C1
Sentimental Analysis Overview
Chapter 1: Introduction
1.1 What is Sentimental Analysis
Sentimental is also known as opinion mining. It is an NLP (Natural
language processing) technique used to determine the
sentiments/emotions expressed in a text. It involves analyzing text to
clarify positive, negative, and natural based on the given text. The
main goal is to understand the opinion, emotion, and tone conveyed in
the provided text and or the entire topic.
Key aspects of sentimental analysis:
Text classification: Analysis categorizes text into various
categories such as positive, negative, and natural tone.
Advanced systems may classify sentiments into more natural
categories such as precise positive and mild negative.
Applications: Sentimental analysis is widely used in monitoring
social media, feedback from customers, research related to the
market, and monitoring brands. Sentimental analysis is used in
many organizations to make reasonable outcomes of public
opinion, customer experience understanding, and making
decisions based on data-driven.
Techniques: It can be approached using various techniques such
as ml Algorithms, lexicon-based methods, and hybrid
approaches for improving accuracy by combining multiple
techniques.
Challenges: Challenges in this analysis include removing
sarcasm, irony, and context-specific text, and adopting
languages with different dialects.
1|Page
C1
Sentimental Analysis Overview
1.2 Concepts of sentimental analysis:
1. Sentiment Categories
Positive: Text expresses a favourable opinion or emotion.
Negative: Text conveys an unfavourable opinion or emotion.
Neutral: The text does not show any strong opinion or emotion.
2. Granularity
Document-level Sentiment Analysis: Determines the overall
sentiment of an entire document.
Sentence-level Sentiment Analysis: Analyses the sentiment of
individual sentences.
Aspect-based Sentiment Analysis: Identifies sentiment concerning
specific aspects or features within the text.
3. Methods
Rule-based Approaches: Use predefined lists of words and
linguistic rules to determine sentiment.
Machine Learning Approaches: Utilize algorithms that learn from
data to classify sentiment. Common techniques include:
Supervised Learning: Involves training a model on labeled data
(e.g., using algorithms like Naive Bayes, Support Vector
Machines, or neural networks).
Unsupervised Learning: Clusters data without predefined labels
(less common in sentiment analysis).
2|Page
C1
Sentimental Analysis Overview
4. Deep Learning
Recurrent Neural Networks (RNNs): Especially Long Short-Term
Memory Networks (LSTMs) and Gated Recurrent Units (GRUs)
are effective in capturing contextual information in sequences of
text.
Convolutional Neural Networks (CNNs): Used for capturing local
patterns in text.
Transformers: Models like BERT (Bidirectional Encoder
Representations from Transformers) and GPT (Generative Pre-
trained Transformer) have significantly advanced sentiment
analysis by understanding context better.
3|Page
C1
Sentimental Analysis Overview
1.3 Importance of Sentimental Analysis:
Sentimental analysis has its importance across various domains and
industries.
Customer Insights and Feedback: Businesses may use customer
reviews, social media comments, and poll findings to understand
better customer attitudes regarding products, services, and
brands. This aids in acknowledging strengths, flaws, and
possibilities for growth.
Brand Monitoring & Reputation Management: Watching social
media and news sentiment enables businesses to manage their
brand reputation proactively. They can respond promptly to
unfavorable sentiments or capitalize on positive remarks.
Market and Competitive Research: Sentiment analysis may help
you understand market developments, customer preferences, and
rivalries' performance. Organizations may make data-driven
decisions about strategies for advertising, product development,
and competitive positioning.
Customer Service Optimization: Analyzing customer assistance
contacts (e.g., emails, chats) can assist in determining customer
satisfaction levels and typical trouble spots. Such data can help
improve service quality and availability.
Political Analysis and Public Opinion: Sentiment analysis is
used in political campaigns to gauge public opinion, evaluate
voter sentiment, and determine the influence of policy actions.
Risk Management: Financial institutions and investors may use
mood research to measure sentiment about the markets,
anticipate fluctuations in the market, and manage risk associated
with investments based on the opinions of firms or sectors.
