INDUSTRY ORIENTED MAJOR PROJECT REPORT

ON

SOCIAL MEDIA OPINION ANALYSIS USING NLP

Submitted in Partial Fulfilment of the Requirements for
The Award of
BACHELOR OF TECHNOLOGY

IN

COMPUTER SCIENCE & ENGINEERING

BY

S. SHIVAMANI (21C31A0558)

UNDER THE GUIDANCE OF

MR. G. SHIVA PRASAD
Assistant Professor, Department of CSE

DEPARTMENT OF COMPUTER SCIENCE AND ENGINEERING

BALAJI INSTITUTE OF TECHNOLOGY AND SCIENCE

Accredited by NBA (UG-CE, ECE, ME, CSE, EEE Programs) & NAAC A+ Grade
(Affiliated by JNTU Hyderabad and Approved by the AICTE, New Delhi)

NARSAMP ET, WARANGAL – 506331.

2021-2025

i
 BALAJI INSTITUTE OF TECHNOLOGY AND SCIENCE

Accredited by NBA (UG-CE, ECE, ME, CSE, EEE Programs) & NAAC A+ Grade
(Affiliated by JNTU Hyderabad and Approved by the AICTE, New Delhi)

NARSAMP ET, WARANGAL – 506331.

DEPARTMENT OF COMPUTER SCIENCE AND ENGINEERING

CERTIFICATE

This is to certify that S. SHIVAMANI (21C31A0558) of B. Tech IV–II Semester has

satisfactorily completed the Major Project entitled “SOCIAL MEDIA OPINION ANALYSIS
USING NLP”, as part of curriculum in Computer Science and Engineering during academic
year 2024 – 2025.

Supervisor HoD
Mr. G. SHIVA PRASAD Dr. BANDI KRISHNA
Assistant Professor, CSE Head of Department-CSE

Project Coordinator

Dr. V. RAMDAS EXTERNAL EXAMINER

Assistant Professor, CSE

ii
 ACKNOWLEDGEMENT

I would like to express my sincere gratitude to my guide Mr. G. SHIVA PRASAD

whose knowledge and guidance gave motivated me to achieve goals. He has consistently
been a source of motivation, encouragement and inspiration. The time I have spent working
under his supervision has truly been a pleasure.

I thank my HoD Dr. BANDI KRISHNA of his effort and guidance and all senior faculty
members for their help during my course. Thanks to programmers and non-teaching staff of
CSE Department of my college.

I heartily thank my Principal Dr. V. S. HARIHARAN for giving this great Opportunity
and his support to complete my project.

I would like to appreciate the guidance given by project coordinator Dr. V. RAMDAS as
well as the panels especially in my project presentation that has improved my presentation skills
by their comment and tips.

Finally Special thanks to my parents for their support and encouragement throughout life
and during the course. Thanks to all my friends and well-wishers for their constant support.

S. SHIVAMANI (21C31A0558)

iii
 ABSTRACT

Social Media Opinion Analysis using NLP is an advanced application aimed at

detecting sentiments expressed in user-generated content across various platforms. Sentiment
analysis plays a crucial role in understanding public opinion, brand monitoring, and market
research. This project leverages Natural Language Processing (NLP) techniques, specifically
using Long Short-Term Memory (LSTM) networks, to classify social media posts into three
categories: Positive, Negative, and Neutral. The model learns to identify patterns in the text
such as keywords, sentence structures, and contextual cues that help determine sentiment.
The Sentiment140 dataset, which consists of over 1.6 million labeled tweets, is used to
train the model. Preprocessing steps such as tokenization, padding, and text normalization are
applied to prepare the data for input. The LSTM model is employed due to its ability to capture
sequential dependencies in the text, making it well-suited for sentiment analysis tasks. The
model's effectiveness is evaluated through its accuracy and confidence scores, with results
demonstrating its ability to accurately classify sentiments in real-world social media content.
The system is deployed as a Flask web application, allowing users to input posts for real-time
sentiment analysis.

iv
 TABLE OF CONTENTS

S.NO. CONTENTS PAGE NO.

1. Introduction 1
1.1 Aim 1
1.2 Objective 2

2. Literature Survey 3

3. System Analysis 4
3.1 Methodology 4
3.2 Modules of the System 9
3.2.1 User Interface Module 9
3.2.2 Text Preprocessing & 10
Feature Extraction Module
3.2.3 Sentiment Analysis Module 12
3.2.4 Sentiment Analysis Interface Module 13
3.3 Existing System 14
3.4 Proposed System 14

4. Feasibility Study 15
4.1 Technical Feasibility 15
4.2 Operational Feasibility 16
4.3 Economical Feasibility 16

5. Software Requirement Specification 17

5.1 Requirement Analysis 17
5.1.1 Functional Requirements 17
5.1.2 Non-Functional Requirements 18
5.1.3 Performance Requirements 19
5.1.4 Software Requirements 20

v
5.2 Libraries 21

6. System Design 23
6.1 UML Diagrams 23
6.1.1 Use Case Diagram 24
6.1.2 Class Diagram 25
6.1.3 Sequence Diagram 26
6.1.4 Activity Diagram 27

7. Implementation 28
7.1 Code 29
7.2 Screenshots 38
8. Conclusion 44
9. References 45

vi
 Social Media Opinion Analysis Using NLP

1. INTRODUCTION

1.1 Aim

Social media platforms have become a central hub for communication, opinion sharing,
and information dissemination. With millions of posts made daily, the ability to analyze and
understand public sentiment has become invaluable for businesses, governments, and
organizations. Understanding the emotional tone behind social media posts allows for better
decision-making, targeted marketing, and timely responses to public opinion. However,
manually analyzing vast amounts of user-generated content is impractical, necessitating the
development of automated systems for sentiment analysis.
This project aims to address this challenge by developing an automated Social Media
Opinion Analysis System using Natural Language Processing (NLP). The system leverages
machine learning techniques, particularly Long Short-Term Memory (LSTM) networks, to
classify social media posts into three sentiment categories: Positive, Negative, and Neutral. The
model is trained on a robust dataset, Sentiment, containing millions of labeled tweets, enabling
it to accurately predict sentiment based on the content of social media posts. The project is
implemented using Python and various libraries such as TensorFlow and Keras for building the
model, as well as Flask for creating a web-based interface where users can input social media
posts and receive sentiment analysis results.
The significance of this project lies in its potential to automate the sentiment analysis
process, making it faster and more efficient. Traditional methods of analyzing sentiment often
require extensive human intervention, which can be both time-consuming and subjective. By
leveraging machine learning, this system can analyze large volumes of data in real time,
offering businesses and organizations valuable insights into public opinion with greater
accuracy and reliability. The project also integrates a user-friendly web interface, allowing
anyone to easily submit text for sentiment analysis.
As the reliance on social media grows, the ability to understand the sentiment of public
discourse becomes increasingly important. This system can assist in various applications, such
as product feedback analysis, social media monitoring, and crisis management. In summary,
this project represents a step forward in the field of NLP by developing an automated system
for sentiment analysis, offering significant benefits in terms of operational efficiency and
decision-making.
CSE Department, BITS 1
 Social Media Opinion Analysis Using NLP

1.2 Objective
Sentiment Classification (Positive, Negative, or Neutral):
The system is built to classify the sentiment of social media posts, such as tweets or
captions, into three categories: Positive, Negative, and Neutral. By analyzing the text and
identifying emotional tones, it helps in understanding public opinion on various topics. This
classification aids businesses, marketers, and organizations in making data-driven decisions
based on the sentiment trends of social media posts.
Automated Sentiment Analysis of Social Media Posts:
The project focuses on automating the process of sentiment analysis, which traditionally
required manual efforts. By leveraging Natural Language Processing (NLP) and Deep Learning
techniques, it analyzes social media content at scale without the need for human intervention,
thus enabling real-time sentiment monitoring. This system is well-suited for analyzing user-
generated content on platforms like Twitter, Instagram, and Facebook, offering valuable
insights into public perception.
The project achieves several key objectives:
 Sentiment Detection and Analysis:
The system reliably differentiates between positive, negative, and neutral sentiments
expressed in social media posts, providing valuable insights for brands, organizations,
and social media analysts.
 Process Automation:
Automating sentiment analysis reduces the reliance on manual review, streamlines
operations, and offers real-time sentiment feedback, making it possible to analyze large
volumes of data efficiently.
 Real-time Monitoring and Feedback:
By analyzing posts in real-time, the system provides instant feedback to users, enabling
businesses and organizations to monitor trends and adjust their strategies accordingly.
In conclusion, this project addresses the growing need for sentiment analysis in
the digital age by utilizing advanced machine learning and NLP techniques. It enables
accurate, efficient, and scalable analysis of social media content, offering actionable
insights to enhance decision-making, marketing strategies, and public relations efforts.

