Sentilytics: Chrome Extension for Real-Time Sentiment

Analysis and AI-Driven Insights from Social Media

Comments
Sonu Kumar1, Moulik Khatri2, Kapil Kumar3
deepakkumar23887@gmail.com1, moulikkhatri14@gmail.com2, kapilbansal204@gmail.com3
Department of Computer Science & Engineering
Sharda University, Greater Noida, IN, 201310

ABSTRACT summary of sentiment patterns [4]. Sentilytics uses

Llama3 (12b) to extract important themes from the
Essential for engagement and content strategy, online comments, so identifying reoccurring user complaints
sites lack efficient tools for summarizing and and creating actionable solutions [5] so improving
analyzing user sentiment in comments. This work usability. At last, these realizations and visual analytics
presents a Chrome add-on with sentiment analysis, are assembled into a coherent report for simple user
real-time comment scraping, and AI-driven insights understanding.
automated. Visualizing results using several graphs
and word clouds, it labels comments as good, bad, or
neutral using NLP approach and DistilBERT. The The design, technique, and application of Sentilytics
system uses Llama3 (12b) for practical advice and are investigated in this work together with their
underlines important concerns and recommendations benefits, constraints, and possible developments.
by users. This paper investigates the architecture, Important difficulties include ethical issues in data
advantages, difficulties, and developments of the collecting, sentiment misclassification resulting from
extension to show its possibility to improve social linguistic subtleties, and API limits [6]. The study
media sentiment analysis and promote data-driven shows how combining NLP, artificial intelligence-
decision-making. driven insights, and data visualization into a single
application can improve user sentiment, therefore
Keywords enabling content authors to better target their efforts
depending on audience comments.
Sentiment Analysis, Social Media Analytics, AI-
Driven Insights, Chrome Extension, Real-Time
Analysis

INTRODUCTION
Social media platforms' explosive expansion has
produced an enormous amount of user-generated
material, including comments, reviews, and debates.
These remarks offer insightful analysis of audience
perspective, attitudes, and expectations. Content
providers, companies, and organizations all depend on
an awareness of user sentiment if they are to properly
handle user problems, enhance engagement, and
optimize content strategies. Manual examination of
hundreds of comments is not feasible, though, and
most current sentiment analysis systems lack real-time
processing and useful insights [1]. This work presents Figure 1: Overview of the Sentilytics Analysis Workflow
Sentilytics, a Chrome add-on automating comment
scraping, sentiment categorization, visualization, and
AI-driven insights, to close this gap.
LITERATURE REVIEW
Deep learning and NLP have advanced sentiment
analysis and allowed insights from social media
The Sentilytics add-on guarantees flawless data comments. Existing studies, books, papers, or
processing and insight development by means of a documentation on sentiment classification, data
disciplined workflow. As shown in Figure 1, the collecting, visualization, and AI-driven insights are
procedure starts with structured CSV file storage of compiled in this part. Finding important developments
scraping comments made using the YouTube API [1]. and gaps, we show how Sentilytics combines
DistilBERT then sorts the gathered data such that sentiment analysis, real-time scraping, and actionable
every comment is assigned a positive, negative, or insights into one platform.
neutral [3] classification. The system creates visual
representations such as histograms, word clouds, and
pie charts following categorization, therefore
offering an easy
Users of social media now mostly express their ideas This section outlines the methodology of Sentilytics,
on this platform; hence sentiment analysis is a very covering the system architecture, real-time comment
important instrument for public opinion understanding. scraping, sentiment analysis, AI-driven insights,
B. Liu (2022) in Sentiment Analysis: Mining and report generation. The workflow ensures
Opinions, Sentiments, and Emotions [1] emphasizes seamless extraction, processing, and visualization of
the need for obtaining sentiments from user-generated user comments for sentiment-based decision-making.
information. While recent developments in deep
learning, especially transformer-based models, have 3.1. System Architecture
greatly increased accuracy and contextual
understanding, traditional sentiment analysis Sentilytics uses a disciplined approach for sentiment
techniques depended on lexicon- based methodologies analysis, as Figure 1 illustrates. Extensive comments
and machine learning models. extracted via the Chrome Extension Interface from
platform APIs (YouTube, Instagram, X) are stored as a
Deep learning models—especially transformer-based
CSV file. Eliminating stopwords, links, and special
architectures—have shown recent success in sentiment
characters, the preprocessing module cleans the text.
classification applications. By means of a comparative
investigation of BERT, DistilBERT, and RoBERTa, While Matplotlib and Seaborn create sentiment
Joshy and Sundar (2024) [4] discovered that visualizations like count plots, histograms, word
DistilBERT achieves equivalent accuracy with low clouds, and pie charts, DistilBERT labels comments as
computing cost. Jethani (2024) also investigated the either "positive," "negative," or "neutral." Extensive
use of DistilBERT for emotional recognition, stressing themes are extracted from the AI Insights Module
its effectiveness in handling vast textual data volumes. (Llama3) together with practical advice. Results are
shown at last via a Flask web interface for user
Sentiment analysis's application in social media has
involvement.
been investigated in several sectors of business use.
Sprout Social (2024) [6] explores how companies use
sentiment data to improve client involvement.
Likewise, Venngage [7] offers methods for visualizing
sentiment data, stressing the need of graphical
representation in obtaining practical understanding
from big data.

