SceneAlert: A Mass Media Brand Listening Tool

Nowadays, a company's reputation is a main determinant of its success; it is beneficial for companies to be aware of what observers are saying about them in the media so that they may respond correctly. The proposed system is a website application that facilitates staying up to date with the latest mentions of companies. The system allows clients to enter keywords that represent what they want to monitor and select the media platforms they want to search. Our website application supports different media platforms, including social media platforms, blogs and articles, radio stations, podcasts, and television. Various tools and APIs implement data gathering from these platforms. The system accesses Twitter through the Twitter API v2, blogs and articles through the Bing Search API, and accesses podcasts through the iTunes Search API, and YouTube, television and radio using the y_dl tool. Audio preprocessing is implemented on the retrieved audio files, after which the Automated Speech Recognition (ASR) process is achieved using the Google Web Speech API, followed by the text preprocessing phase. The system detects the keyword mentions and alerts clients with the awaited context. It processes the data received to provide clients with data visualization on a web application after the data processing and analysis phase, which essentially includes sentiment analysis using the Mazajak API operating on a Convolutional Neural Network - Long Short-Term Memory (CNN-LSTM) model as well as topic modelling using BERTopic and text summarization using the Latent Dirichlet Allocation (LDA) topic model.