CSE Department, BITS 2

 Social Media Opinion Analysis Using NLP

2. LITERATURE SURVEY
Sentiment analysis has evolved significantly, progressing from rule-based systems to
machine learning, and now to deep learning methods. Early systems relied on sentiment
lexicons—predefined lists of positive and negative words—to identify the polarity of text.
While simple and interpretable, these rule-based approaches struggled with context, sarcasm,
slang, and negations.
With the rise of labeled datasets, machine learning methods such as Naive Bayes, Support
Vector Machines (SVM), and Logistic Regression became popular. These models learned from
data and used statistical features like Bag-of-Words or TF-IDF. However, they still lacked the
ability to capture semantic meaning or word order effectively.
Deep learning revolutionized sentiment analysis by introducing Recurrent Neural
Networks (RNNs) and Long Short-Term Memory (LSTM) networks, which are well-suited for
sequence data. LSTMs are capable of retaining contextual information across sentences,
making them ideal for analyzing the informal and brief nature of social media text. The use of
pre-trained embeddings such as Word2Vec, GloVe, and more recently, transformer-based
models like BERT, has further enhanced performance by capturing semantic and syntactic
relationships between words.
One of the most widely used datasets in this domain is the Sentiment dataset, which
contains 1.6 million labeled tweets and is extensively used to train and evaluate sentiment
classification models for Twitter data.
Recent studies focus on solving real-world challenges in sentiment analysis, such as
incorporating emoji sentiment (e.g., 😊 indicating positivity), handling phrase negations like
“not bad,” and detecting sarcasm—where literal and intended meanings differ. Advanced
models now integrate these complexities using deep learning and attention mechanisms.
In our work, we built upon these research insights by customizing our text preprocessing
to include emoji normalization, negation handling, and informal word corrections. We
experimented with both rule-based (VADER) and deep learning models (LSTM) using pre-
trained embeddings to ensure context-aware sentiment detection. These enhancements align
our project with current advancements in NLP for social media opinion mining.

CSE Department, BITS 3

 Social Media Opinion Analysis Using NLP

3. SYSTEM ANALYSIS

3.1 Methodology
The methodology adopted for this project encompasses a comprehensive pipeline from
data acquisition to deployment, carefully structured to ensure accurate sentiment analysis of
social media content using deep learning and Natural Language Processing (NLP) techniques.
The main stages involved in the methodology are:
1. Data Collection and Preprocessing:

Fig 3.1.1: Text Preprocessing

1.1 Dataset Used

The project utilizes the Sentiment dataset, a large and widely-used dataset for sentiment
analysis research. It consists of 1.6 million social media posts, each labeled as positive,
negative, or neutral, based on the sentiment expressed in the text. These labels are automatically
derived using emoticons present in the original tweets, which helps in building a robust training
dataset.

CSE Department, BITS 4

 Social Media Opinion Analysis Using NLP

1.2 Data Cleaning

Before training the model, the raw data must be cleaned to remove noise and irrelevant
elements. The following cleaning steps were performed:
 URL Removal: Any hyperlinks (e.g., "https://...") were removed as they do not
contribute to sentiment understanding.
 Username and Hashtag Removal: Handles like @user and hashtags such as #topic
were stripped out to focus purely on text content.
 Punctuation and Special Characters: All punctuation marks, special characters, and
digits were removed.
 Emoji Handling: While most emojis were removed for text analysis, some were
mapped to their corresponding sentiments using a sentiment-aware emoji dictionary to
improve expressiveness.
 Case Normalization: All text was converted to lowercase to maintain uniformity.
1.3 Tokenization
The cleaned text was then broken down into individual words (tokens) using Keras’s
Tokenizer. Tokenization is essential to convert text into a sequence of indices based on a
defined vocabulary. This allows the model to work with numerical inputs instead of raw strings.
1.4 Stemming and Lemmatization
To reduce dimensionality and improve learning efficiency, stemming and lemmatization
were applied:
 Stemming reduces words to their base/root form (e.g., “playing” → “play”).
 Lemmatization considers grammatical context and reduces words to their meaningful
root forms (e.g., “better” → “good”).
Both processes help in minimizing redundancy and improving semantic interpretation by the
model.
1.5 Stopword Removal
Common words such as “is,” “are,” “the,” and “of” (known as stopwords) were
removed. These words typically do not add much value to sentiment detection and are filtered
out to reduce noise.

CSE Department, BITS 5

 Social Media Opinion Analysis Using NLP

1.6 Padding and Sequence Standardization

All token sequences were padded to a fixed length (e.g., 50 or 100 tokens) using Keras’s
pad_sequences function. This step ensures that all input sequences are of equal length, which
is essential for LSTM model input.
2. Feature Extraction
Once preprocessing was complete, the next step was to extract meaningful features
from the text. These features serve as the input to the sentiment classification model.
2.1 Word Embeddings
Word embeddings are dense vector representations of words that capture semantic
meaning. Pre-trained embeddings like GloVe or Word2Vec were explored to enhance the
model’s understanding of word relationships. These embeddings allow the model to recognize
that words like "happy" and "joyful" are similar in meaning.
2.1.1 Sentiment Lexicons and Negations
Additional features included:
 Sentiment-bearing words: Words with known positive or negative sentiments, such as
"love," "hate," "disgusted," and "excited."
 Negation phrases: Patterns like "not good," "never liked," or "didn't enjoy" were
specially treated during preprocessing to reverse the sentiment where applicable.
These linguistic cues are critical in identifying subtle or complex sentiments that might
otherwise be misinterpreted.
3. Model Development:
A deep learning model based on Long Short-Term Memory (LSTM) networks is employed
for sentiment analysis. LSTM is a type of recurrent neural network (RNN) that is particularly
effective for sequence data, like text. The architecture includes the following components:
 Embedding Layer: The first layer is the embedding layer, which converts the
tokenized words into dense vectors that represent their meanings.
 LSTM Layer: The LSTM layer captures the sequential dependencies in the text,
learning long-range relationships between words.
 Dense Layer: A fully connected dense layer is added at the end of the network to
classify the sentiment into three categories: Positive, Negative, and Neutral.
CSE Department, BITS 6
 Social Media Opinion Analysis Using NLP

The model is trained using a loss function (e.g., categorical crossentropy) and an optimization
algorithm (e.g., Adam optimizer) to minimize errors during training.
3.1 Model Selection
A Long Short-Term Memory (LSTM) based deep learning model was selected due to its
strength in handling sequential data and capturing long-term dependencies. Unlike traditional
feedforward networks, LSTMs are designed to retain context, which is essential in sentiment
classification.
3.2 Architecture Overview
The architecture includes the following layers:
 Embedding Layer: Converts each tokenized word into its corresponding dense vector
(learned during training or initialized from pre-trained embeddings).
 LSTM Layer: Processes sequences of embeddings, learning contextual dependencies
between words.
 Dropout Layer: Helps reduce overfitting by randomly deactivating a fraction of
neurons during training.
 Dense Output Layer: A fully connected layer with softmax activation is used to predict
the final sentiment class (Positive, Negative, or Neutral).
3.3 Training Configuration
 Loss Function: categorical_crossentropy was used to handle multiclass classification.
 Optimizer: Adam optimizer was chosen for its efficiency and adaptive learning rate
capabilities.
 Batch Size & Epochs: The model was trained using a batch size of 64 and up to 10–
20 epochs, depending on validation performance.
4. Model Evaluation:
Once the sentiment analysis model has been trained on the labeled data, it is essential
to evaluate how well it performs on new, unseen content. This is done to ensure that the model
doesn't just memorize training data but can generalize effectively to real-world social media
posts.
To do this, a separate set of posts, not used during training, is used for testing. This test
data simulates how the system would behave when deployed and interacting with real users.
CSE Department, BITS 7
 Social Media Opinion Analysis Using NLP

By analyzing the model’s performance on this data, we can measure its accuracy and identify
areas where it may need improvement.
Assessing the Model’s Accuracy
The main goal of the evaluation is to understand how accurately the model predicts
sentiments — whether a given post is positive, negative, or neutral. The total number of correct
predictions is compared to the total number of posts in the test set. A high number of correct
predictions indicates that the model has learned well and can be trusted to analyze social media
content.
However, accuracy alone doesn't tell the whole story. Sometimes, the model might do
well on certain types of posts (for example, clear positive messages) but may struggle with
more subtle or ambiguous ones (like sarcastic or neutral statements). Therefore, it's important
to look deeper.
5. System Deployment
After successful model training and evaluation, the system was deployed as a web
application using the Flask framework. The deployment involved integrating the trained model
with a user-friendly frontend that allows real-time sentiment predictions.
Web Interface Features:
 Text Input: Users can enter any sentence or post for analysis.
 Backend Processing: The text is preprocessed and passed to the LSTM model in the
backend.
 Sentiment Output: The predicted sentiment (Positive, Negative, or Neutral) is
displayed along with a confidence score.
 Emoji Display: To enhance user experience, an appropriate emoji is displayed based

on the sentiment detected (😊, 😐, 😞).

The application is responsive and designed to work across devices, making sentiment
analysis accessible to end users in a simple and intuitive way.

CSE Department, BITS 8

 Social Media Opinion Analysis Using NLP

3.2 Modules of the System

The system developed for Social Media Opinion Analysis using Natural Language
Processing (NLP) is organized into several functional modules. Each module is responsible for
a specific part of the workflow, from user interaction to text preprocessing, sentiment
classification, and result display. These modules work in coordination to deliver accurate, real-
time sentiment analysis of user-provided text inputs.

3.2.1 User Interface Module:

The User Interface (UI) Module serves as the primary point of interaction between the
user and the sentiment analysis system. It allows users to input social media posts or comments
for sentiment analysis and displays the results in an intuitive and easy-to-understand manner.
The UI is built using Flask, HTML5, and CSS, designed to be user-friendly and responsive for
various devices.
Key Features:
 Text Input: Users can input social media posts or comments through a text box.
 Sentiment Display: Once the input text is processed, the UI displays the sentiment
analysis results, including the predicted sentiment (Positive, Negative, or Neutral) and
the corresponding confidence score.
 Interactive Features: The UI supports interactive elements such as a refresh button to
reset inputs and a history display to show the results of previous analyses.

