YouTube Text Sentiment Analysis

Abstract

An innovative approach to YouTube content analytics by combining sentiment analysis of viewer comments with video performance metrics. Our methodology integrates natural language processing techniques and machine learning algorithms to classify comments as positive, negative, neutral, relevant, and irrelevant, while also extracting key insights from video transcripts. By synthesizing video metrics such as views, likes, and engagement with sentiment intelligence, the system delivers comprehensive understanding of audience behavior. The results, visualized through interactive dashboards, empower content creators to refine their strategies for maximum impact. Our implementation achieved 90% accuracy in sentiment classification, demonstrating the effectiveness of our approach for content evaluation and strategic decision-making.Through the fusion of quantitative metrics and qualitative sentiment analysis, our system enables creators to identify content themes that resonate with audiences, recognize patterns in viewer engagement, and optimize content strategy accordingly. The interactive visualization tools facilitate intuitive exploration of complex relationships between sentiment distribution and performance indicators. This holistic analytical framework provides actionable insights that transcend traditional analytics, allowing for more nuanced understanding of audience preferences and behavior. Our approach has significant implications for content optimization and audience development strategies in the growing digital content ecosystem.

Introduction

Overview:
YouTube, with over 2 billion monthly users, is a central platform for digital content. However, its native analytics tools are limited in assessing audience sentiment and content quality. Content creators struggle to interpret raw data like comments and engagement metrics, while users face inefficiencies in gauging video relevance manually. This research proposes a data-driven YouTube Video Analyzer that enhances content evaluation through sentiment analysis, relevance classification, and transcript summarization.

Key Objectives and Contributions:

1. Improve content quality assessment by integrating advanced analytics into YouTube data.

2. Automate sentiment analysis of viewer comments using Natural Language Processing (NLP) and machine learning.

3. Generate extractive summaries of video transcripts to allow users to grasp core content without watching the full video.

4. Visualize insights via a user-friendly web interface showing sentiment trends, comment relevance, and summaries.

Related Research:

• Previous studies have applied basic NLP for YouTube sentiment analysis, classifying comments as positive, neutral, or negative.

• Tools like TubeBuddy and VidIQ assist in metadata optimization but lack deep sentiment and summarization features.

• Research by Khan et al. and Baravkar et al. demonstrated the potential of comment sentiment analysis, while Deepali and Mahender explored methods for extractive and abstractive text summarization.

Proposed Model Features:

1. Architecture:

   • A three-tier system: Data collection (YouTube API), processing (sentiment, relevance, summarization), and visualization (Flask-based interface).

2. Data Collection:

   • Extracts comments using the YouTube Data API and transcripts via youtube-transcript-api.

   • Filters out irrelevant metadata and formats text for analysis.

3. Text Preprocessing:

   • Includes lowercasing, tokenization, lemmatization, and removal of noise (special characters, timestamps).

4. Sentiment Analysis:

   • Uses TextBlob to assign polarity scores.

   • Sentiment classified as positive, neutral, or negative.

   • Extreme sentiment scores are also used to judge comment relevance.

5. Machine Learning Models:

   • Logistic Regression for sentiment classification (positive, neutral, negative).

   • Gaussian Naive Bayes for relevance classification (relevant, irrelevant).

   • Achieved 70–90% accuracy in testing.

6. Transcript Summarization:

   • Extracts key sentences based on word frequency scoring to create concise, high-information summaries (~70% content reduction).

7. Web Interface:

   • Built with Flask.

   • Users input a YouTube video URL and receive:

     • Sentiment breakdown (pie/bar charts).

     • Relevance classification of comments.

     • Transcript summary in a scrollable box.

Simulation & Results:

1. Sentiment Classification Accuracy:

   • 90% overall accuracy.

   • High precision for positive (1.0) and neutral (0.83) sentiments.

   • Poor performance on negative comments (F1-score: 0.00) due to low data volume or imbalance.

2. Relevance Classification Accuracy:

   • 70% overall accuracy.

   • Good at predicting relevant comments (F1-score: 0.80).

   • Weak at identifying irrelevant comments (F1-score: 0.40), needing model improvement.

3. User Interface & Visualization:

   • Presents three major outputs:

     • Sentiment pie charts/bar graphs (positive, neutral, negative).

     • Relevance graphs (relevant vs. irrelevant comments).

     • Transcript summary for quick content preview.

   • Enhances decision-making by reducing users' time spent evaluating video quality.

Conclusion

This research successfully combines sentiment analysis of comments with performance metrics and transcript summarization to create a powerful tool for YouTube content evaluation. By delivering actionable insights through an interactive dashboard, the system empowers both creators and viewers: For Creators: The tool provides valuable insights into audience sentiment and engagement, helping to refine content strategies and improve performance. For Viewers: The system enables quick assessment of video quality and relevance before investing time in watching the full content. Our approach demonstrates the effectiveness of integrating NLP-based sentiment analysis with performance metrics for comprehensive YouTube content evaluation. The system can help users make informed decisions about content consumption, thereby saving time and enhancing the overall YouTube experience. Future enhancements include multilingual analysis by integrating translation services and training models for diverse languages. Real-time analytics will provide instant insights, with push notifications for sudden view or sentiment changes. Competitor analysis will track engagement and sentiment trends, offering strategic content optimization. Audio-based assessment will analyze videos lacking transcripts or comments using speech-to-text and tone analysis. Abstractive summarization will enhance summaries with human-like phrasing via advanced NLP models. Lastly, cross-platform integration will aggregate performance insights from social media platforms like Instagram and Twitter, providing a comprehensive view of audience behavior and content impact.

