狭义的情感分析(sentiment analysis)是指利用计算机实现对文本数据的观点、情感、态度、情绪等的分析挖掘。广义的情感分析则包括对图像视频、语音、文本等多模态信息的情感计算。简单地讲,情感分析研究的目标是建立一个有效的分析方法、模型和系统,对输入信息中某个对象分析其持有的情感信息,例如观点倾向、态度、主观观点或喜怒哀乐等情绪表达。

知识荟萃

情感分析 ( Sentiment Analysis ) 专知荟萃

入门学习

  1. 斯坦福大学自然语言处理第七课“情感分析(Sentiment Analysis)” [http://52opencourse.com/235/%E6%96%AF%E5%9D%A6%E7%A6%8F%E5%A4%A7%E5%AD%A6%E8%87%AA%E7%84%B6%E8%AF%AD%E8%A8%80%E5%A4%84%E7%90%86%E7%AC%AC%E4%B8%83%E8%AF%BE-%E6%83%85%E6%84%9F%E5%88%86%E6%9E%90%EF%BC%88sentiment-analysis%EF%BC%89] [https://class.coursera.org/nlp/]
  2. 情感分类方法简介 [http://www.jianshu.com/p/61212b11769a]
  3. NLP 笔记 - Sentiment Analysis [http://www.shuang0420.com/2017/06/01/NLP%20%E7%AC%94%E8%AE%B0%20-%20Sentiment%20Analysis/]
  4. 斯坦福CoreNLP —— 用Java给Twitter进行情感分析 [http://zqdevres.qiniucdn.com/data/20131225114906/index.html]
  5. TEXT CLASSIFICATION FOR SENTIMENT ANALYSIS – NLTK + SCIKIT-LEARN [https://streamhacker.com/2012/11/22/text-classification-sentiment-analysis-nltk-scikitlearn/]
  6. Sentiment Analysis in Python [http://andybromberg.com/sentiment-analysis-python/]
  7. Basic Sentiment Analysis with Python [http://fjavieralba.com/basic-sentiment-analysis-with-python.html]
  8. 中文情感分析 (Sentiment Analysis) 的难点在哪?现在做得比较好的有哪几家? [https://www.zhihu.com/question/20700012]

综述

  1. Sentiment analysis and opinion mining https://www.cs.uic.edu/~liub/FBS/SentimentAnalysis-and-OpinionMining.pdf~
  2. Sentiment analysis algorithms and applications: A survey [https://pan.baidu.com/s/1miR4DD2] http://www.sciencedirect.com/science/article/pii/S2090447914000550
  3. Sentiment Analysis:A Comparative Study On Different Approaches https://www.researchgate.net/profile/Amal_Ganesh/publication/303848210_Sentiment_Analysis_A_Comparative_Study_on_Different_Approaches/links/576a633208aeb526b69b84d7/Sentiment-Analysis-A-Comparative-Study-on-Different-Approaches.pdf
  4. 文本情感分析 [http://jos.org.cn/ch/reader/create_pdf.aspx?file_no=3832&journal_id=jos]
  5. Opinion Mining and Sentiment Analysis Bo Pang1 and Lillian Lee2 [https://www.cse.iitb.ac.in/~pb/cs626-449-2009/prev-years-other-things-nlp/sentiment-analysis-opinion-mining-pang-lee-omsa-published.pdf]

进阶论文

2002

  1. Bo Pang, Lillian Lee, Shivakumar Vaithyanathan. Thumbs up? Sentiment Classification using Machine Learning Techniques. EMNLP, 2002.
    [https://wenku.baidu.com/view/efa9391d650e52ea551898e8.html]

2004

  1. Minqing Hu and Bing Liu. Mining and summarizing customer reviews. KDD: 168-177, 2004.
    [https://dl.acm.org/citation.cfm?id=1014073]

2011

  1. Maite Taboada, Julian Brooke, Milan Tofiloski, Kimberly Voll, and Manfred Stede. Lexicon-Based Methods for Sentiment Analysis. Computational Linguistics: 37(2), 267-307. 2011.
    [https://dl.acm.org/citation.cfm?id=1014073]
  2. Dmitriy Bespalov, Bing Bai, Yanjun Qi, Ali Shokoufandeh. Sentiment Classification Based on Supervised Latent n-gram Analysis. Proceedings of the Conference on Information and Knowledge Management, 2011.
    [https://dl.acm.org/citation.cfm?id=2063576.2063635]