 Emoji Visualization: Sentiments are represented with corresponding emojis (😊 for

Positive, 😞 for Negative, 😐 for Neutral) to make the results more engaging.

Technologies Used:
 Flask
 HTML5
 CSS
Interaction with Other Modules: The User Interface communicates with the Sentiment Analysis
Module via HTTP requests. Text input from the UI is sent to the backend for analysis, and the

CSE Department, BITS 9

 Social Media Opinion Analysis Using NLP

sentiment prediction along with the confidence score is returned to the UI for display.

3.2.2 Text Preprocessing and Feature Extraction Module

The Text Preprocessing and Feature Extraction Module processes raw social media text
and extracts features that are necessary for sentiment analysis. It involves cleaning and
transforming the text into a suitable format for the machine learning model, ensuring that only
relevant linguistic features are used for classification.
Key Features:
 Text Cleaning: Removes unnecessary elements such as URLs, special characters, and
stop words to focus on the meaningful content.
 Tokenization: Breaks the text into smaller units (tokens) such as words or phrases for
further processing.
 Vectorization: Converts the tokenized text into numerical vectors, using techniques like
word embeddings (e.g., Word2Vec or GloVe) to represent the semantic meaning of
words.
 Padding: Ensures that all input sequences have the same length by adding padding,
making the data suitable for feeding into the LSTM model.
Technologies Used:
 NLTK (Natural Language Toolkit)
 Keras (for tokenization and padding)
 GloVe or Word2Vec (for word embeddings)
Interaction with Other Modules: The processed text and feature vectors are passed on to the
Sentiment Analysis Model for training or prediction. These features enable the model to
understand and classify the sentiment expressed in the text.

CSE Department, BITS 10

 Social Media Opinion Analysis Using NLP

Fig 3.2.2: Workflow of Sentiment Analysis

CSE Department, BITS 11

 Social Media Opinion Analysis Using NLP

3.2.3 Sentiment Analysis Model Module

The Sentiment Analysis Model is the core of the system, responsible for classifying the
sentiment of social media posts or comments as Positive, Negative, or Neutral. The model uses
a Long Short-Term Memory (LSTM) network, a type of recurrent neural network (RNN)
designed for sequence data like text.
Key Features:
 LSTM Architecture: The model uses an LSTM layer to capture long-term dependencies
in the text, understanding context and sentiment over multiple words.
 Fully Connected Layers: After processing the sequence with the LSTM layer, the output
is passed through fully connected layers to make the final sentiment prediction.
 Softmax Activation: The final output layer uses a softmax activation function to classify
the sentiment into three categories: Positive, Negative, or Neutral.
 Training with Sentiment140 Dataset: The model is trained using a labeled dataset of
social media posts (Sentiment140) that has positive, negative, and neutral sentiment
labels.
Technologies Used:
 Keras (for LSTM model building)
 TensorFlow (for training and evaluation)
 NumPy (for data manipulation)
Interaction with Other Modules: The processed text features from the Text Preprocessing and
Feature Extraction Module are fed into the LSTM model, which predicts the sentiment. The
output is sent back to the Sentiment Analysis Interface Module for display.

CSE Department, BITS 12

 Social Media Opinion Analysis Using NLP

3.2.4 Sentiment Analysis Interface Module

The Sentiment Analysis Interface Module is responsible for taking the output from the
Sentiment Analysis Model and formatting it for presentation to the user. It manages the
communication between the model and the User Interface, ensuring that results are displayed
in a user-friendly way.
Key Features:
 Prediction Display: Displays the sentiment prediction (Positive, Negative, or Neutral)
along with the confidence score and corresponding emoji.
 Interactive Feedback: Provides the user with a visual indication of sentiment, such as
color coding (green for positive, red for negative, gray for neutral).
 Real-Time Updates: Once the user submits a social media post, the interface updates in
real-time with the sentiment analysis result.
Technologies Used:
 Flask (for routing and communication)
 HTML5/CSS (for result display)
 JavaScript (for interactive content updates)
Interaction with Other Modules: The module takes the sentiment prediction and confidence
score from the Sentiment Analysis Model and updates the User Interface with the results. It
ensures that the display is visually appealing and informative for the user.

CSE Department, BITS 13

 Social Media Opinion Analysis Using NLP

3.3 Existing System

The existing system for social media sentiment analysis primarily relies on traditional
machine learning algorithms like Naive Bayes, Support Vector Machines, and Logistic
Regression. These methods often depend on simple feature extraction techniques such as bag-
of-words or tf-idf, which are effective for structured text but fall short when dealing with
informal language and noisy data typical of social media. Existing systems also struggle with
handling contextual shifts in sentiment, such as sarcasm or irony, and often overlook the
importance of emojis, hashtags, and other contextual markers that play a significant role in
social media communication. Furthermore, these models are limited in terms of accuracy and
real-time analysis capabilities, making them less adaptable to the fast-paced and dynamic
nature of social media platforms.

3.4 Proposed System

The proposed system, on the other hand, utilizes an advanced LSTM-based deep
learning model to analyze social media posts. LSTM networks excel at capturing long-term
dependencies in text, making them better equipped to understand the context and nuances of
informal and complex language used in social media. The system integrates emojis and
hashtags, providing a richer understanding of the sentiment expressed in the post. Additionally,
it supports real-time sentiment analysis, allowing for immediate processing of user-generated
content. This approach not only enhances the accuracy of sentiment predictions but also
improves the model's robustness in dealing with noisy and unstructured data, making it more
suitable for analyzing large volumes of social media content.

CSE Department, BITS 14

 Social Media Opinion Analysis Using NLP

4. FEASIBILITY STUDY
The feasibility study for the proposed Social Media Opinion Analysis using NLP
project involves evaluating various factors that determine the practicality and success of the
system. These factors include technical feasibility, operational feasibility, economic feasibility,
and legal and ethical feasibility.
4.1 Technical Feasibility
The technical feasibility of this project is strong, as the required technologies and
frameworks for building and deploying the sentiment analysis system are readily available and
well-supported. The project uses a Long Short-Term Memory (LSTM) network, a state-of-the-
art approach in natural language processing, which excels in handling long-term dependencies
in text, making it ideal for analyzing the context and sentiment of social media posts. Python,
along with libraries such as TensorFlow, Keras, and NLTK, provides a comprehensive
environment for building deep learning models and handling textual data. Additionally, the
Flask web framework ensures smooth integration of the model with a user-friendly interface
for real-time analysis.
The project's integration of emoji analysis, contextual sentiment detection, and real-
time processing is technically feasible as there are pre-trained models and datasets that handle
these aspects. Existing datasets like Sentiment offer a reliable foundation for training the
sentiment analysis model. Moreover, the system's real-time capabilities are enabled by Flask’s
lightweight nature, which facilitates the smooth processing of data and response generation for
users. Thus, the required technical expertise and tools are available to develop and deploy the
system effectively.

CSE Department, BITS 15

 Social Media Opinion Analysis Using NLP

4.2 Operational Feasibility

Operational feasibility concerns whether the system can be successfully implemented
within the operational constraints of the intended environment. The proposed sentiment
analysis system can easily be deployed as a web application, accessible via any modern web
browser. This accessibility ensures that users can input text in real-time and receive immediate
11 sentiment analysis. The system’s integration with a Flask web interface ensures minimal
operational overhead, as it operates in an intuitive, user-friendly environment. The system's
real-time processing ability makes it highly applicable for monitoring and analyzing social
media conversations, which can be vital for various stakeholders such as marketing teams,
brands, or government agencies monitoring public opinion. With user friendly interaction, the
system is well-suited for integration into existing operational workflows without requiring
significant changes to the user experience or backend infrastructure.

4.3 Economic Feasibility

Economic feasibility refers to the cost-effectiveness of the system in terms of
development, deployment, and maintenance. The proposed system is economically feasible as
it makes use of open-source technologies like Python, Flask, TensorFlow, and Keras, which
eliminate the need for expensive proprietary software. The primary costs for development will
be related to computational resources (such as GPU for model training) and potential cloud
hosting costs for web deployment. Once the system is developed and tested, the operational
costs are expected to be relatively low, with the main ongoing costs being server maintenance
and computational power for model inference. Since the system’s design aims for efficient
resource usage and scalability, it can be deployed on affordable cloud platforms, thus keeping
operational costs manageable. Furthermore, the system’s automation of sentiment analysis in
real-time significantly reduces the manual effort involved in monitoring social media,
providing a cost-saving benefit in terms of time and human resources.

CSE Department, BITS 16

 Social Media Opinion Analysis Using NLP

5. SOFTWARE REQUIREMENT SPECIFICATION

The main purpose for preparing this document is to give a general insight into the
analysis and requirements of the existing system or situation and for determining the operating
characteristics of the system. This document plays a vital role in the development life cycle and
it describes the complete requirement of the system. It is meant for use by the developers and
will be the basic during testing phase. Any changes made to the requirements in the future will
have to go through formal changes and approval process.

The Software Requirement Specification (SRS) document outlines the functional and non-
functional requirements, performance requirements, software requirements, and libraries
needed to build the Social Media Opinion Analysis using NLP system. This section provides
detailed information about the system's functional behavior, user requirements, performance
expectations, and the software components used to build the system.