References

[1] Mohammed Arsalan Khan, Sumit Baraskar, AnshulGarg, Shineyu Khanna, Asha M. Pawar“YouTube Comment Analyzer”, International Journal of Scientific Research in Computer Science and Engineering, August 2021. [2] Aditya Baravkar, Rishabh Jaiswal, Jayesh Chhoriya,“Sentimental Analysis of YouTube Videos”, International Research Journal of Engineering and Technology (IRJET), Volume:07, Issue:12, Dec 2020. [3] K.M. Kavitha, Asha Shetty, Bryan Abreo, Adline D’Souza, Akarsha Kondana, “Analysis and Classification of User C o m m e n t s o n Yo u Tu b e Vi d e o s ” , ScienceDirect, Nov 2020. [4] Deepali K., Gaikwad, C. Namrata Mahender, “A Review Paper on Text Summarization”, International Journal of Advanced Research i n C o m p u t e r a n d C o m m u n i c a t i o n Engineering. [5] DuyDucAn Bui PhD, Guilherne DelFiol MD, PhD, John F. Hurdle MD, PhD, Siddhartha Jonnalagadda PhD, “Extractive text summarization system to aid data extraction from full text in systematic r e v i e w d e v e l o p m e n t ” , J o u r n a l o f Biomedical Informatics, Oct 2016. [6] Shi Yuan, Junjie Wu, Lihong Wang and Qing Wang, \"A Hybrid Method for Multi-class Sentiment International Journal of Intelligent Systems and Applications in Engineering IJISAE, 2024, 12(16s), 597– 601 | 600 Analysis of Microblogs\", ISBN-978-1-5090-2842-9, 2016. [7] Neethu M S and Rajasree R, “Sentiment Analysis in Youtube using Machine Learning Techniques”. [8] Aliza Sarlan, Chayanit Nadam, and Shuib Basri, \"Youtube Sentiment Analysis\", 2014 International Conference on Information Technology and Multimedia (ICIMU), Putrajaya, Malaysia November 18 – 20, 2014. [9] B. Gupta, M. Negi, K. Vishwakarma, G. Rawat, and P. Bandhani, \"Study of Youtube Sentiment Analysis using Machine Learning Algorithms on Python,\" Int.J. Comput. Appl., vol. 165, no. 9, pp. 29–34, May 2017. [10] “Computationally Efficient Learning of Quality Controlled Word Embeddings for Natural Language Processing,\" 2019 IEEE Comput. Soc. Annu. Symp. On, p. 134, 2019. Opinion Mining”, Kluwer Academic Publishers. Printed in the Netherlands, 2006. [11] Hearst, M., “Direction-based text interpretation as an information access refinement”, In Paul Jacobs, editor, Text Based Intelligent Systems. Lawrence Erlbaum Associates, 1992. [12] Das, S., and Chen, M., “Yahoo! for Amazon: Extracting market sentiment from stock message boards”, In Proc. of the 8th Asia Pacific Finance Association Annual C o n f e r e n c e ( A P FA 2 0 0 1 ) , 2 0 0 1 . unsupervised classification of reviews”. In Proc. of the ACL, 2002. [13] Argamon-Engelson, S., Koppel, M., a n d Av n e r i , G . , “ S t y l e b a s e d t e x t categorization: What newspaper am I reading? ”, In Proc. of the AAAI Workshop on Text Categorization, pages 1–4, 1998. [14] Pang, B. & Lee, L., “A Sentimental Education: Sentiment Analysis Using Subjectivity Summarization Based on M i n i m u m C u t s ” , A s s o c i a t i o n of Computational Linguistics (ACL), 2004. [10]Jin, W., & HOH. H., “A novel l e x i c a l i z e d H M M - b a s e d l e a r n i n g framework for web opinion mining”, P r o c e e d i n g s o f t h e 2 6 t h A n n u a l International Conference on Machine Learning. Montreal, Quebec, Canada,ACM: 465-472, 2009. [15] Brody, S., & Elhadad, N., “An unsupervised aspect-sentiment model for online reviews”, Human Language Te c h n o l o g i e s : T h e 2 0 1 0 A n n u a l Conference of the North American Chapter of the Association for Computational Linguistics. Los Angeles, California, Association for Computational Linguistics: 804-812, 2010. [16] Wiebe, J., Wilson, T., and Cardie, C., “Annotating expressions of opinions and e m o t i o n s i n l a n g u a g e ” . L a n g u a g e Resources and Evaluation, 2005.

Copyright

Copyright © 2025 Aditya Chauhan, Sumit Chaudhary, Ashwini Tomar, Ritika Singh, Ajai Kumar. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.