• Wednesday, Oct 1st, 2025

International Journal of Advanced Research in Education and TechnologY(IJARETY)
International, Double Blind-Peer Reviewed & Refereed Journal, Open Access Journal
|Approved by NSL & NISCAIR |Impact Factor: 8.152 | ESTD: 2014|

|Scholarly Open Access Journals, Peer-Reviewed, and Refereed Journal, Impact Factor-8.152 (Calculate by Google Scholar and Semantic Scholar | AI-Powered Research Tool), Multidisciplinary, Bi-Monthly, Citation Generator, Digital Object Identifier(DOI)|

Article

TITLE Spoiler Shield: AI Powered NLP for Real Time Detection and Protection of Website Content
ABSTRACT Spoilers—revealing plot details of movies, TV shows, books, or games—pose a frequent annoyance for users browsing reviews or social media. Spoiler Shield is a proposed system that uses Natural Language Processing (NLP) and machine learning to detect, mask or block spoiler content in real time on websites, thereby protecting users from unwanted reveals. The system scans incoming text content (user reviews, forum posts, comments) and uses a model trained to classify sentences/paragraphs as “spoiler” or “nonspoiler.” It applies masking/blurring/hiding for spoiler content depending on user settings. Key contributions include: (1) a robust dataset of spoiler vs nonspoiler content gathered from multiple domains (movie reviews, fan forums, book discussion sites); (2) development of a transformer‐based classifier fine‐tuned on this dataset; (3) integration into browser extension / website middleware allowing live content filtering; (4) evaluation of accuracy, latency, user acceptability, and impact on user experience. In experiments, Spoiler Shield achieved an F1score of ~0.93, precision ~0.90, recall ~0.95 on test data, with average detection latency <150 ms per block of text, and minimal perceptible delay for users. A user study (n = 50) indicated high satisfaction: 85% of users felt spoilers were correctly blocked, 90% appreciated control over how spoilers are hidden, and less than 5% reported false positives that hindered content enjoyment. The system shows promise as a tool for enhancing content privacy in online browsing. Limitations include domain drift (novel plot devices), linguistic variability, and potential overblocking. Future work involves expanding to multilingual support, multimodal content (images / video transcripts), adaptive learning from user feedback, and enhancing granularity (blocking only certain spoiler levels).
AUTHOR Nayantara Sahgal Kapoor P. A. College of Engineering (PACE) Mangalore, India
PUBLICATION DATE 2025-09-21
VOLUME 12
DOI DOI:10.15680/IJARETY.2025.1201037
PDF 37_Spoiler Shield AI Powered NLP for Real Time Detection and Protection of Website Content.pdf
KEYWORDS