This document discusses machine learning approaches for sentiment analysis. It begins by defining sentiment analysis as identifying the orientation of opinions in text through predicting the attitude, opinions, and emotions. The objective is to determine a writer's attitude on a given topic by analyzing text at the document, sentence, and phrase level. Feature selection methods and sentiment classification techniques are discussed, including lexicon-based approaches using dictionaries and corpora, and machine learning approaches using supervised and unsupervised learning with classifiers like naive Bayes and SVMs. Deep learning models for sentiment analysis including CNNs, RNNs, and LSTMs are also covered. The document concludes by discussing applications and potential future work exploring the cognitive aspects of sentiment analysis.
1. MACHINE LEARNING BASED
APPROACH FOR SENTIMENT ANALYSIS
Presented by
Alpna Patel
Guided by
Dr. Arvind kumar Tiwari
(Associate Professor)
KNIT, Sultanpur(228118)
2. WHAT IS SENTIMENT ANALYSIS?
Identify the orientation of opinion in a piece of text
It helps to understand the attitude, opinions and emotion
in the text.
It involves predicting or analysing the hidden information
present in the text.
Sentiment analysis can also be applied to audio, images
and videos.
The movie
was fabulous!
The movie
stars Mr. X
The movie
was horrible!
3. OBJECTIVE
The objective of sentiment analysis is to determine the
attitudes of a writer or a speaker for a given topic.
Sentiment analysis aims at getting sentiment-related knowledge
especially from the huge amount of information on the internet.
Can be generally used to understand opinion in a set of
documents.
4. IMPORTANT NOTATIONS
Subjectivity: Identify the subjective and objective
text.
Polarity: It is related to positive text, negative text
and neutral text.
Sentiment level: Sentiment analysis can be
performed at various level such as document level,
sentence level and phrase level.
5. FEATURE SELECTION METHODS
N-grams: It refers n-terms in text. It could be
unigram and bigram up to n accordingly.
POS tagging: Noun, pronouns, adjectives etc are
the example of parts of speech. It holds most of the
sentiment in text.
Stemming: It is the process of removing prefixes
and suffixes.
Stop words: Articles(a, the, an), prepositions are
stop words.
Negation handling: Negation words like ‘not’
inverts the meaning of whole sentence.
6. SENTIMENT CLASSIFICATION TECHNIQUES
Here the different techniques:
Sentiment Analysis
Machine Learning Approach Lexicon Based
Approach
Corpous Based Dictionary
Based
Supervised
Learning
Unsupervised
Learning
SemanticStatistical
Probabilistic
Classifier
Linear
Classifier
Rule Based
Classifier
Decision Tree
Classifier
Bayesian
Network
Naïve Bayes
Maximum
Entropy
Neural
Network
SVM
7. CONTINUE…
Lexicon based approach: It deals with counting
the number of positive and negative words in the
text. It uses two approaches named Dictionary
based approach, Corpus based approach.
Machine learning based approach: This is an
automatic classification technique. Classification is
performed using text features.
Supervised: System is trained using labeled
training . Each class represents different features
and has a label associated with it.
8. CONTINUE…
Supervised method uses different algorithms to
train and test datasets named as Naïve Bayes,
Maximum entropy, SVM, Neural Network etc.
9. DEEP LEARNING BASED MODELS
Deep Learning was firstly proposed by G.E. Hinton
in 2006.
Deep learning networks are capable for providing
training to both supervised and unsupervised
categories
Neural networks are very beneficial in text
generation, sentence classification, sentence
modeling and feature presentation.
Deep learning includes many networks such as
CNN, RNN , Recursive Neural Networks, DBN
(Deep Belief Networks) and many more.
10. CONTINUE…
Convolutional Neural Network(CNN): Convolutional
Neural Networks are very similar to ordinary Neural
Network.
They are made up of neurons that have learnable
weights and biases.
Figure: CNN
11. CONTINUE…
Recurrent Neural Network:
Each node at a time step takes an input from the
previous node and this can be represented using a
feedback loop.
Backpropagation in recurrent neural networks
occurs in the opposite direction of the arrows
12. CONTINUE…
The disadvantage with RNN is that as the time
steps increase, it fails to derive context from time
steps which are much far behind.
To solve above problem, it uses LSTM(Long Short
term memory)
13. APPLICATIONS
Review-related analysis.
Question-answering (Opinion-oriented questions
may involve different treatment).
Support in decision making.
Business application.
Predictions and trend analysis.
14. CONCLUSION & FUTURE WORK
Lexical Resources have been developed to
capture sentiment-related nature.
Subjective extracts provide a better
accuracy of sentiment prediction.
Several approaches use algorithms like
Naïve Bayes, clustering, etc. to perform
sentiment analysis.
The cognitive angle to Sentiment Analysis
can be explored in the future.
15. REFERENCES
Opinion Mining and Sentiment Analysis, Foundations and Trends in
Information Retrieval, B. Pang and L. Lee, Vol. 2, Nos. 1–2 (2008) 1–
135, 2008.
Bo Pang, Lillian Lee; ‘A Sentimental Education: Sentiment Analysis
Using Subjectivity Summarization Based on Minimum Cuts’;
Proceedings of the 42nd ACL; pp. 271–278; 2004.
https://towardsdatascience.com/introduction-to-sequence-models-
rnn-bidirectional-rnn- lstm-gru-73927ec9df15
http://cs231n.github.io/convolutional-networks/
https://towardsdatascience.com/convolutional-neural-networks-from-
the-ground-up-c67bb41454e1
https://pdfs.semanticscholar.org/8892/24a64a5bc5f9e965f418a63b6
768f7164993.pdf