Twitter Sentiment Analysis - Indian Institute of ... · Twitter Sentiment Analysis Text length...
Transcript of Twitter Sentiment Analysis - Indian Institute of ... · Twitter Sentiment Analysis Text length...
![Page 1: Twitter Sentiment Analysis - Indian Institute of ... · Twitter Sentiment Analysis Text length Topic relevance Noisy text Data sparsity Negation Stopwords Tokenization Multilingual](https://reader033.fdocuments.in/reader033/viewer/2022050213/5f5fcb08f376a44e181cd32b/html5/thumbnails/1.jpg)
Twitter Sentiment AnalysisTwitter Sentiment Analysis
Presented by:Loitongbam Gyanendro Singh
![Page 2: Twitter Sentiment Analysis - Indian Institute of ... · Twitter Sentiment Analysis Text length Topic relevance Noisy text Data sparsity Negation Stopwords Tokenization Multilingual](https://reader033.fdocuments.in/reader033/viewer/2022050213/5f5fcb08f376a44e181cd32b/html5/thumbnails/2.jpg)
What is “Sentiment analysis”?● Study that aims to identify the orientation of opinions in a text
![Page 3: Twitter Sentiment Analysis - Indian Institute of ... · Twitter Sentiment Analysis Text length Topic relevance Noisy text Data sparsity Negation Stopwords Tokenization Multilingual](https://reader033.fdocuments.in/reader033/viewer/2022050213/5f5fcb08f376a44e181cd32b/html5/thumbnails/3.jpg)
Source ofSentimentInformation
Source of image: http://jameskaskade.com/?p=2336
![Page 4: Twitter Sentiment Analysis - Indian Institute of ... · Twitter Sentiment Analysis Text length Topic relevance Noisy text Data sparsity Negation Stopwords Tokenization Multilingual](https://reader033.fdocuments.in/reader033/viewer/2022050213/5f5fcb08f376a44e181cd32b/html5/thumbnails/4.jpg)
Advent of various social media platforms➔ Given netizen the liberty to openly express their views and opinions➔ Large volume of data to get these information➔ Knowing “what people think”➔ Studies of SA deals:
◆ Product and services reviews,◆ Celebrities,◆ Government policies,◆ Event,◆ and many more…
Why Sentiment Analysis?
OM: Study the subjectivity of opinionSA: Study the sentiment of opinion
![Page 5: Twitter Sentiment Analysis - Indian Institute of ... · Twitter Sentiment Analysis Text length Topic relevance Noisy text Data sparsity Negation Stopwords Tokenization Multilingual](https://reader033.fdocuments.in/reader033/viewer/2022050213/5f5fcb08f376a44e181cd32b/html5/thumbnails/5.jpg)
➔ An opinion is quintuple: (Bing Liu, 2012)◆ (ei,aij,sijkl,hk,tl)
➔ Example:◆ The picture quality of my new Nikon V3 camera is great◆ (Nikon V3, picture quality, positive, User, Time)
➔ Where can we find opinionated text?◆ Blogs◆ Microblogs◆ Consumer forum/sites, etc.