5.1 Requirement Analysis

5.1.1 Functional Requirements
The requirement analysis process involves identifying the user needs and system
functionalities that the proposed system must meet. These requirements guide the development
of the sentiment analysis system and ensure that it fulfills the intended objectives.
Functional Requirements:
 User Interaction: The system must allow users to input social media posts for sentiment
analysis, providing results in terms of sentiment classification (positive, negative,
neutral) and associated confidence scores.
 Real-time Processing: The system must process user inputs in real-time and provide
immediate feedback on the sentiment of the inputted post.
 Multi-Class Sentiment Detection: The system must classify the sentiment into three
categories: positive, negative, and neutral, while also detecting any related emojis or
contextual meanings.
 Visualization: Sentiment results should be displayed with visual cues such as color
coding (e.g., green for positive, red for negative, and yellow for neutral) along with
confidence scores.
CSE Department, BITS 17
 Social Media Opinion Analysis Using NLP

 Data Storage: The system should record historical sentiment analysis results and allow
users to review previous inputs.

5.1.2 Non-Functional Requirements

Non-functional requirements define the overall attributes of the system that ensure it meets
user expectations regarding usability, performance, and reliability. These are not about specific
functions but rather about how the system should perform and behave under different
conditions.
 Performance:
o The system should provide sentiment analysis results within 2-5 seconds for
each post.
o It should be able to handle up to 10,000 simultaneous users with minimal
degradation in performance.
o Latency should not exceed 3 seconds between user input and result display.
 Scalability:
o The system should be designed to scale horizontally to handle an increasing
volume of data, especially during social media trends or viral events.
o It should be able to expand the backend infrastructure (e.g., cloud resources) to
support a growing number of users.
 Availability:
o The system should have an uptime of at least 99.9%, ensuring that users can
reliably access it at any time.
 Security:
o The system should implement secure user authentication if required.
o Data encryption must be applied to protect sensitive user data (especially private
social media posts) during both transmission and storage.
o Regular security audits should be carried out to ensure that there are no
vulnerabilities in the system.
 Compatibility:
o The system should be accessible through modern web browsers such as Google
Chrome, Firefox, Safari, and Microsoft Edge.
CSE Department, BITS 18
 Social Media Opinion Analysis Using NLP

o It should support mobile devices as well, providing a responsive design that

adjusts to different screen sizes.
 Localization:
o The system should support multiple languages, at least for the analysis of posts
in English and other widely spoken languages such as Spanish, French, and
German.

5.1.3 Performance Requirements

Performance requirements focus on the system’s ability to perform under expected
conditions and ensure the system can handle a variety of operational demands.
 Response Time:
o The average response time for sentiment analysis should be under 5 seconds for
each post submitted by the user.
o Response times must remain consistent, even as the number of concurrent users
or requests increases.
 Throughput:
o The system should be capable of processing 1,000 posts per minute without
significant delays.
o At peak usage, the system should maintain an acceptable performance level,
processing sentiment analysis requests without exceeding a response time of 10
seconds per request.
 Concurrency:
o The system should support up to 10,000 simultaneous active users without
crashing or compromising performance.
o It must handle burst traffic, such as surges in posts related to trending topics or
events, while maintaining a high level of performance.

 Data Storage and Retrieval:

o The system must be able to store sentiment analysis results efficiently and
retrieve data within 2 seconds when requested by the user.

CSE Department, BITS 19

 Social Media Opinion Analysis Using NLP

 Model Accuracy:
o The sentiment analysis model should achieve an accuracy rate of at least 85%
in classifying posts as positive, negative, or neutral.
o The model should be fine-tuned regularly to improve accuracy and handle
domain-specific social media language or emerging trends.

5.1.4 Software Requirements

The software requirements specify the tools, frameworks, and technologies required to
develop and deploy the Social Media Opinion Analysis using NLP system. The software
components ensure the smooth functioning of various modules such as sentiment analysis,
model training, web interface, and database management.
 Programming Languages:
o Python: Python will be the primary programming language for building the
sentiment analysis model, handling data preprocessing, and developing the web
application using the Flask framework.
 Web Framework:
o Flask: Flask will be used for building the backend of the application, handling
user requests, serving the sentiment analysis results, and managing the
interaction with the frontend.
 Libraries:
o TensorFlow/Keras: For building and training the Long Short-Term Memory
(LSTM) model for sentiment analysis.
o NLTK: Natural Language Toolkit (NLTK) will be used for text preprocessing,
such as tokenization, stop word removal, and stemming.
o Pandas: For data manipulation and handling structured data such as user posts
and sentiment analysis results.
o Scikit-learn: For model evaluation and additional feature extraction if necessary.
 Data Processing & Preprocessing Tools:
o OpenCV: For image processing tasks like emoji extraction (if necessary).
o SpaCy: An alternative to NLTK for advanced NLP tasks such as named entity
recognition (NER) and dependency parsing.
CSE Department, BITS 20
 Social Media Opinion Analysis Using NLP

 Deployment Tools:
o Docker: For containerizing the application and ensuring consistent deployment
environments.
o AWS/GCP: For cloud hosting and deploying the web application and models at
scale.
o Heroku: For initial deployment during development and testing phases.
 Version Control:
o Git: For source code management and version control.

5.2 Libraries
To develop the Social Media Opinion Analysis using NLP system, a range of powerful
and open-source Python libraries have been utilized. These libraries provide essential
functionalities that support everything from data preprocessing and natural language
processing to model training, deployment, and advanced analysis. Below is a detailed overview
of each key library used in the system:
1. TensorFlow: TensorFlow is an open-source machine learning framework developed by
Google. It serves as the backbone for constructing and training the deep learning model used
in this project. TensorFlow provides powerful tools for building neural networks and efficiently
handling large-scale computations, making it ideal for training the LSTM-based sentiment
analysis model on the Sentiment140 dataset.
2. Keras: Keras is a high-level neural network API that runs on top of TensorFlow. It simplifies
the model building process with an easy-to-use interface, allowing developers to define,
compile, and train deep learning models quickly. Keras is used to build the LSTM architecture
in this project, manage model layers, and streamline the training and evaluation process.
3. NLTK (Natural Language Toolkit): NLTK is a leading library for natural language
processing in Python. It is used in the preprocessing phase for tasks such as tokenization, stop
word removal, stemming, lemmatization, and text normalization. These operations are crucial
for cleaning raw social media text data and making it suitable for input into the deep learning
model.
4. Scikit-learn: Scikit-learn provides various tools for data preprocessing, model selection,

CSE Department, BITS 21

 Social Media Opinion Analysis Using NLP

and evaluation. In this project, it is mainly used for calculating performance metrics such as

accuracy, precision, recall, and F1-score. It also assists in generating the confusion matrix,
which helps visualize the performance of the sentiment classifier.
5. Flask: Flask is a lightweight Python web framework used to create the web interface for the
sentiment analysis system. It allows users to input text data through a simple web form and
receive real-time sentiment predictions. Flask routes the user input to the backend, processes
the request, and displays the predicted sentiment and confidence score.
6. Pandas: Pandas is a powerful data analysis library that provides flexible data structures like
Series and DataFrames. It is used to manage and manipulate datasets, store user inputs and
sentiment results, and prepare data for training or export. Its ease of use makes it ideal for
handling structured information in the system.
7. OpenCV (Open Source Computer Vision Library): OpenCV is a library designed for real-
time computer vision tasks. In the context of this project, OpenCV can be optionally used for
processing visual elements like emojis or images that are included in social media posts.
Analyzing visual features alongside text can enhance the accuracy of sentiment detection,
especially when emojis convey emotional tone that may contradict or complement the written
text.
8. SpaCy: SpaCy is an advanced NLP library optimized for performance and production use.
While NLTK handles basic preprocessing, SpaCy is useful for more complex NLP tasks such
as named entity recognition (NER), part-of-speech tagging, and dependency parsing. It can
also be employed for more nuanced sentiment analysis scenarios where syntactic structure and
entity context are essential for understanding meaning.
By combining these libraries, the system is capable of performing end-to-end sentiment
analysis on social media text in a robust, scalable, and efficient manner. Each library
contributes to a specific part of the development pipeline, ensuring that the system remains
modular, maintainable, and easy to upgrade for future enhancements or use cases.

CSE Department, BITS 22

 Social Media Opinion Analysis Using NLP

6. SYSTEM DESIGN
System design is a critical phase that outlines how various components of the
application interact and function together. It provides a structured blueprint for implementation,
ensuring that the system meets both functional and non-functional requirements. The goal of
this phase is to create a scalable, maintainable, and efficient system capable of accurately
analyzing sentiments from social media posts.
The system is composed of multiple layers such as:
 Presentation Layer: Frontend (Flask-based web interface)
 Logic Layer: Python backend with pre-trained LSTM model for sentiment prediction
 Data Layer: Tokenizer, embedded weights, model files, and input preprocessing

6.1 UML Diagrams

A UML diagram is a diagram based on the Unified Modelling Language with the
purpose of visually representing a system along with its main actors, roles, actions, artifacts,
or classes. UML provides a modern approach to modeling and documenting software systems.
It is one of the most popular and effective business and software process modeling techniques.
Since UML is based on diagrammatic representations, it allows developers to quickly identify
design flaws or inefficiencies. In our project, UML diagrams have been used to represent the
logical structure, system behavior, and interaction flow between components involved in Social
Media Sentiment Analysis.

CSE Department, BITS 23

 Social Media Opinion Analysis Using NLP

6.1.1 Use Case Diagram

The Use Case diagram provides a high-level overview of the interactions between the
user and the system.
Actors:
 User
 System
Use Cases:
 Input a social media post
 Submit the post for sentiment analysis
 View sentiment result with confidence and emoji
 Reset or clear the input
 (Optional) View previously analyzed posts
This diagram helps us understand the functional scope of the system and the user's interaction
at different stages.

