Copyright © 2026 Authors retain the copyright of this article. This article 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.

@article{200944,
        author = {Navin Kumar P and Prathip S and Ragul Gandhi G and SyrilKumarL and Dr. A. Jagan},
        title = {Unified Customer Feedback Analysis Tool for Multi-Platform Social Networks},
        journal = {International Journal of Innovative Research in Technology},
        year = {2026},
        volume = {12},
        number = {no},
        pages = {78-99},
        issn = {2349-6002},
        url = {https://ijirt.org/article?manuscript=200944},
        abstract = {The exponential rise of social media platforms such as Twitter/X, Reddit, YouTube, and Instagram has generated an unprecedented volume of user-generated content containing valuable customer feedback. Extracting actionable insights from this multi-platform data is a critical challenge for businesses seeking to understand customer sentiment in real time. Existing tools are largely platformspecific, rely on shallow lexicon-based classifiers, and fail to handle the linguistic complexity of social media text including slang, sarcasm, emojis, and code-switching. This paper presents the Unified Customer Feedback Analysis Tool (UNIFIED CUSTOMER FEEDBACK ANALYSIS TOOL FOR MULTI-PLATFORM
SOCIAL NETWORKS), a comprehensive NLP-based system designed to collect, preprocess, analyze, and visualize customer feedback from multiple social networks simultaneously. The proposed system employs a multiagent architecture in which four specialized agents — Data Acquisition, Text Preprocessing, Semantic Analysis, and Reporting
— handle the complete pipeline under the coordination of a central Orchestration Hub. Text preprocessing uses NLTK and TextBlob for an eight-stage normalization pipeline. Sentiment classification uses a fine-tuned DistilBERTa-large transformer model with a contrastive learning objective for sarcasm detection. Interactive reporting is delivered via a Streamlit dashboard with Matplotlib and Seaborn visualizations including pie charts, temporal trend plots, aspect heatmaps, and keyword clouds. The system classifies feedback into Positive, Negative, and Neutral categories and provides aspect-level breakdowns identifying specific topics driving customer sentiment. The two-stage classification pipeline combines TextBlob rapid polarity pre-screening with DistilBERTalarge inference, reducing transformer load by 23% while achieving 3,200 posts per second throughput. Experimental evaluation demonstrates 95.8% classification accuracy on the Sentiment140 benchmark dataset, outperforming all prior baselines including VADER (68.4%), Naive Bayes (74.1%), BERT (91.7%), and standalone DistilBERTa-large (94.3%). Sarcasm detection F1-score improved by 8.3 percentage points through contrastive learning. The complete implementation uses Python with Pandas, NLTK, TextBlob, Matplotlib, Streamlit, and HuggingFace Transformers, making the system fully opensource, reproducible, and accessible at zero licensing cost for academic and enterprise use.},
        keywords = {.},
        month = {May},
        }