Opinion
![Page 6: Twitter Sentiment Analysis - Indian Institute of ... · Twitter Sentiment Analysis Text length Topic relevance Noisy text Data sparsity Negation Stopwords Tokenization Multilingual](https://reader033.fdocuments.in/reader033/viewer/2022050213/5f5fcb08f376a44e181cd32b/html5/thumbnails/6.jpg)
➔ Microblogs contains a large amount of opinionated text➔ There are many microblogging platforms available
◆ Twitter◆ Tumbler◆ FourSquare◆ Google+◆ LinkedIn
➔ Twitter provides an easy way to access and download published posts
Microblogs
![Page 7: Twitter Sentiment Analysis - Indian Institute of ... · Twitter Sentiment Analysis Text length Topic relevance Noisy text Data sparsity Negation Stopwords Tokenization Multilingual](https://reader033.fdocuments.in/reader033/viewer/2022050213/5f5fcb08f376a44e181cd32b/html5/thumbnails/7.jpg)
➔ Microblogs contains a large amount of opinionated text➔ There are many microblogging platforms available
◆ Twitter◆ Tumbler◆ FourSquare◆ Google+◆ LinkedIn
➔ Twitter provides an easy way to access and download published posts
Microblogs
![Page 8: Twitter Sentiment Analysis - Indian Institute of ... · Twitter Sentiment Analysis Text length Topic relevance Noisy text Data sparsity Negation Stopwords Tokenization Multilingual](https://reader033.fdocuments.in/reader033/viewer/2022050213/5f5fcb08f376a44e181cd32b/html5/thumbnails/8.jpg)
➔ Majority of TSA studies deals on building sentiment classifierTwitter Sentiment Analysis
![Page 9: Twitter Sentiment Analysis - Indian Institute of ... · Twitter Sentiment Analysis Text length Topic relevance Noisy text Data sparsity Negation Stopwords Tokenization Multilingual](https://reader033.fdocuments.in/reader033/viewer/2022050213/5f5fcb08f376a44e181cd32b/html5/thumbnails/9.jpg)
➔ Text length➔ Topic relevance➔ Noisy text➔ Data sparsity➔ Negation➔ Stopwords➔ Tokenization➔ Multilingual content➔ Multimodal content
TSA challenges
![Page 10: Twitter Sentiment Analysis - Indian Institute of ... · Twitter Sentiment Analysis Text length Topic relevance Noisy text Data sparsity Negation Stopwords Tokenization Multilingual](https://reader033.fdocuments.in/reader033/viewer/2022050213/5f5fcb08f376a44e181cd32b/html5/thumbnails/10.jpg)
➔ Semantic➔ Syntactic➔ Stylistic➔ Twitter specific features
Features Opinion words, Sentiment words,Semantic concepts,Negation, etc.
![Page 11: Twitter Sentiment Analysis - Indian Institute of ... · Twitter Sentiment Analysis Text length Topic relevance Noisy text Data sparsity Negation Stopwords Tokenization Multilingual](https://reader033.fdocuments.in/reader033/viewer/2022050213/5f5fcb08f376a44e181cd32b/html5/thumbnails/11.jpg)
➔ Semantic➔ Syntactic➔ Stylistic➔ Twitter specific features
Features Unigrams,Bigrams,N-grams,Terms’ frequencies,POS,Dependency tree, etc
![Page 12: Twitter Sentiment Analysis - Indian Institute of ... · Twitter Sentiment Analysis Text length Topic relevance Noisy text Data sparsity Negation Stopwords Tokenization Multilingual](https://reader033.fdocuments.in/reader033/viewer/2022050213/5f5fcb08f376a44e181cd32b/html5/thumbnails/12.jpg)
➔ Semantic➔ Syntactic➔ Stylistic➔ Twitter specific features
FeaturesEmoticons,Intensifiers,Abbreviations,Slang terms,Punctuation marks, etc.
![Page 13: Twitter Sentiment Analysis - Indian Institute of ... · Twitter Sentiment Analysis Text length Topic relevance Noisy text Data sparsity Negation Stopwords Tokenization Multilingual](https://reader033.fdocuments.in/reader033/viewer/2022050213/5f5fcb08f376a44e181cd32b/html5/thumbnails/13.jpg)
➔ Tweet➔ User➔ Mention➔ Replies➔ Follower➔ Retweet➔ Hashtag➔ Privacy
Twitter Specific Features
![Page 14: Twitter Sentiment Analysis - Indian Institute of ... · Twitter Sentiment Analysis Text length Topic relevance Noisy text Data sparsity Negation Stopwords Tokenization Multilingual](https://reader033.fdocuments.in/reader033/viewer/2022050213/5f5fcb08f376a44e181cd32b/html5/thumbnails/14.jpg)
➔ Manual selection
➔ Statistical analysis
➔ Dimensionality reduction
➔ Representation learning
Features Selection
![Page 15: Twitter Sentiment Analysis - Indian Institute of ... · Twitter Sentiment Analysis Text length Topic relevance Noisy text Data sparsity Negation Stopwords Tokenization Multilingual](https://reader033.fdocuments.in/reader033/viewer/2022050213/5f5fcb08f376a44e181cd32b/html5/thumbnails/15.jpg)
Statistical Approach● Entropy,
○ H(X) = -∑i C∈ [P(xi) * log(P(xi))]
● Strength of Association via Pointwise Mutual Information, ○ PMI(x,S) = log(P(x,S)/{P(x)*P(S)})○ SOA(x,S) = PMI(x,S) - PMI( x,S)⅂
![Page 16: Twitter Sentiment Analysis - Indian Institute of ... · Twitter Sentiment Analysis Text length Topic relevance Noisy text Data sparsity Negation Stopwords Tokenization Multilingual](https://reader033.fdocuments.in/reader033/viewer/2022050213/5f5fcb08f376a44e181cd32b/html5/thumbnails/16.jpg)
Latent Representation Methods● Eigen Value Decomposition (EVD)
● Singular Value Decomposition (SVD)
● Word Embedding via Word2Vec, etc.