2012

  1. Bing Liu. 2012. Sentiment analysis and opinion mining. In Synthesis lectures on human language technologies, 1-167.
    [http://download.csdn.net/download/kevin_done_register/6750185]

2014

  1. Simpler is better? Lexicon-based ensemble sentiment classification beats supervised methods.
    [https://www.cs.rpi.edu/szymansk/papers/C3-ASONAM14.pdf]
  2. Duyu Tang, Furu Wei, Bing Qin, Ting Liu, Ming Zhou. 2014. Building Large-Scale Twitter-Specific Sentiment Lexicon : A Representation Learning Approach. International Conference on Computational Linguistics(COLING).
    [http://www.aclweb.org/anthology/C14-1018]

2015

  1. Sentiment Analysis: mining sentiments, opinions, and emotions 图书
    [https://www.cs.uic.edu/~liub/FBS/sentiment-opinion-emotion-analysis.html]
  2. Rie Johnson and Tong Zhang. Effective use of word order for text categorization with convolutional neural networks. In NAACL 2015.
    [https://arxiv.org/abs/1412.1058]
  3. Rie Johnson, and Tong Zhang. Semi-supervised convolutional neural networks for text categorization via region embedding. In NIPS 2015.
    [http://pubmedcentralcanada.ca/pmcc/articles/PMC4831869/]
  4. Xiang Zhang, Junbo Zhao, and Yann LeCun. Character-level convolutional networks for text classification. In NIPS 2015.
    [http://arxiv.org/abs/1509.01626]

2016

  1. Comprehensive Study on Lexicon-based Ensemble Classification Sentiment Analysis.
    [http://www.mdpi.com/1099-4300/18/1/4]
  2. Duyu Tang, Furu Wei, Bing Qin, Nan Yang, Ting Liu, Ming Zhou. 2016. Sentiment Embeddings with Applications to Sentiment Analysis. IEEE Transactions on Knowledge and Data Engineering (TKDE).
    [https://www.mendeley.com/research-papers/sentiment-embeddings-applications-sentiment-analysis/]
  3. Yafeng Ren, Yue Zhang, Meishan Zhang, and Donghong Ji. 2016. Improving Twitter Sentiment Classification Using Topic-Enriched Multi-Prototype Word Embeddings. In Proceedings of AAAI.
    [https://aaai.org/ocs/index.php/AAAI/AAAI16/paper/view/11925]
  4. Zichao Yang, Diyi Yang, Chris Dyer, Xiaodong He, Alex Smola, Eduard Hovy. 2016. Hierarchical Attention Networks for Document Classification. In NAACL 2016.
    [https://www.microsoft.com/en-us/research/publication/hierarchical-attention-networks-document-classification/]
  5. Rie Johnson, and Tong Zhang. Supervised and Semi-Supervised Text Categorization using LSTM for Region Embeddings. In ICML 2016.
    [https://arxiv.org/abs/1602.02373]
  6. Alexis Conneau, Holger Schwenk, Loïc Barrault, and Yann Lecun. 2016. Very Deep Convolutional Networks for Natural Language Processing. arXiv.org 1606.01781.
    [https://arxiv.org/abs/1606.01781v1]
  7. Huimin Chen, Maosong Sun, Cunchao Tu, Yankai Lin, Zhiyuan Liu. Neural Sentiment Classification with User and Product Attention. In EMNLP 2016.
    [http://nlp.csai.tsinghua.edu.cn/~lzy/publications/emnlp2016.pdf]
  8. Lin Gui, Dongyin Wu, Ruifeng Xu*, Qin Lu, Yu Zhou. Event-Driven Emotion Cause Extraction with Corpus Construction. In EMNLP 2016.
    [http://pdfs.semanticscholar.org/120b/d71c72f9477dec6b5291c32f73ae4afbf163.pdf]

