Text classification & sentiment analysis

Text Classification & Sentiment Analysis

Muhammad Atif QureshiArjumand Younus

2

Contents

● An Introduction to Text Classification

– Text Classification Examples

– Text Classification Methods● Naive Bayes

– Formalization

– Learning● Applications of Sentiment Analysis

● Baseline Algorithm for Sentiment Analysis

● Sentiment Lexicons

● Sentiment Analysis for the Political Domain (Personal Research)

3

Text Classification Examples

● News filtering and organization

● Document organization and retrieval

● Sentiment analysis/Opinion mining

● Email classification and spam filtering

● Authorship attribution

4

Spam Classification Example

Slide borrowed from Coursera Lectures on “Natural Language ProcessingBy Prof. Dan Jurafsky

5

Text Classification

● Set of training documents D = {d1,....,dN} such that each record is labeled with a class value 'c' from C = {c1,....,cJ}

● Features in training data are related to labels by means of classification model

● Classification model helps predict label for an unknown (test) record

● With text classification, model uses text-based features

6

Text Classification Methods

● Hand-coded rules

● Supervised machine learning

– Naive bayes

– Logistic regression

– Support vector machines

– K-nearest neighbors

7

Naive Bayes

● Simple (“naive”) classification method based on Bayes rule

● Relies on simple document representation namely bag of words

I love this movie. It's sweet but with satirical humor. The dialogue Is great and the adventure scenes are great fun...It manages to be whimsical and romantic while laughing at the conventions of the fairy tale genre. I would recommend it to just about anyone. I've seen it several times as I love it so much, and I'm always happy to see it again whenever I have a friend who hasn't seen it yet.

8

Bag of Words Representation: Subset of Words

I love this movie. It's sweet but with satirical humor. The dialogue is great and the adventure scenes are great fun...It manages to be whimsical and romantic while laughing at the conventions of the fairy tale genre. I would recommend it to just about anyone. I've seen it several times as I love it so much, and I'm always happy to see it again whenever I have a friend who hasn't seen it yet.

great 2

love 2

recommend 1

laugh 1

happy 1

..... ....

9

Bayes' Rule Applied to Documents and Classes

● For a document d and a class c

P (d /c )P (c )

P(d)P(c /d) =

10

Naive Bayes Classifier (1/3)

CMAP argmax P(c /d)= c∈C

argmaxP (d /c )P (c )

P(d)=

c∈C

argmax P(d /c)P(c)c∈C

=

11

Naive Bayes Classifier (2/3)

CMAP =

=

argmax P(d /c)P(c)c∈C

argmax P(x 1, x2,. . , xn /c)P(c)c∈C

Document represented asfeatures x1....xn

How often does this class occur?We can just count the relativefrequencies in a corpus.

12

Naive Bayes Classifier (3/3)

CMAP =

=

argmax P(d /c)P(c)c∈C

argmax P(x 1, x2,. . , xn /c)P(c)c∈C

O(|Xn|.|C|) parameters

Could only be estimated if a very,very large number of training exampleswas available.

argmax P(x 1, x2,. . , xn /c)P(c)

13

Multinomial Naive Bayes Independence Assumptions

Bag of Words assumption: Assume position doesn't matter

● Conditional Independence: Assume the feature probabilities P(xi/cj) are independent given the class c.

P(x 1,x 2,. . , xn /c)

P(x 1,x 2,. . , xn /c)=P(x 1/c )x .....P (xn/c )

14

Multinomial Naive Bayes Classifier

positions ← all word positions in test document

cNB

= cj∈C

argmax P(cj) ∏i∈positions

P(xi /cj)

15

Multinomial Naive Bayes Classifier

CMAP =

argmax P(x 1, x2,. . , xn /c)P(c)c∈C

argmax P(cj)∏x ∈X

P (x /c )c∈C

cNB

=

16

Learning the Multinomial Naive Bayes Model

● First attempt: maximum likelihood estimates

– simply use frequencies in the data

17

Parameter Estimation

● Create mega-document for topic j by concatenating all docs in this topic

– Use frequency of w in mega-document

18

Problem with Maximum Likelihood

● What if we have seen no training documents with the word fantastic and classified as positive

● Zero probabilities cannot be conditioned away, no matter the other evidence!

19

Laplace (add-1) Smoothing for Naive Bayes

20

Multinomial Naive Bayes: Learning

● From training corpus, extract Vocabulary

21

Multinomial Naive Bayes: A Worked Example

22

Sentiment Analysis Overview

23

Sentiment Analysis Applications (1/4)

● Movie: is this review positive or negative?

● Products: what do people think about the new iPhone?

● Public sentiment: how is consumer confidence? Is despair increasing?

● Politics: what do people think about this candidate or issue?

● Prediction: predict election outcomes or market trends from sentiment

24

Sentiment Analysis Applications (2/4)

25

Sentiment Analysis Applications (3/4)

26

Sentiment Analysis Applications (4/4)

27

Formal Definition of Sentiment Analysis● Sentiment analysis is the detection of attitudes

“enduring, affectively colored beliefs, dispositions towards objects or persons”

1. Holder (source) of attitude

2. Target (aspect) of attitude

3. Type of attitude

➢ From a set of types• like, love, hate, value, desire, etc.

➢ Or (more commonly) simple weighted polarity:• positive, negative, neutral together with strength

4. Text containing the attitude

➢ Sentence or entire document

28

Sentiment Analysis Tasks

● Simplest:

– Is the attitude of this text positive or negative?● More complex:

– Rank the attitude of this text from 1 to 5● Advanced:

– Detect the target, source, or complex attitude types

29

Sentiment Analysis: A Baseline Algorithm

● Polarity detection in movie reviews:

– Is an IMDB movie review positive or negative?● Data: Polarity Data 2.0:

– http://www.cs.cornell.edu/people/pabo/movie-review-data/

30

Baseline Algorithm (adapted from Pang and Lee)

● Tokenization

● Feature Extraction

● Classification using different classifiers

– Naive Bayes

– MaxEnt

– SVM

31

Sentiment Tokenization Issues

● Deal with HTML and XML markup

● Twitter markup (names, hash tags)

● Capitalization (preserve for words in all caps)

● Phone numbers, dates

● Emoticons

32

Extracting Features for Sentiment Classification

● How to handle negation

– I didn't like this movie

vs

– I really like this movie● Which words to use?

– Only adjectives

– All words

33

Negation

● Add NOT_ to every word between negation and following punctuation:

Didn't like this movie, but I

Didn't NOT_like NOT_this NOT_movie but I

34

Reminder: Naive Bayes

35

Sentiment Lexicons

● Dictionary of well-known “sentiment” words

– Abusive terms

– Adjectives like bad, worse, good, better, ugly, pretty● Available for use in research

– LIWC: Linguistic Inquiry and Word Count

– SentiStrength

– Bing Liu's Opinion Lexicon

36

My Research: Election Trolling on Twitter (Pakistan Elections 2013)

Twitterer Tweet

A @B Yeh...#Shame with fake account, this is how PTIians think they will get votes

B @A Stop making a fuss and fuck off.

A @B A dumb leader like IK can produce followers like you.

B @A A corrupt leader like Noora can hire paid trolls like you