TIDM: Topic-Specific Information Detection Model

Abstract Nowadays information control and detection on the social network have become a problem that we should solve as soon as possible. Unfortunately, due to the informal expressions, detecting the massive data on the internet is a big challenge based on the traditional text mining methods such as Topic Model. In our paper, we propose a simple 4-Tuple Structure instead of the raw text event which usually contains many meaningless words. Using the word embedding technique, we propose the Topic-Specific Information Detection Model (TIDM) for detecting the specific information. For training the words and idiomatic phrases, we adopt the supervise learning technique: manually constructing a specific Semantic Dataset for training our model. Our experiments based on the Amazon Reviews demonstrate that the TIDM can effectively detect and recognize the information.