Automated detection of profanities for film censorship using deep learning
Given the excessive profanities identified in audio and video files and the detrimental consequences to an individual‟s character and behaviour, content censorship is crucial to filter profanities from young viewers with higher exposure to uncensored content. Although manual detection and censorship...
Saved in:
| Main Author: | |
|---|---|
| Format: | Thesis |
| Published: |
2022
|
| Subjects: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Given the excessive profanities identified in audio and video files and the detrimental consequences to an individual‟s character and behaviour, content censorship is crucial to filter profanities from young viewers with higher exposure to uncensored content. Although manual detection and censorship were implemented, the methods proved tedious. Inevitably, misidentifications involving profane language owing to human weariness and the low performance in human visual and hearing systems concerning long screening time occurred. Another challenge in this task is the lack of well-structured profane language dataset for research and application purposes. The thesis proposes profanities classification and detection models utilizing deep neural network and a novel profanity language dataset. This research proposed an intelligent model for profane words censorship through automated and robust detection by deep learning models. A dataset of profanities was collected and processed for the computation of audio spectral feature that serve as an input to evaluate the classification of spoken profane and normal words. The proposed classification models were first tested for 2-class (Profanity vs Normal) classification problem, the profanity class is then further decomposed into multi-class classification problem for exact detection of profanity. Experimental results show the viability of proposed system by demonstrating high performance of profane words classification with high speed. The proposed systems outperformed state-of-the-art pre-trained and baseline neural networks on the novel profanities dataset and proved to reduce the computational cost with minimal trainable parameters. Hence, proposed system was proven to be fast on screening and detection of audible profane content for films‟ censorship, as the time taken to process an input sample is only about 46% of the input sample‟s duration. |
|---|