learning and Deep learning course notes \

  • Machine Learning Cheat Sheet – understand Machine Learning like reciting TOEFL Vocabulary

  • Introduction to Deep Learning – Python Deep Learning, annotated version of the original code in Chinese and ebook

  • Zotero paper Management tool

  • The mathematical foundations of machine learning

  • Machine learning essential treasure book – “statistical learning methods” python code implementation, ebook and courseware

  • Blood vomiting recommended collection of dissertation typesetting tutorial (complete version)

  • The encyclopaedia of Machine learning introduction – A collection of articles from the “Beginner machine Learning” public account in 2018