AI-DRIVEN INAPPROPRIATE CONTENT DETECTION AND CLASSIFICATION IN YOUTUBE VIDEOS USING DEEP LEARNING
Keywords:
Deep Learning, Content Filtering, Neural Networks, Convolution Neural Networks (CNN).Abstract
Video information is one of the most convenient and contemporary methods of staying informed about current events. The popularity of video content on the internet is on the rise, and it is having a significant impact on various aspects of our lives, such as education, entertainment, and communication. Video content is one of the most captivating forms of information, as it not only captivates viewers with its visuals but also facilitates the acquisition of knowledge and comprehension. The primary resource for the development and categorization of the text that we intend to use for the endeavor is YouTube. YouTube is considered to be one of the most pleasurable methods of acquiring global information. The primary objective of our endeavor is to generate and organize video content into distinct categories. We consider YouTube videos that include translations. The primary objective is to extract and categorize data from videos. The process involves the extraction of text that may contain undesirable letters or symbols through the use of natural language processing (NLP), which necessitates text cleaning. In essence, NLP is employed to evaluate pertinent data. Special text processing methods, such as tokenization and stemming, may be necessary to derive meaningful information from text. A replica of the YouTube URL has been incorporated into the front-end web page. The entire process commences upon the URL's upload. The connected dataset is employed to produce text that includes subtitles and natural language processing (NLP), which eliminates stop words and generates keywords. The preprocessed records and keywords generated are contained in the CSV file. A summary is generated after pre-processing, and the extracted text is arranged according to synonyms and keywords. The entire process is transmitted into the LSTM model to train and validate it for precise output. The system will generate a categorized summary based on the URLs that users submit. The Flask framework is employed to produce an interactive webbased output for the project.
