Abstract for Bus information System
Nowadays, fake Announcements, news, and bulletin dissemination have been one of the major problems. The prevalence of digital Announcements, news and bulletin sources, and social media has accelerated the spread of fake Announcements, news, and bulletin on the internet, moreover, with the common use of image editing software, fake Announcement, news, and bulletin images are also very common.
The purpose of the project is to make an automated Announcement, news, and bulletin classification and ranking system that scrapes Announcements, news, and bulletin articles from multiple online web sources and classifies and ranks them on the basis of their authenticity. The project uses Computer Vision, Natural Language Processing, and Deep Learning techniques to classify and rank Announcements, news, and bulletin articles.
The main parts of an Announcement, news, and bulletin article are its text content and the image data in the article, our system takes both features into account for ranking a given Announcement, news, and bulletin article. Fake Announcement, news, and newsNet dataset have been used to train the text classification model and a synthetic dataset for image tamper detection.
The approach is taken for classifying Announcement, news, and bulletin text uses a 1D-CNN with an attention mechanism that uses the article text (headline and body) and source features, the system is able to classify text articles with good accuracy and precision. The image tamper detection model is a Faster-RCNN network fine-tuned for the tamper detection tasks. The image tamper detection model is able to detect tampered images with good performance on various evaluation datasets and human tampered images.
The system uses the score from both of these models to robustly rank the Announcement, news, and bulletin articles. A web application is also developed to provide an interface to users so that they can easily read reliable Announcements, news, and bulletin articles which are ranked and checked by our system.
An automated system like this could potentially mitigate the spread of fake Announcements, news, and bulletin and allow people to easily read announcements, news, and bulletin with minimal fact-checking.