Fig 6.1.1: Use Case Diagram

CSE Department, BITS 24

 Social Media Opinion Analysis Using NLP

6.1.2 Class Diagram

The Class Diagram models the structure of the system by showing the system's classes,
attributes, methods, and the relationships among objects.
Key Classes in the Project:
 UserInput
o Attributes: postText
o Methods: validateInput()
 SentimentAnalyzer
o Attributes: model, tokenizer
o Methods: preprocessText(), predictSentiment(), getConfidenceScore()
 ResultDisplay
o Attributes: sentiment, emoji, confidence
o Methods: displayResult(), colorCodeResult()
This diagram helps in designing the system with reusable and modular components.

Fig 6.1.2: Class Diagram

CSE Department, BITS 25

 Social Media Opinion Analysis Using NLP

6.1.3 Sequence Diagram

The Sequence Diagram captures the interaction between different components when a
user inputs a post for sentiment prediction.
Flow:
1. User enters a post and clicks "Analyze".
2. The User Interface sends the post to the SentimentAnalyzer.
3. The text is preprocessed and tokenized.
4. The model is invoked to predict sentiment.
5. The SentimentAnalyzer returns the result.
6. ResultDisplay shows the sentiment label, confidence, and emoji.
7. (Optional) HistoryManager saves the post and result.
This diagram is useful for understanding the order and logic of interactions in real time.

Fig 6.1.3: Sequence Diagram

CSE Department, BITS 26

 Social Media Opinion Analysis Using NLP

6.1.4 Activity Diagram

The Activity Diagram represents the flow of control or workflow of the system in terms
of activities.
Activity Steps:
 Start
 User enters social media post
 Input is validated
 Text is preprocessed
 Sentiment is predicted using LSTM model
 Confidence score is calculated
 Sentiment label and emoji are assigned
 Output is displayed on UI
 Store result in history
This helps to visualize the decision-making paths and workflow of the entire sentiment analysis
process.

Fig 6.1.4: Activity Diagram

CSE Department, BITS 27
 Social Media Opinion Analysis Using NLP

7. IMPLEMENTATION
The implementation phase involves translating the design into actual code. For the Social
Media Opinion Analysis using NLP project, the system has been developed using Python with
the Flask web framework for the backend, and HTML/CSS for the frontend. The core of the
system is an LSTM-based deep learning model trained on the Sentiment140 dataset to classify
sentiments into Positive, Negative, or Neutral.
The implementation includes:
 Text preprocessing (cleaning, tokenizing, padding)
 Model loading and inference
 Confidence score calculation
 Emoji and sentiment display
 Web interface with input form and output section

CSE Department, BITS 28

 Social Media Opinion Analysis Using NLP

7.1 Code
The entire system has been modularized for clarity and maintainability. Key
components include:
7.1.1. app.py (Main Flask Application)
Handles routing and interaction between frontend and backend.
python
from flask import Flask, request, render_template
import pickle
import numpy as np
import os
from tensorflow.keras.models import load_model
from preprocess import clean_text, prepare_text
from datetime import datetime

app = Flask( name )

model = load_model('sentiment_analysis.keras')
with open('tokenizer.pkl', 'rb') as f:
tokenizer = pickle.load(f)

results_dir = "results"
os.makedirs(results_dir, exist_ok=True)

@app.route("/", methods=["GET"])
def home():
return render_template("index.html")

@app.route("/result", methods=["POST"])
def result():
post = request.form["post"]
cleaned = clean_text(post)
padded = prepare_text(cleaned, tokenizer)

CSE Department, BITS 29

 Social Media Opinion Analysis Using NLP

prediction = model.predict(padded)[0]
sentiment_label = np.argmax(prediction)
confidence = round(float(np.max(prediction)) * 100, 2)

if sentiment_label == 0:
sentiment = "Negative 😞"
elif sentiment_label == 1:
sentiment = "Neutral 😐"
else:
sentiment = "Positive 😊"

timestamp = datetime.now().strftime("%Y%m%d_%H%M%S")
filename = os.path.join(results_dir, f"result_{timestamp}.txt")

# Save with UTF-8 encoding to avoid emoji issues

with open(filename, "w", encoding="utf-8") as f:
f.write(f"Post: {post}\nSentiment: {sentiment}\nConfidence: {confidence}%\n")

return render_template("result.html", sentiment=sentiment, confidence=confidence,

post=post)

if name == " main ":

app.run(debug=True)

CSE Department, BITS 30

 Social Media Opinion Analysis Using NLP

7.1.2. preprocess.py
Handles clean the text, remove URLs, mentions, and non-alphabetical characters
import re
import numpy as np
from tensorflow.keras.preprocessing.sequence import pad_sequences

MAX_LEN = 50 # Maximum length of padded sequence

def clean_text(text):
# Convert text to lowercase
text = text.lower()
# Remove URLs, mentions, and non-alphabetical characters
text = re.sub(r"http\S+", "", text)
text = re.sub(r"@\w+", "", text)
text = re.sub(r"[^a-zA-Z\s]", "", text)

# Handle negation: If 'not' appears before certain words, reverse sentiment

negations = ["not", "isn't", "wasn't", "can't", "won't", "don't", "never", "no"]
words = text.split()
if any(neg in words for neg in negations):
text = " ".join(["not" if word in negations else word for word in words])

return text.strip()

def prepare_text(text, tokenizer):

# Convert the text into a sequence of integers using the tokenizer
seq = tokenizer.texts_to_sequences([text])

# Pad the sequence to ensure it matches the maximum length

padded = pad_sequences(seq, maxlen=MAX_LEN, padding='post', truncating='post')

return padded

CSE Department, BITS 31

 Social Media Opinion Analysis Using NLP

7.1.3. train.py (Model Prediction Logic)

Handles text preprocessing, model loading, and sentiment prediction.
python
import pandas as pd
import numpy as np
import pickle
from sklearn.model_selection import train_test_split
from tensorflow.keras.models import Sequential
from tensorflow.keras.layers import Embedding, Bidirectional, LSTM, Dense, Dropout
from tensorflow.keras.preprocessing.text import Tokenizer
from tensorflow.keras.preprocessing.sequence import pad_sequences

# Constants
MAX_LEN = 60
VOCAB_SIZE = 15000
EMBEDDING_DIM = 128

# Load data
df = pd.read_csv("archive/sentiment_train.csv")
df['text'] = df['text'].astype(str).str.lower()
df['label'] = df['label'].astype(int)

# Tokenization
tokenizer = Tokenizer(num_words=VOCAB_SIZE, oov_token="<OOV>")
tokenizer.fit_on_texts(df['text'])
sequences = tokenizer.texts_to_sequences(df['text'])
padded = pad_sequences(sequences, maxlen=MAX_LEN, padding="post", truncating="post")

X = padded
y = pd.get_dummies(df['label']).values

X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2)

CSE Department, BITS 32

 Social Media Opinion Analysis Using NLP

# Build model
model = Sequential([
Embedding(VOCAB_SIZE, EMBEDDING_DIM, input_length=MAX_LEN),
Bidirectional(LSTM(64, return_sequences=True)),
Dropout(0.3),
Bidirectional(LSTM(32)),
Dropout(0.3),
Dense(64, activation='relu'),
Dropout(0.3),
Dense(3, activation='softmax')
])

model.compile(loss='categorical_crossentropy', optimizer='adam', metrics=['accuracy'])

# Train model for 15 epochs without early stopping

model.fit(X_train, y_train, validation_data=(X_test, y_test),
epochs=15, batch_size=64)

# Save model and tokenizer

model.save("sentiment_analysis.keras")
with open("tokenizer.pkl", "wb") as f:
pickle.dump(tokenizer, f)

print("✅ Training done. Model and tokenizer saved.")

CSE Department, BITS 33

 Social Media Opinion Analysis Using NLP

7.1.4. templates
4.1 templates/index.html (Frontend UI)
html
<!DOCTYPE html>
<html lang="en">
<head>
<meta charset="UTF-8">
<title>Social Media Opinion Analyzer</title>
<style>
body {
font-family: Arial, sans-serif;
background: url('https://img.freepik.com/free-vector/realistic-social-media-elements-
background_79603-1521.jpg') no-repeat center center;
background-size: cover;
padding: 40px;
}
.container {
max-width: 600px;
margin: auto;
background: #fff;
padding: 30px;
border-radius: 15px;
box-shadow: 0 0 15px rgba(0,0,0,0.1);
}
textarea {
width: 100%;
height: 100px;
padding: 10px;
font-size: 16px;
border-radius: 8px;
border: 1px solid #ccc;
resize: none;
margin-bottom: 15px;
CSE Department, BITS 34
 Social Media Opinion Analysis Using NLP

}
button {
padding: 10px 25px;
background-color: #f8b500;
border: none;
color: white;
font-weight: bold;
font-size: 16px;
border-radius: 5px;
cursor: pointer;
}
button:hover {
background-color: #e0a800;
}
</style>
</head>
<body>
<div class="container">
<h2>📱 Social Media Opinion Analyzer</h2>
<form method="POST" action="{{ url_for('result') }}">
<label for="post">Enter your post or caption:</label><br>
<textarea name="post" required></textarea><br>
<button type="submit">Analyze</button>
</form>
</div>
</body>
</html>