Real-time sentiment analysis depends on comment

scraping being automated. While web scraping tools
such as BeautifulSoup help data extraction from other Figure 2: Sentilytics System Architecture
sites, the Google API Client Library (2024) [2] offers
capabilities to extract comments from YouTube. 3.2. Real-Time Comment Scraping
Effective data collection made possible by these
instruments provides the basis for sentiment analysis in Sentilytics collects comments using the official APIs
social media systems. of YouTube, Instagram, and X, ensuring compliance
with platform policies. The process follows these
Good visualization helps sentiment analysis results to steps:
be interpretable. Histograms, word clouds, and pie
charts [8] can be produced by using programs  Authenticate using API keys and retrieve
including Matplotlib and Seaborn. The 2024 Flask [8] comments from the active post or video.
documentation shows how sentiment reports might be  Store extracted comments in a CSV file
dynamically displayed using web frameworks, hence along with metadata (user ID, timestamp,
enabling end-user access to insights. likes).

Recent developments in AI-driven insights have made

However, certain challenges and ethical considerations
sentiment trends possible for automated response
arise in this process. API rate limits may restrict large-
production. The 2024 Hugging Face documentation [5]
scale data extraction, requiring optimization strategies
illustrates how well fine-tuned transformer models
for efficient data retrieval. Additionally, privacy
perform in text classification problems. By including
concerns necessitate the anonymization of user data to
such models in programs like Sentlytics, content
ensure compliance with data protection regulations and
creators can get real-time feedback that lets them make
maintain user confidentiality.
data-driven changes to their content plans.

Research already in publication emphasizes the value 3.3. Sentiment Analysis Using
of deep learning in sentiment analysis, the success of
automated comment scraping, and the part data DistilBERT
visualization plays in enabling insights. For content
Sentilytics uses a lightweight transformer model
creators, there is a gap in combining these elements
tailored for sentiment classification, DistilBERT, once
into a single, user-friendly platform, though.
the comments are gathered. Preprocessing starts the
Combining real-time scraping, sentiment
process; NLTK and regex help to remove special
categorization, and AI- driven insights into a complete
characters, links, and stopwords from the text, thus
sentiment analysis platform, Sentilytics seeks to close
preparing it for analysis.
this gap.
DistilBERT sorts the comments into three sentiment
3. Methodology categories—positive, negative, and neutral—after
preprocessing. Various visualizations like count plots
and histograms to indicate the distribution of sentiment  Countplot & Histogram – Shows positive,
categories, pie charts to show the proportion of negative, and neutral comment distribution.
positive, negative, and neutral comments, and word
clouds stressing often used phrases in the comments
are produced to communicate the data.

3.4. Generating Actionable Insights

Beyond sentiment classification, Sentilytics uses
Llama3 (12b) to derive deeper understanding from
user comments and create ideas for content
enhancement. Together with a boilerplate template
prompt, Llama3 (12b) handles the comments to offer
AI-generated recommendations. While for positive Figure 3: Sample Countplot and Histogram
sentiment it points up important content strengths that
might be strengthened, for negative sentiment it
emphasizes common problems and suggests
corrections.

3.5. Report Generation and User

Interaction
The final step compiles analyzed data into a structured
report, presented through an interactive Flask-based
interface.

Features of the Report:

Figure 5: Sample Pie Chart
 Summary of positive, negative, and neutral
comments.  Pie Chart – Shows for a given post or video
the percentage of every sentiment type.
 Count plots, pie charts, and word clouds.

 Extracted themes and recommended actions.

 Users can download the report for further

analysis.

3.7. Conclusion
Sentilytics generates insights using a disciplined
approach for real-time sentiment analysis. It starts with Figure 4: Sample Word Cloud Graph
comment extraction using platform APIs and then uses
DistilBERT to preprocess text and classify sentiment.  Word Cloud – It highlights often used
Llama3 (12b) helps to clarify insights by pointing out words while stressing shared themes in user
recurrent themes and recommendations for comments.
development. An interactive Flask interface presents the
The sentiment distribution graphs (Figures 3 and 4)
visual results. As shown in Figure 2: Sentilytics
expose general audience reception. While a lot of
System Architecture, this flawless procedure helps
negative comments point to certain problems that
content authors to make data-driven decisions
require attention, a bigger ratio of positive remarks
depending on sentiment patterns and practical insights.
implies great involvement. The word cloud (Figure 5)
enables content authors to discover audience
4. Results and Discussion expectations and criticism, therefore helping to
The examined sentiment data is presented in this part pinpoint often discussed subjects.
together with discussions of sentiment distribution
patterns, interpretation of important visualizations, and 4.2. Case Studies and Use Cases
comparison of Sentilytics' performance with other
To demonstrate Sentilytics real-world impact, we
sentiment analysis techniques.
present example use cases:
4.1. Visualization and Sentiment  Case Study 1: Creator of YouTube content
Distribution finds consistent unfavourable comments on
audio quality. They increase production
Multiple graphical approaches help to clearly show settings by means of AI-driven insights,
the processed sentiment data, hence guiding user thereby increasing engagement.
thoughts.  Case Study 2: Brand Monitoring on
Instagram — A brand uses Sentilytics to
react proactively, therefore enhancing
consumer
 confidence when it finds customer discontent sentiment classification, artificial intelligence-driven
with delivery times. insights utilizing Llama3, and visual analytics. The
results show its efficiency in spotting sentiment trends,
4.3. Why DistilBERT? displaying comments, and creating actionable
recommendations, thereby supporting data-driven
Driven by DistilBERT, Sentilytics is compared with
decision-making.
conventional sentiment analysis models:
Approaches Accuracy
Processing Multilingual Sarcasm
Sentlytics has difficulties including API rate constraints,
Speed Support Handling language complexity in sentiment classification, and
DistilBERT High Fast Yes Good possible AI biases in insights generating, even with its
VADER Moderate
Fast Limited Poor
success [2][6].
Traditional ML Moderate
Slow Limited Weak
(SVM,
etc.) Improvements ahead include

