Smart Content Moderation in Digital Media Classrooms
Bibliographic Data
| ID | 22197366 |
|---|---|
| Authors | Baliram N Gaikwad (University of Mumbai), Mayank Deep Khare (0000-0002-0726-2711, Jaypee Institute of Information Technology), Lokesh Verma (0000-0001-5443-1963, Chitkara University), Sudharsan M (Presidency University), Prachi Rashmi (Noida International University), Anand Bhargava (Parul University), Subhash Kumar Verma, Subhash Verma (Noida International University), Subramanian Karthick (International Institute of Information Technology) |
| Year | 2025 |
| Volume | 6 |
| Issue | 4s |
| Pages | 495-505 |
| Publication date | 2025-12-25 |
| Open Access | Yes |
| Type | ARTICLE |
| Venue | ShodhKosh: Journal of Visual and Performing Arts (JOURNAL) |
| Journal identifiers | ISSN: 2582-7472 • E-ISSN: 2582-7472 |
| Publisher | Granthaalayah Publications and Printers (PUBLISHER • IN) |
| DOI | 10.29121/shodhkosh.v6.i4s.2025.6859 |
| OpenAlex | W7117411144 |
| Language | EN |
| References cited | 16 |
The high growth of digital media classrooms has turned the learning classes into the multimodal and much interacting environments where students are exposed to text, images, audio and video materials. On the one hand, this change increases the level of creativity and cooperation, on the other hand, it creates serious issues with balancing unsuitable, damaging, or biased content in the real-time. The traditional rule based systems tend to miss the nuance, context and multimodal aspect of the contemporary digital interactions. The paper has introduced a complete framework of smart content moderation that can be adapted to digital media classroom, utilizing the advanced natural language processing, computer vision and multimodal fusion techniques. The suggested system plan combines the real-time content acquisition, contextual filtering pipelines and toxicity detecting modules that can detect offensive words, graphic images, misinformation signals and gentle biases. A teacher-in-the-loop process provides human control over the system so that teachers can adjust the system responses and decrease the number of false positives with the help of constant feedback. Their methodology implies close curation and annotation of the datasets, followed by training transformer-based NLP models and CNN/ViT visual classifiers training them on the educational domain. An effective hybrid moderation system that is a combination of machine learning and rules filters guarantee sound and transparent decision making. The findings indicate that it has been found to improve accuracy, latency, and cross-modal consistency over traditional tools of moderation
Avatar · Digital content · Digital media · Misinformation · Moderation · Emotion and Mood Recognition · Hate Speech and Cyberbullying Detection · Sentiment Analysis and Opinion Mining
| Highly cited | No |
|---|