Tutorial

  1. 面向社会媒体的文本情感分析 秦兵 哈尔滨工业大学 2017.9.16 北京 第六届全国社会媒体处理大会 [https://pan.baidu.com/s/1i5qxd1V]
  2. 文本情绪分类关键技术研究 李寿山 苏州大学,自然语言处理实验室 2017.9.16 北京 第六届全国社会媒体处理大会 [https://pan.baidu.com/s/1pLLsV3d]
  3. Affective Computing on Social Media Data 贾珈 - 清华大学 2017.9.16 北京 第六届全国社会媒体处理大会 [https://pan.baidu.com/s/1mhDHrxY]
  4. Sentiment Analysis with Neural Network 唐都钰、张梅山  深度学习与情感分析 2016 [https://pan.baidu.com/s/1c2NHlNM] [https://pan.baidu.com/s/1c2ETG0S]
  5. A Short Overview On Sentiment Analysis 黄民烈 清华大学 2016 [https://pan.baidu.com/s/1o7XVV0u]
  6. LingPipe Sentiment 一个java自然语言处理包 [http://alias-i.com/lingpipe/demos/tutorial/sentiment/read-me.html]

代码

  1. Sentiment TreeBank 斯坦福结构依存情感分析演示 [http://nlp.stanford.edu:8080/sentiment/rntnDemo.html]
  2. Sentiment Analysis with Python NLTK Text Classification [http://text-processing.com/demo/sentiment/]
  3. Vivekn's sentiment model [https://github.com/vivekn/sentiment/]
  4. nltk -sentiment analysis tool, Lexical, Dictionary-based, Rule-based. [http://www.nltk.org/]
  5. twitter-sent-dnn Supervised Machine Learning, Deep Learning, Convolutional Neural Network. [https://github.com/xiaohan2012/twitter-sent-dnn]

视频教程

  1. 斯坦福大学自然语言处理第七课-情感分析 [https://class.coursera.org/nlp/]

数据集

  1. Stanford Sentiment Treebank [https://nlp.stanford.edu/sentiment/code.html]
  2. Amazon product dataset  [http://jmcauley.ucsd.edu/data/amazon/]
  3. IMDB movies reviews dataset [http://ai.stanford.edu/~amaas/data/sentiment/]
  4. Sentiment Labelled Sentences Data Set  [https://archive.ics.uci.edu/ml/datasets/Sentiment+Labelled+Sentences]

领域专家

  1. 黄民烈 [http://www.tsinghua.edu.cn/publish/cs/4616/2013/20131122151220708543803/20131122151220708543803_.html]
  2. 李寿山 [http://nlp.suda.edu.cn/~lishoushan/]
  3. Bing Liu [https://www.cs.uic.edu/~liub/]
  4. John Blitzer  [http://john.blitzer.com/]
  5. 万小军 [https://sites.google.com/site/wanxiaojun1979/]
  6. 唐都钰 哈尔滨工业大学 [https://www.microsoft.com/en-us/research/people/dutang/]

初步版本,水平有限,有错误或者不完善的地方,欢迎大家提建议和补充,会一直保持更新,本文为专知内容组原创内容,未经允许不得转载,如需转载请发送邮件至fangquanyi@gmail.com 或 联系微信专知小助手(Rancho_Fang)

敬请关注http://www.zhuanzhi.ai 和关注专知公众号,获取第一手AI相关知识

VIP内容

摘要

图像可以传达丰富的语义,促使读者产生多种情感。近年来,随着情感智力的快速发展和视觉数据的爆炸式增长,情感图像内容分析(AICA)得到了广泛的研究。在此,我们将全面回顾近二十年来AICA的发展,特别是针对情感鸿沟、感知主观性、标签噪声和缺失三个主要挑战的最先进的方法。首先,我们介绍了在AICA中广泛应用的关键情感表示模型,以及可用数据集的描述,以进行标签噪声和数据集偏差的定量比较评估。然后,我们总结并比较了情感特征提取的代表性方法 (1)包括手工特征和深度特征,(2) 主导情感识别的学习方法,个性化情感预测,情感分布学习,以及从噪声数据或少量标签中学习的方法,以及(3)基于AICA的应用。最后,讨论了图像内容与语境理解、群体情感聚类、观众-图像交互等方面的挑战和未来的研究方向。