Healthcare and Public Health: Analyzing opinions on social
media and healthcare forums may reveal public health patterns,
feedback from patients, and attitudes regarding medical care or
medical professionals.
Sentiment-Driven Promotional Campaigns: Businesses can
modify marketing campaigns based on sentiment assessment
4|Page
C1
Sentimental Analysis Overview
results, ensuring that messages resonate well with target
audiences and enhance engagement.
Sentiment-Based Content The collection: Media companies and
content providers employ sentiment assessment to select news
stories, social media postings, and other material based on
reader preference.
Real-time Surveillance and Crisis Managing: Rapid detection of
negative opinion patterns enables firms to respond quickly and
deploy crisis management measures, reducing feasible
reputation or brand harm.
5|Page
C1
Sentimental Analysis Overview
1.4 Techniques Used in Sentimental Analysis
Text preprocessing:
Tokenization is a method of breaking together text into smaller
components, such as phrases or words.
Normalization is the procedure of turning text into a
standardized format, such as "won't" to "will not".
Stop word Removal: The removal of frequent words (like "and"
and "the") that have no essential importance.
Originating and the lemmatization: Reduce words to their
fundamental or root form (for example, "running" becomes
"run").
Feature Extraction:
Bag of Words (BoW): Presenting text as a collection of words,
disregarding the syntax and word order yet conserving
multiplicity.
Term frequency-Inverse Book Frequency (TF-IDF): Measure
the significance of a term in a document in contrast to a
collection of papers.
Lexicons and Dictionaries for Sentiment Analysis:
Sentiment Lexicons are word collections categorized by their
sentiment polarity (positive, negative, neutral).
Senti Word Net, AFINN, and VADER are all sentiment lexicons
utilized in sentiment analysis.
Analysis focusing on different aspects of a subject to determine
sentiments related to those specific aspects.
Examining emotions in greater detail, concentrating on
particular elements or characteristics of a product, service, or
subject.
Approaches based on rules:
Classifying sentiment through linguistic rules or patterns that are
predefined.
6|Page
C1
Sentimental Analysis Overview
Methods that combine different approaches.
Merging various methods (such as lexicon-based and machine
learning) to enhance precision and resilience.
Assessment Criteria:
Evaluating sentiment analysis models' performance through
metrics such as accuracy, precision, recall, and F1-score.
7|Page
C1
Sentimental Analysis Overview
1.5 The algorithm used in the sentimental analysis
Sentimental analysis algorithms are broadly categorized into various
types on the basis of their approach and underlying techniques.
Naïve bayes classifier:
This method uses Bayes' theorem based on substantial (naive)
independence constraints to evaluate the text's traits (words). It's
simple, quick, and frequently used as an initial model for text
classification tasks like sentiment evaluation.
SVM (Support vector Machine):
SVMs are models of supervised learning that use data to
perform classification and regression analysis. They may be
utilized for sentiment analysis by converting incoming data into
a high-dimensional feature space.
Logistic regression forecasts a binary cause (e.g., positive or
negative sentiment) utilizing text attributes as variables.
Decision trees repeatedly allocate data into segments that utilize
attribute values, making them beneficial to sentiment analysis.
8|Page
C1
Sentimental Analysis Overview
Random Forest is an assembly of decision trees. They use many
decision trees to increase broadness and resilience. Random
forests can be applied to sentiment analysis jobs to improve
accuracy.
9|Page
C1
Sentimental Analysis Overview
1.6 Applications of Sentimental Analysis
The application of sentimental analysis involves extracting
information from the textual data and forming it into useful
information. This information will then clarify the positive, negative
and netural tone.
Businesses use sentiment analysis to analyze customer tests,
surveys, and social media comments to better understand how
customers feel, uncover product flaws, and improve the whole
customer experience.
Organizations look at social media, headlines, and online forums
to determine public opinion on their brand. This helps in brand
reputation administration, fixing bad sentiment swiftly, and
using useful sentiment for promotional purposes.
