B04 Sentiment Analysis

Question 1

What is Sentiment Analysis?

Answer 1

Sentiment analysis (or
opinion mining) is the
process of extracting an
author’s emotional intent
from text.

The task of sentiment analysis is not only about finding
the opinions about the whole entity but also the opinions about individual attributes of the entity and summarizing them.

Question 2

Sentiment Analysis

Individual Attributes

Answer 2

aspects

Question 3

Sentiment Analysis

The person making the opinion

Answer 3

Opinion holder

Question 4

Sentiment Analysis

The nature of the sentiment expressed

Answer 4

orientation or polarity

Question 5

Sentiment Analysis

The entity or aspect that the opinion is expressed about

Answer 5

Opinion target

Question 6

Some attributes of sentiment analysis

Answer 6

1. Does a piece of text represent a positive
or a negative sentiment?
2. What are the entities being discussed
about, and are they being discussed
about in a positive or negative way?
3. What attributes of the
entity are discussed, and
what are the sentiments
expressed about them?
4. What do people think about this candidate or issue?

Question 7

Challenges of Sentiment Analysis

Answer 7

• Cultural and demographic differences between authors.
• Discerning between feature-specific sentiment.
• Quantifying the hundreds of emotional states which are part of the human condition.

Question 8

Plutchik emotion
framework classifies 8
evolutionary emotions

Answer 8

Anger, Fear, Anticipation, Surprise, Joy, Sadness, Trust, Disgust

Question 9

Document Polarity

Answer 9

• Instead of trying to predict emotional states, an easier
  approach is to simply state whether a document is positive or negative.
• This is referred to as the polarity of a document.
• The approaches to calculating polarity vary and can either be fairly straightforward or rather sophisticated.

Question 10

Opinion Words

Answer 10

A straightforward approach to calculating polarity involves the use of certain words that are associated with a particular emotional state.

Question 11

Some examples of opinion words

Answer 11

Opinion words are often adjectives and adverbs (e.g. “good”, “bad”, “excellent”, etc.), although nouns (e.g., “trash”) or verbs (e.g., “annoy”) are also sometimes used.

Question 12

Opinion or Sentiment Lexicon

Answer 12

• A collection of opinion words along with their polarity form what is known as an opinion or sentiment lexicon.
• Sentiment lexicons are created either through crowd
  sourcing or by the labor of an author, and then validated by crowd sourcing or research.
• We can calculate the polarity of a document by simply
  adding up positive words and subtracting negative words.

Question 13

Bing

Answer 13

Categorizes words in
a binary fashion into
positive and negative
categories.

Question 14

AFINN

Answer 14

Assigns words with a
score that runs
between -5 and 5,
with negative scores
indicating negative
sentiment and
positive scores
indicating positive
sentiment.

Question 15

NRC

Answer 15

Categorizes words
into categories of
positive, negative,
anger, anticipation,
disgust, fear, joy,
sadness, surprise,
and trust.

Question 16

Loughran

Answer 16

Categorizes words
into categories of
negative, litigious,
positive, uncertainty,
constraining and
superfluous.

Question 17

What does the get_sentiments() function do?

Answer 17

• Returns a specific sentiment lexicon in tidy format.
• Values for lexicon are either “afinn”, “bing”, “nrc”, or “loughran”.
• Other than using the get_sentiments(), we can also refer to the sentiments dataset directly.

Question 18

Note with sentiments that:

Answer 18

• Not every English word is represented in the lexicons.
• The words do not take into account qualifiers. For example “no good” or “not true”.
• The size of the text analyzed can have an impact on the results.

Question 19

Sentence Level Sentiment Analysis

Answer 19

Takes valence shifters into consideration in an effort to do more accurate sentiment analysis.

Question 20

What are valence shifters?

Answer 20

Valence shifters are words that have an impact on the

overall polarity of a message.

Question 21

The 4 categories of valence shifters:

Answer 21

1. negators
2. amplifiers
3. de-amplifiers
4. adversative conjunctions

Question 22

Negators

Answer 22

Flip the sign of the polarized word.

“I do not love apple pie.”

Question 23

Amplifier

Answer 23

Increases or intensifies the impact of a polarized word.

“I really love apple pie.”

Question 24

De-amplifier

Answer 24

Reduces the impact of a polarized word.

“I hardly like apple pie.”

Question 25

Adversative Conjunction

Answer 25

Overrules the previous clause containing a polarized word.

“I would love to bake apple pie but it’s not worth it.”

Question 26

What does the sentimentr package do?

Answer 26

• Designed to quickly calculate text
  polarity sentiment at the sentence level.
• Optionally allows for aggregation by
  rows or grouping variable(s).

Question 27

What does the get_sentences() function do?

Answer 27

• Performs sentence boundary disambiguation.

- Returns a list of vectors of sentences.

Question 28

What does the sentiment() function do?

Answer 28

• Approximates the sentiment (polarity) of text by sentence.
• Several polarity and valence shifter dictionaries can be used.
• Returns data table of element_id, sentence_id, word_count, and sentiment.

Question 29

What does the sentiment_by() function do?

Answer 29

• Approximates the sentiment (polarity) of text by group.

- Returns data table of element_id, sentence_id, word_count, sd and ave_sentiment.