https://www.zhuanzhi.ai/paper/a3ad209b82c286810a4b2a27981757e2

引言

明斯基(1970年图灵奖得主)在《心灵社会》[1]中声称:“问题不在于智能机器是否可以有任何情感,而在于机器是否可以没有情感的智能。”虽然情感在机器和人工智能中扮演着至关重要的角色,但人们对情感计算的关注远远少于对客观语义理解的关注,比如计算机视觉中的对象分类。人工智能的快速发展在语义理解方面取得了显著的成功,并对情感交互提出了更高的要求。例如,能够识别和表达情感的陪伴机器人可以为人类提供更和谐的陪伴,特别是老年人和独生子女。要想拥有类似人类的情感,机器首先应该了解人类如何通过多种渠道表达情感,比如言语、手势、面部表情和生理信号[2]。虽然其他信号可以很容易地被抑制或掩盖,但由交感神经系统控制的生理信号是独立于人类意志的,因此提供了更可靠的信息。然而,捕捉准确的生理信号是相当困难和不切实际的,因为它需要特殊类型的可穿戴传感器。另一方面,移动设备中摄像头的便捷接入和社交网络(如Twitter、Flickr、Weibo)的广泛流行,使得人们习惯使用图片、视频和文本[3]在网上分享他们的经验和表达他们的观点。识别这大量多媒体数据的情感内容提供了一种理解用户行为和情感的替代方法。

我们知道,“一图胜千言”,这说明图像可以传达丰富的语义。不同于现有的分析图像感知方面的研究,如物体检测和语义分割,情感图像内容分析(AICA)侧重于理解更高层次的语义——认知层次,即理解观看者对图像所能诱发的情感。哪个更有挑战性。利用AICA对人的情感状态进行自动推断,有助于评估人的心理健康状况,发现情感异常,防止对自己甚至对整个社会的极端行为。例如,在图1中,发布图片(b)的用户比发布图片(a)的用户更容易产生负面情感。

主要目标。给定一个输入图像,AICA的主要目的是(1)识别特定读者或大多数人的情感(基于心理学,情感可能以不同的模型表示,例如分类或维度。(2)分析图像中包含的刺激激发了这种情感(例如特定的物体或颜色组合),(3)将识别出的情感应用于不同的现实应用中,以提高情感智力能力。

在这项综述中,我们集中回顾了最先进的方法,并概述了研究趋势。首先,我们介绍了1.3节的简要历史,以及与1.4节中其他相关主题的比较。其次,我们在第二节中描述了广泛使用的情感表征模型。第三,我们总结了在第3节中用于进行AICA评估的可用数据集,并定量比较了标签噪声和数据集偏差。第四,基于1.1节的主要目标和挑战,我们总结并比较了情感特征提取、学习方法(主导情感识别、个性化情感预测、情感分布学习、从嘈杂数据或少量标签中学习)等方面的代表性方法。第4、5、6部分分别基于AICA的应用,如图6所示。最后,我们在第7节讨论了可能的研究方向。

成为VIP会员查看完整内容
0
22

最新内容

Contrastive learning techniques have been widely used in the field of computer vision as a means of augmenting datasets. In this paper, we extend the use of these contrastive learning embeddings to sentiment analysis tasks and demonstrate that fine-tuning on these embeddings provides an improvement over fine-tuning on BERT-based embeddings to achieve higher benchmarks on the task of sentiment analysis when evaluated on the DynaSent dataset. We also explore how our fine-tuned models perform on cross-domain benchmark datasets. Additionally, we explore upsampling techniques to achieve a more balanced class distribution to make further improvements on our benchmark tasks.

0
0
下载
预览

最新论文

Contrastive learning techniques have been widely used in the field of computer vision as a means of augmenting datasets. In this paper, we extend the use of these contrastive learning embeddings to sentiment analysis tasks and demonstrate that fine-tuning on these embeddings provides an improvement over fine-tuning on BERT-based embeddings to achieve higher benchmarks on the task of sentiment analysis when evaluated on the DynaSent dataset. We also explore how our fine-tuned models perform on cross-domain benchmark datasets. Additionally, we explore upsampling techniques to achieve a more balanced class distribution to make further improvements on our benchmark tasks.

0
0
下载
预览
Top