Sentiment analysis gives data on customer preferences,
movements, and behavior. Companies utilize this information
for better marketing tactics, create new goods, and stay ahead of
their rivals.
Social media networks create enormous quantities of textual
data. Sentiment analysis is useful in determining public
sentiment, identifying influencers, and analyzing the effect of
social media initiatives.
During polls or campaigns in politics, sentiment evaluation is
used to evaluate public opinion on candidates, policies, and
party affiliations. It aids in analyzing voter feelings and
predicting election results.
Sentiment analysis of customer service interaction (e.g., emails,
chats, or call transcripts) helps firms by assessing customer
satisfaction levels, identifying continuing issues, and enhancing
response times.
In finance, the sentiment analysis of articles, online forums, and
market reports assists traders and investors in gauging market
mood, predicting market moves, and making enlightened
investment decisions.
10 | P a g e
C1
Sentimental Analysis Overview
Sentiment analysis is used to assess patient comments, reviews
of healthcare services, and views about medical treatments. It
helps healthcare practitioners for enhancing patient care while
discovering the potential for service improvement.
Advertisers use sentiment monitoring to personalize promotions
depending on what consumers think. Assessing audience
sensation enables them to produce more tailored yet wealthy
messages for marketing.
11 | P a g e
C1
Sentimental Analysis Overview
1.7 Examples
Illustrated some of the examples of the sentimental analysis.
Ex 1: Product Review - Text: “I love this phone! The battery life is
amazing and the build quality is top-notch”
Analysis result: Positive
Sentiments: Love, Amazing, and top-notch.
Ex 2: Social-media - Text: “I can’t believe the service at the
restaurant is the worst experience ever.”
Analysis result: Negative
Sentiments: Worst, can’t believe.
Ex 3: News Article – Text: “The economy is showing signs of
recovery, with job growth exceeding expectations in the last
quarter.”
Analysis: Positive
Sentiments: Signs of recovery, Exceeding expectations.
12 | P a g e
C1
Sentimental Analysis Overview
1.8 Background
This section essentially explains the entire theoretical framework for
sentiment analysis, which encompasses all of the conventional and
numerous learning techniques. A review follows.
Theoretical Background
This background section presents all of the sentiment detection
techniques that are widely utilized and highly successful classification
systems that have been employed in the past, highlighting the most
well-liked ones.
TRADITIONAL CLASSIFICATION METHOD
Sentiment analysis text categorization has demonstrated a wide range
of previous supervised learning techniques, such as support vector
machines, logistic regression, and naïve Bayes. Twitter is the most
widely used and productive social media site in this situation. People
can always discover the finest viewpoint and honestly express their
feelings through both short and long messages on Twitter. The best
survey on Twitter indicates that it is between 75% and 80% accurate,
and the data is concluded based solely on this accuracy. While certain
businesses and works are conducted using social media platforms like
YouTube, it is discovered that Twitter offers the most accurate events.
TEXT ANALYSIS USING DEEP LEARNING
Given the numerous sentiment analysis shortcomings, the researchers
simply adopted the mentality that they needed to discover more
sophisticated and customized approaches to handling text and data.
For efficient sentiment analysis, deep learning has shown to be the
most potent and sophisticated version. It uses NLP algorithms to
execute all of the tasks with remarkable task and success.
13 | P a g e
C1
Sentimental Analysis Overview
Deep learning is a traditional machine learning approach in which all
computational modules are gathered and composed of numerous
layers known as non-linear layers, which are then converted into a
low-level representation of the higher-level data. Working with neural
networks has produced CNNs, RNNs, and several variants.
RECURRENT NEURAL NETWORK
RNNs are a particular class of neural network used in deep learning
that retains all of the information about the input data that is
processed for subsequent and future inputs in memory. In deep
learning, this is also referred to as reliance capturing. Sequential data
is entered into the RNN.
14 | P a g e