 Increasing accessibility by means of several

Faster and more lightweight than BERT, DistilBERT languages.
is selected for speed and efficiency; hence, it is perfect  Include e-commerce sites like Amazon,
for real-time analysis. Its rich contextual learning helps Flipkart, etc., as well as other social media
it also to properly manage sarcasm and slang. sites like Reddit and Facebook.
DistilBERT also supports multilingual text so that  Optimizing Llama3 (12b) for more context-
sentiment classification across many languages is aware recommendations can help to improve
accurate. AI-driven insights.
 Investigating other scraping techniques
4.4. Why Llama3? outside of API-based extraction to help to
reduce rate restrictions.
Llama3 is used to extract recurrent themes, user problems,
and practical advice to improve analysis outside of Beyond content creation, Sentilytics offers a potent tool for
sentiment categorization. real-time audience perspective comprehension in
marketing analytics, customer sentiment tracking, and
 Advanced Context Understanding: Llama3 brand reputation management.
searches vast comment sets for hidden user
issues and subtle themes. 6. REFERENCES
 AI-Generated Suggestions: Based on
sentiment patterns, offers content authors 1. Liu, Bing. “Sentiment analysis and opinion
practical advice. mining”. Springer Nature, 2022.
 Scalability guarantees real-time artificial 2. Google Developers. Google API Client
intelligence by effectively managing Library for Python. 2024.
enormous volumes of text data. 3. A. Joshy and S. Sundar, “Analyzing the
Performance of Sentiment Analysis using
Llama3 gives better contextual awareness, efficiency, and
BERT, DistilBERT, and RoBERTa”, 2022
real-time processing than FLAN-T5-Large, Gemma,
IEEE International Power and Renewable
and others.
Energy Conference (IPRECON), Kollam,
For instance, Llama3 can recommend shorter, more India, 2022, pp. 1-6, doi:
succinct content forms to increase interaction if 10.1109/IPRECON55716.2022.10059542.
Sentilytics notes recurrent complaints about video 4. Jethani, Aditya. “Decoding Emotions:
length. Sentiment Analysis with DistilBERT”
Medium, 2024.
4.5 Conclusion 5. Hugging Face. DistilBERT: A Smaller,
Faster, Cheaper Version of BERT. Hugging
The findings show how well Sentilytics analyzes Face, 2024.
sentiment distribution, charts patterns, and offers 6. Sprout Social. “Social Media Sentiment
useful insights. While case studies show practical uses, Analysis: Benefits and Guide for 2025”.
graphical depictions, including count graphs, pie Sprout Social, 2024.
charts, and word clouds, assist to analyzing audience 7. Venngage. “Social Media Sentiment
comments. Comparisons reveal DistilBERT's better Analysis Infographic Template”. Venngage,
sentiment categorization performance as well as 2024.
Llama3's capacity for deeper insights. These 8. Flask Documentation. Flask
technologies used together help content makers to Documentation (Stable Version). Flask Web
make informed, data-driven decisions for maximum Framework, 2024.
audience pleasure and engagement. 9. "Top Sentiment Analysis Tools & Software
in 2024." Brandwatch, 13 Feb. 2024.
5. Conclusion and Future Work Accessed 1 Apr. 2025.
10. DS-AI with Khanh Vy, “How to Use Google
This work presented Sentlytics, a Chrome add-on meant for Gemma 3.0 Multimodal Models FREE - 3
real-time social media remark sentiment analysis. WAYS with Ollama, HuggingFace, Google
Sentlytics offers content creators and companies a Studio,” YouTube, Mar. 23, 2025.
complete and easily available solution by combining
automatic comment scraping, DistilBERT-based
11. L. von Werra, "Sentiment Analysis with
Python," Hugging Face Blog, Jun. 21, 2022.
12. Google, "Extensions documentation,"
Chrome Developers, [Online].