CSE Department, BITS 35

 Social Media Opinion Analysis Using NLP

4.2 templates/result.html
<!DOCTYPE html>
<html lang="en">
<head>
<meta charset="UTF-8">
<title>Analysis Result</title>
<style>
body {
font-family: Arial, sans-serif;
background: url('https://img.freepik.com/free-vector/realistic-social-media-elements-
background_79603-1521.jpg') no-repeat center center;
background-size: cover;
padding: 40px;
}
.container {
max-width: 600px;
margin: auto;
background: #fff;
padding: 30px;
border-radius: 15px;
box-shadow: 0 0 15px rgba(0,0,0,0.1);
text-align: center;
}
.result {
font-size: 20px;
margin-top: 20px;
}
.back-button {
margin-top: 20px;
}
.back-button a {
text-decoration: none;
color: #f8b500;
CSE Department, BITS 36
 Social Media Opinion Analysis Using NLP

font-weight: bold;
font-size: 16px;
}
</style>
</head>
<body>
<div class="container">
<h2>📝 Analysis Result</h2>
<div class="result">
<p><strong>Post:</strong> {{ post }}</p>
<p><strong>Sentiment:</strong> {{ sentiment }}</p>
<p><strong>Confidence:</strong> {{ confidence }}%</p>
</div>
<div class="back-button">
<a href="{{ url_for('home') }}">🔙 Analyze Another Post</a>
</div>
</div>
</body>
</html>

CSE Department, BITS 37

 Social Media Opinion Analysis Using NLP

7.2 Screenshots
To provide a clear understanding of how the Social Media Opinion Analysis system
operates in a real-world environment, this section presents several key screenshots from the
web application. These screenshots showcase various stages of user interaction, including data
input, result generation, and error handling. Each stage highlights the user-friendly interface
and real-time sentiment analysis functionality.

Screenshot 1: Initial Input Page

The initial input page is the landing screen of the web application where users begin their
interaction with the system. It features a clean and simple layout that is intuitive even for first-
time users.

 At the center of the page, there is a text input field where users can type or paste a social
media post (such as a tweet or caption) they wish to analyze.
 Below the input field, there is a prominent “Analyze” button. When clicked, it sends
the entered text to the backend system, which performs sentiment analysis using the
trained deep learning model.

This page ensures that users can quickly and easily submit their posts without any distractions
or unnecessary steps.

Figure 7.2.1: Initial Input Page

CSE Department, BITS 38

 Social Media Opinion Analysis Using NLP

Features a "Analyze" button to process the post.

Figure 7.2.2: Analyzing a Positive Post in Progress

Screenshot 2: Sentiment Analysis Result

Once the text is submitted, the application processes the input and displays the sentiment
result on the same interface. The result page is designed to provide instant feedback and
includes the following key features:
 A sentiment label indicating whether the post is Positive, Negative, or Neutral based on
the model's prediction.
 A confidence score, typically displayed as a percentage (e.g., 91.49%), showing how
certain the model is about its prediction.
 Emoji representation or a color-coded background (green for positive, red for negative,
gray for neutral) to visually reinforce the sentiment outcome.
This result page ensures that users can quickly interpret the sentiment of their post with clarity
and ease.
Additionally, a “Analyze Another Post” button is provided, which allows users to return to the
input screen and analyze new content. This encourages repeated use and supports batch
analysis of multiple social media posts.

CSE Department, BITS 39

 Social Media Opinion Analysis Using NLP

Figure 7.2.3: Sentiment Analysis Result for a Positive Post

 After getting the result if we want to text another tweet then click on the Analyze
Another Post and then enter into initial input page and check the remaining tweets.

Figure 7.2.4: Analyzing a Neutral Post

CSE Department, BITS 40

 Social Media Opinion Analysis Using NLP

Figure 7.2.5: Sentiment Analysis Result for a Neutral Post

Figure 7.2.6: Analyzing a Negative Post

CSE Department, BITS 41

 Social Media Opinion Analysis Using NLP

Figure 7.2.7: Sentiment Analysis Result for a Negative Post

Screenshot 3: Error Handling and Edge Cases

To ensure smooth operation and a good user experience, the system includes proper error
handling mechanisms. If a user submits the form without entering any text or with content that
is too short to analyze meaningfully, the application responds with an appropriate validation
message.

 A prompt appears, clearly instructing the user to "Please fill in this field" or a similar
message that helps them understand what went wrong.
 This validation ensures that only meaningful text is sent for analysis, improving both
performance and accuracy of the results.

This error handling feature prevents system misuse and guides users to correct their input
without confusion.

CSE Department, BITS 42

 Social Media Opinion Analysis Using NLP

Figure 7.2.8: Attempt to Analyze Without Entering Text

These screenshots illustrate how the system transitions smoothly from input to analysis to result
display, with built-in safeguards for invalid input. The overall user experience is designed to
be seamless, responsive, and informative — making it suitable for both casual users and
professionals monitoring sentiment trends in real time.

Advantages
 Real-time sentiment prediction
 Trained on large social media dataset
 Accurate classification using deep learning
 Easy-to-use interface for public use
Limitations
 Can misinterpret sarcasm or complex negations
 Dependent on the dataset quality
 Not multilingual (currently supports English only)
Future Enhancements
 Include multilingual support (Hindi, Telugu, etc.)
 Add emoji-based sentiment detection
 Use transformer-based models like BERT for better accuracy
 Analyze images or voice data for sentimen
CSE Department, BITS 43
 Social Media Opinion Analysis Using NLP

8. CONCLUSION
This project, Social Media Opinion Analysis using NLP, successfully demonstrates the
use of deep learning techniques for analyzing sentiment in social media content. By leveraging
an LSTM-based neural network trained on real-world datasets like Sentiment, the system is
capable of understanding and classifying complex user expressions into emotional categories.
The intuitive web interface ensures that end-users can access the sentiment analysis tool
effortlessly. The model provides both sentiment labels and confidence scores, offering a
transparent view of prediction certainty. The solution can be extended further to analyze bulk
social media data for brand monitoring, political opinion mining, mental health tracking, and
more.
Overall, the project bridges Natural Language Processing, deep learning, and web development
to create a practical and impactful application

CSE Department, BITS 44

 Social Media Opinion Analysis Using NLP

9. REFERENCES
Below is a list of resources and references used during the development of this project:
1. Sentiment140 Dataset – Kaggle https://www.kaggle.com/kazanova/sentiment140
2. Keras Doc LSTM Layers https://keras.io/api/layers/recurrent_layers/lstm/
3. Flask Documentation https://flask.palletsprojects.com/
4. PyTorch Documentation https://pytorch.org/docs/stable/index.html
5. Goldberg, Y. (2016). A Primer on Neural Network Models for Natural Language
Processing. Journal of Artificial Intelligence Research
6. Jurafsky, D., & Martin, J. H. (2020). Speech and Language Processing (3rd Edition
Draft) https://web.stanford.edu/~jurafsky/slp3/
7. IMDbMovieReviewsDatasetKaggle
https://www.kaggle.com/datasets/lakshmi25npathi/imdb-dataset-of-50k-movie-
reviews
8. Natural Language Toolkit (NLTK) Documentation https://www.nltk.org/
9. TextBlob: Simplified Text Processing https://textblob.readthedocs.io/en/dev/
10. Scikit-learn Documentation https://scikit-learn.org/stable/documentation.html
11. Word2VecExplainedGoogleResearch
https://research.google.com/archive/word2vec.html
12. TensorFlow Documentation https://www.tensorflow.org/api_docs
13. Hugging Face Transformers Library https://huggingface.co/docs/transformers/index
14. Mikolov, T., Sutskever, I., Chen, K., Corrado, G. S., & Dean, J. (2013). Distributed
representations of words and phrases and their compositionality. Advances in Neural
Information Processing Systems.
15. Pang, B., & Lee, L. (2008). Opinion mining and sentiment analysis. Foundations and
Trends® in Information Retrieval.
16. BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding –
Devlin, J., et al. (2019). https://arxiv.org/abs/1810.04805

CSE Department, BITS 45

 International Journal of Information and Electronics Engineering, Vol. 15, No. 4, April 2025

SOCIAL MEDIA OPINION ANALYSIS

Using NLP
G. Shiva prasad1, Ch. Navya2, G. Sahithi3, S. Shivamani4, K. Rakesh5, Ramdas Vankdothu6,

P.Nagaraju7
2,3,4,5
BTech Student, Department of CSE, Balaji Institute of Technology and Science, Laknepally,
Warangal, India
1,7
Assistant Professor, Department of CSE, Balaji Institute of Technology & Science, Laknepally, Warangal,
India

ABSTRACT
Sentiment analysis on social media contributes towards public opinion perception, guiding
market strategies, political commentary, as well as buyer sentiment. Attitudes and brand loyalty
are crafted by social networking sites like Facebook and "X" (Twitter), and having sentiment
analysis on these is instrumental in tracking directions. The present study analyzes over 1000
newscast-related posts for the purpose of identifying sentiment direction. Social networking not
only facilitates communication but affects user behavior too. Sentiment analysis is vital for
organizations, policymakers, and businesses to take decisions based on facts, improve customer
engagement, and react to public matters effectively. With user-generated content expanding
exponentially, sentiment analysis has become a must to make effective use of the digital world.

The proposed system enhances existing methodologies by incorporating advanced word

embedding and transformer-based models like BERT to improve accuracy and context
understanding.

doi: 10.48047/ijiee.2025.15.4.37 396

 International Journal of Information and Electronics Engineering, Vol. 15, No. 4, April 2025

1. INTRODUCTION
In the context of contemporary socio-technical systems, social media has transformed
profoundly in the means of communication and expression of feeling and sharing information.
The social media platforms like Facebook, X (formerly Twitter), Instagram, and YouTube with
their millions of active users produce massive amounts of content that captures emotions,
attitudes, and opinions. These sentiments are crucial in shaping public opinion, brand image, and
buying behavior.