![Page 17: Twitter Sentiment Analysis - Indian Institute of ... · Twitter Sentiment Analysis Text length Topic relevance Noisy text Data sparsity Negation Stopwords Tokenization Multilingual](https://reader033.fdocuments.in/reader033/viewer/2022050213/5f5fcb08f376a44e181cd32b/html5/thumbnails/17.jpg)
➔ Machine Learning
➔ Lexicon-based
➔ Hybrid-based
Classification Approach
![Page 18: Twitter Sentiment Analysis - Indian Institute of ... · Twitter Sentiment Analysis Text length Topic relevance Noisy text Data sparsity Negation Stopwords Tokenization Multilingual](https://reader033.fdocuments.in/reader033/viewer/2022050213/5f5fcb08f376a44e181cd32b/html5/thumbnails/18.jpg)
DNN Classification approach
Convolution Neural Networkd = embedding dimensionm = window sizes = max lengthn = no. of filters
Paper Title: Twitter Sentiment Analysis with Deep Convolutional Neural Networks - Aliaksei Severyn and Alessandro Moschitti
![Page 19: Twitter Sentiment Analysis - Indian Institute of ... · Twitter Sentiment Analysis Text length Topic relevance Noisy text Data sparsity Negation Stopwords Tokenization Multilingual](https://reader033.fdocuments.in/reader033/viewer/2022050213/5f5fcb08f376a44e181cd32b/html5/thumbnails/19.jpg)
➔ Accuracy
➔ Precision
➔ Recall
➔ F-score
Evaluation Metrics
![Page 20: Twitter Sentiment Analysis - Indian Institute of ... · Twitter Sentiment Analysis Text length Topic relevance Noisy text Data sparsity Negation Stopwords Tokenization Multilingual](https://reader033.fdocuments.in/reader033/viewer/2022050213/5f5fcb08f376a44e181cd32b/html5/thumbnails/20.jpg)
Related fields● Twitter-based Opinion Retrieval
● Tracking Sentiment over Time
● Irony Detection on Tweets
● Emotion Detection on Tweets
● Tweet Sentiment Quantification
![Page 21: Twitter Sentiment Analysis - Indian Institute of ... · Twitter Sentiment Analysis Text length Topic relevance Noisy text Data sparsity Negation Stopwords Tokenization Multilingual](https://reader033.fdocuments.in/reader033/viewer/2022050213/5f5fcb08f376a44e181cd32b/html5/thumbnails/21.jpg)
References● Like It or Not: A Survey of Twitter Sentiment Analysis Methods
(Authors: Anastasia Giachanou, Fabio Crestani)
● Sentiment Analysis and Opinion Mining(Author: Bing Liu)
![Page 22: Twitter Sentiment Analysis - Indian Institute of ... · Twitter Sentiment Analysis Text length Topic relevance Noisy text Data sparsity Negation Stopwords Tokenization Multilingual](https://reader033.fdocuments.in/reader033/viewer/2022050213/5f5fcb08f376a44e181cd32b/html5/thumbnails/22.jpg)
Thank you