The volume of data produced every day per user, however, has accelerated to such an extent,
that it poses a challenge to individuals or businesses attempting to gain insights from it.
Sentiment Analysis, which is a subfield of Artificial Intelligence (AI) and Natural Language
Processing (NLP), solves this problem through automated extraction and sorting of feelings into
positive, negative, or neutral. It is now possible for researchers, business people, and
policymakers to effectively and systematically make sense of sentiments, trends, and public
opinion and make informed decisions.

The goal of the Sentiment Analysis for Social Media project is to create an intelligent
software system which can analyze more than 1,000 posts from social media news accounts.
Given the fact that public opinion and company branding as well as the behavior of customers
are formed through online discourse, an adequate feeling classifier is needed. So this is the
approach towards satisfying demand for powerful machine learning sentiment analysis
systems[1-25].

2. LITERATURE SURVEY
2.1 Sentiment Analysis Overview:

Sentiment analysis or opinion mining is among the key areas of research due to the social
media witnessing heightened growth. Earlier stages utilized rule-based methods but were weak at
handling slang, colloquialisms, and culture. Accuracy and flexibility increased with machine
learning and deep learning in subsequent stages.

Some research that helped strengthen sentiment analysis is:

Deep Learning Approaches (2024)

doi: 10.48047/ijiee.2025.15.4.37 397
 International Journal of Information and Electronics Engineering, Vol. 15, No. 4, April 2025

Study: Ali & Kabir (2024) compared deep learning methods for sentiment classification. Strengths:

Identified subtle sentiment patterns well.

Weaknesses: Needed lots of labeled data and high computational.

2. Big Data and Real Time Processing (2023)

Study: This systematic literature review evaluated large-scale social media sentiment analysis in
2023.

Advantages: Real-time trend identification is supported by big data.

Disadvantages: Working with unstructured data in real time constantly proves difficult.

3. Comparative Techniques (2022)

Study: Kapur & Harikrishnan compared lexicon-based, machine learning, and deep learning
techniques.

Advantages: Determines the best technique according to the data set.

Limitations: No method works for every data set.

4. Aspect-Based Sentiment Analysis (2021)

Study: Kumar & Gupta carried out research testing ABSA.

Advantages: Lasts very detailed insight through sentiment analysis of product or service features.

Limitations: Quite complicated to develop deep aspect lexicons and map the sentiments accurately.

3. PROBLEM STATEMENT

Massive amounts of user-generated content are created every day, thanks to the rapid growth
of social media. Understanding public sentiment becomes critical for businesses, organizations,
and policymakers. In fact, it is a tough task as analyzing sentiments of social media posts gets
complicated by informal language, slangs, abbreviations, and speaker sarcasms along with
cultural quirks. Active rule-based methods fail to attend to these complexities, and on the other
hand, machine learning methods require extensive labeled datasets.
doi: 10.48047/ijiee.2025.15.4.37 398
 International Journal of Information and Electronics Engineering, Vol. 15, No. 4, April 2025

This specific study intends to develop a fast and accurate system of Sentiment Analysis
specific to Social Media that will interpret better what users feel. It builds upon Natural
Language Processing (NLP) and Machine Learning (ML) to detect emotional states considering
context variation to deliver greater insight for decision making. The solution will facilitate brand
perception monitoring, customer engagement in businesses, and improve marketing without
compromising ambiguity and scalability challenges in sentiment data.

4. EXISTING SYSTEM
The existing sentiment analysis systems on social media depend on:

1. Lexicon-Based Approaches: It ignores the slang and context in and even while you use
predefined sentiment dictionaries.These can be often lacking deep contextual knowledge,
machine learning models include SVM, naïve bayes, and random forest.

2. By using the LSTMs and CNNs but find a lot of difficulties in sarcasm and long-range
dependencies challenging.

3. Even using the basic Deep Learning Models which are giving the incorrect results.

Limitations of the Current System:

Drawbacks including bad performance on big data and not able to include more thorough
background in sophisticated sentences causes a strong reliance on feature engineering.

5. METHODOLOGY
By considering the drawbacks of the project we propose a sentiment analysis model
which is used to employing transformer-based architectures, especially here by dealing with
BERT and its variants, we can overcome these restrictions. There are some suggested
arrangementlikeContextual word representation with pre-trained embeddings (BERT, RoBERTa)
Use fine-tuning on datasets particular to the domain which is used to improve performance. It
can be possible when using the attention mechanisms to grasp word dependencies.

In the project methodology there exists of a various phases which are as follows:

1. Data Collection:To extract tweets, Facebook comments, and other social media posts using

doi: 10.48047/ijiee.2025.15.4.37 399

 International Journal of Information and Electronics Engineering, Vol. 15, No. 4, April 2025

APIs this is a crucial step.

2. Preprocessing: After the extraction of data remove stopwords, punctuation, emojis, and
perform stemming/lemmatization these steps are mainly used.

3. Feature Extraction:Then convert the text into numerical vectors by using TF-IDF,
Word2Vec, and BERT embeddings.

4. Model Training: Train the model which is from the deep learning includes (LSTM, BERT,
RoBERTa) on labeled sentiment datasets.

5. Evaluation:Calculate the accuracy from the training phase, precision, recall, and F1-score.

6. Deployment: Develop a Flask-based web application for real-time sentiment prediction.

Fig 1: Basic steps of opinion mining on social network platforms.

6. FEATURE EXTRACTION
1. TF-IDF (Term Frequency - Inverse Document Frequency): It represents the text data
numerically.

doi: 10.48047/ijiee.2025.15.4.37 400

 International Journal of Information and Electronics Engineering, Vol. 15, No. 4, April 2025

2. Word2Vec: It captures the semantic meaning of words from the sentences.

3. BERT Embeddings: These provides a context-aware of word representation.

MODEL DEVELOPMENT

We investigate three varying sentiment classification models within this project, each having
its own pros and cons:

1. Long Short-Term Memory (LSTM)

LSTM is efficient in handling sequential dependencies in text, which is helpful in

sentiment analysis. It fails to do so with longer texts, which happen in big data sets, thus
subjecting the contextual meaning to possible loss.

2. Convolutional Neural Networks (CNN)

CNN is effective in feature extraction from text, especially n-gram features, which can be
used to identify sentiment-oriented words and phrases. Yet CNN is missing a sense of sequential
relations between words, which will restrict its performance for sentiment classification.

3. Bidirectional Encoder Representations from Transformers (BERT)

BERT is a cutting-edge deep learning model that learns deep contextualized representations
of text. BERT is different from other models in that it captures the meaning of words in relation
to the context surrounding them, resulting in extremely accurate sentiment predictions.

Training Procedure

To make sure the models are running at their best, we adhere to a systematic training
process:

Dataset: The models are trained with well-established sentiment datasets such as IMDB reviews,
custom-labeled datasets, and Twitter Sentiment140.

Optimizer: We apply the Adam optimizer with a learning rate of 0.0001 to make learning
efficient and update weights properly.

Loss Function: We use the categorical cross-entropy loss function to compute the
discrepancy between predicted and actual sentiment labels.
doi: 10.48047/ijiee.2025.15.4.37 401
 International Journal of Information and Electronics Engineering, Vol. 15, No. 4, April 2025

Hyperparameter Tuning: We tune parameters like batch size, number of epochs for training,
and dropout rate to enhance the performance of the model and avoid overfitting.

Using these methods, we hope to build a sentiment analysis system that accurately
classifies posts on social media while overcoming the shortcomings of current models.

7. PROJECT REQUIREMENTS
Hardware: GPU system (NVIDIA RTX 3060 or comparable) Software:

Python, TensorFlow/PyTorch, Flask, Jupyter Notebook Libraries: Pandas,

NumPy, Scikit-learn, Transformers, NLTK, Flask

8. IMPLEMENTATION
The project is with support of Python with deep learning libraries In order to optimize training,
the models were trained on this system.
Steps to be followed in the process of Implementation:
1. Data Preprocessing
2. Feature Extraction
3. Model Training and Evaluation
4.Result
1. Data Preprocessing
Preprocessing of text data were taken out before training themodel with the following:
Removing stopwords, punctuation, and emojis
Words like the, is, and were were removed because they don't add any sentiment.Punctuation and
emojis were removed to only keep words that have meaning.1.1.Tokenization
Text was separated into individual words or subwords for easy processing. In
BERT, WordPiece tokenizer was utilized for unknown words.
1.2.Stemming and Lemmatization
Stemming converted words to base form (e.g., running → run).
Lemmatization changed words to dictionary form (e.g., better → good).
Example:
Before preprocessing:

doi: 10.48047/ijiee.2025.15.4.37 402

 International Journal of Information and Electronics Engineering, Vol. 15, No. 4, April 2025

"This film was incredible!!! ???????? But the conclusion was disappointing." After
preprocessing:
"movie incredible but end disappoint"
2. Feature Extraction
To transform text into a form that can be processed by the model, various feature extraction
methods were employed:
2.1. TF-IDF (Term Frequency-Inverse Document Frequency): Weighted words according to their
significance.
2.2. Word2Vec: Trapped word relationships by transforming them into numerical
vectors.2.3.BERT Embeddings: Offered more profound comprehension by examining words
accordingto the context surrounding them.
Example of Contextual Understanding:
"I am heading to the bank." → (Bank as a financial institution) "I
am sitting on the river bank." → (Bank as a river shore)
3. Model Training and Evaluation
The below dataset is taken into consideration as follows as The
models were trained and tested on three datasets:
1. IMDB Movie Reviews (general sentiment classification)
2. Twitter Sentiment140 (social media sentiment analysis)
3. Custom-labeled datasets (for specific domains)
3.1. Training Process
The followingdataset classified is into 80% training and 20% testing.
The models were trained with Adam optimizer and a learning rate of 0.0001. The
loss function utilized was Categorical Cross-Entropy.
The models were trained from 10 to 15 epochs, with the hyperparameters being optimizedfor
enhanced performance.

4. Result
After training the models, their performance was evaluated based on accuracy, precision, recall,
and F1-score. The results are summarized in the table below:

doi: 10.48047/ijiee.2025.15.4.37 403

 International Journal of Information and Electronics Engineering, Vol. 15, No. 4, April 2025

Model Accuracy (%) Precision Recall F1-Score

LSTM 82.5 80.2 79.8 80.0

CNN 78.3 76.5 74.2 75.3

BERT 92.1 91.8 91.5 91.6

9. FUTURE SCOPE
The future application scope of sentiment analysis and NLP reaches across some of the main
areas of future innovation and advancement. One exciting area is the scaling of the model across
different languages via multilingual BERT, whereby the sentiment analysis models can
recognize and process text across various languages at high levels of accuracy, allowing them to
better fit in foreign markets. Another thrilling breakthrough is enhancing sarcasm detection by
multimodal learning, blending text and image analysis to better recognize sarcasm. This comes
in handy during sentiment analysis in social media, as sarcasm distorts the textual meaning. The
creation of an in-real-time dashboard for the analysis of sentiment trends also can give
businesspersons and researchers timely insights into people's opinion so that they may respond
appropriately to the growing trends. In addition, the incorporation of business intelligence
platforms for market analysis provides enriched data-driven decision-making through the
correlation of sentiment analysis with financial and customer information. These technologies
will significantly enhance the efficiency and precision of sentiment analysis applications in many
industries.

10. CONCLUSION
This work showcases a state-of-the-art sentiment analysis technique utilizing deep learning
and transformer models. By overcoming the drawback of conventional models, our system with
the new design has high precision and stability in opinion classification on social media.
Sentiment analysis will be further improved by future AI and NLP developments to support
improved decision-making in various industries.
doi: 10.48047/ijiee.2025.15.4.37 404
 International Journal of Information and Electronics Engineering, Vol. 15, No. 4, April 2025

11. REFERENCES
1. Devlin, J., Chang, M., Lee, K., & Toutanova, K. (2019). BERT: Pre-training of Deep Bidirectional
Transformers for Language Understanding.
2. Pang, B., & Lee, L. (2008). Opinion Mining and Sentiment Analysis.
3. Vaswani, A., Shazeer, N., Parmar, N., et al. (2017). Attention Is All You Need.
4. Liu, B. (2012). Sentiment Analysis and Opinion Mining.
5. Sentiment Analysis Symposium 2011, New York, 12/04/2011.
6. Pang, B., Lee, L.: “Opinion Mining and Sentiment Analysis”, in “Foundations and Trends in
Information.
7. Pang, B., Lee, L., Vaithyanathan, S.: “Thumbs up? sentiment classification using machine learning
techniques”, in Proceedings of the ACL-02 conference on Empirical methods in natural language
processing, Volume 10, July 2002, pp. 79-86.
8. Socher, R., et al.: “Semi-supervised recursive autoencoders for predicting sentiment distributions” in
Proceedings of EMNLP '11 - the Conference on Empirical Methods in Natural Language Processing,
ISBN: 978-1-937284-11-4, pp. 151-161
9. Ramdas Vankdothu,Dr.Mohd Abdul Hameed, Husnah Fatima” A Brain Tumor Identification and
Classification Using Deep Learning based on CNN-LSTM Method” Computers and Electrical
Engineering , 101 (2022) 107960
10. Ramdas Vankdothu,.Mohd Abdul Hameed “Adaptive features selection and EDNN based brain image
recognition on the internet of medical things”, Computers and Electrical Engineering , 103 (2022)
108338.
11. Ramdas Vankdothu,.Mohd Abdul Hameed,Ayesha Ameen,Raheem,Unnisa “ Brain image
identification and classification on Internet of Medical Things in healthcare system using support value
based deep neural network” Computers and Electrical Engineering,102(2022) 108196.
12. Ramdas Vankdothu,.Mohd Abdul Hameed” Brain tumor segmentation of MR images using SVM and
fuzzy classifier in machine learning” Measurement: Sensors Journal,Volume 24, 2022, 100440
.
13. Ramdas Vankdothu,.Mohd Abdul Hameed” Brain tumor MRI images identification and classification
based on the recurrent convolutional neural network” Measurement: Sensors Journal,Volume 24, 2022,
100412 .
14. Bhukya Madhu, M.Venu Gopala Chari, Ramdas Vankdothu,.Arun Kumar Silivery,Veerender
Aerranagula ” Intrusion detection models for IOT networks via deep learning approaches ”
Measurement: Sensors Journal,Volume 25, 2022, 100641

15. Mohd Thousif Ahemad ,Mohd Abdul Hameed, Ramdas Vankdothu” COVID-19 detection and
classification for machine learning methods using human genomic data” Measurement: Sensors
Journal,Volume 24, 2022, 100537
doi: 10.48047/ijiee.2025.15.4.37 405
 International Journal of Information and Electronics Engineering, Vol. 15, No. 4, April 2025

16. S. Rakesh a, NagaratnaP. Hegde b, M. VenuGopalachari c, D. Jayaram c, Bhukya Madhu d, MohdAb dul
Hameed a, Ramdas Vankdothu e, L.K. Suresh Kumar “Moving object detection using modified GMM
based background subtraction” Measurement: Sensors ,Journal,Volume 30, 2023, 100898

17. Ramdas Vankdothu,Dr.Mohd Abdul Hameed, Husnah Fatima “Efficient Detectionof Brain Tumor
Using Unsupervised Modified Deep Belief Network in Big Data” Journal of Adv Research in
Dynamical & Control Systems, Vol. 12, 2020.

18. Ramdas Vankdothu,Dr.Mohd Abdul Hameed, Husnah Fatima “Internet of Medical Things of Brain
Image Recognition Algorithm and High Performance Computing by Convolutional Neural
Network” International Journal of Advanced Science and Technology, Vol. 29, No. 6, (2020), pp.
2875 – 2881
19. Ramdas Vankdothu,Dr.Mohd Abdul Hameed, Husnah Fatima “Convolutional Neural Network-
Based Brain Image Recognition Algorithm And High-Performance Computing”, Journal Of Critical
Reviews,Vol 7, Issue 08, 2020(Scopus Indexed)
20. Ramdas Vankdothu, Dr.Mohd Abdul Hameed “A Security Applicable with Deep Learning
Algorithm for Big Data Analysis”,Test Engineering & Management Journal,January-February 2020
21. Ramdas Vankdothu, G. Shyama Chandra Prasad “ A Study on Privacy Applicable Deep Learning
Schemes for Big Data” Complexity International Journal, Volume 23, Issue 2, July- August 2019
22. Ramdas Vankdothu, Dr.Mohd Abdul Hameed, Husnah Fatima “ Brain Image Recognition using
Internet of Medical Things based Support Value based Adaptive Deep Neural Network” The
International journal of analytical and experimental modal analysis, Volume XII, Issue IV,
April/2020
23. Ramdas Vankdothu,Dr.Mohd Abdul Hameed, Husnah Fatima” Adaptive Features Selection and
EDNN based Brain Image Recognition In Internet Of Medical Things “ Journal of Engineering
Sciences, Vol 11,Issue 4 , April/ 2020(UGC Care Journal)
24. Ramdas Vankdothu, Dr.Mohd Abdul Hameed “ Implementation of a Privacy based Deep Learning
Algorithm for Big Data Analytics”, Complexity International Journal , Volume 24, Issue 01, Jan
2020
25. Ramdas Vankdothu, G. Shyama Chandra Prasad” A Survey On Big Data Analytics: Challenges, Open
Research Issues and Tools” International Journal For Innovative Engineering and Management
Research,Vol 08 Issue08, Aug 2019

doi: 10.48047/ijiee.2025.15.4.37 406

 International Journal of Information and Electronics Engineering, Vol. 15, No. 4, April 2025

BIBLIOGRAPHY

I am Navya Chanda from the Department of Computer Science

and Engineering. Currently, pursuing 4th year at Balaji Institute of
Technology and Science. My research is done based on “Social
Media Opinion Analysis Using NLP”.

I am Sahithi Gudimalla from the Department of Computer

Science and Engineering. Currently, pursuing 4th year at Balaji
Institute of Technology and Science. My research is done based
on “Social Media Opinion Analysis Using NLP”.

I am Shivamani Sirimalle from the Department of Computer

Science and Engineering. Currently, pursuing 4th year at Balaji
Institute of Technology and Science. My research is done based
on “Social Media Opinion Analysis Using NLP”.

I am Rakesh Kannam from the Department of Computer Science

and Engineering. Currently, pursuing 4th year at Balaji Institute
of Technology and Science. My research is done based on
“Social Media Opinion Analysis Using NLP”.

doi: 10.48047/ijiee.2025.15.4.37 407