Automatically Temporal Labeled Data Generation Using Positional Lexicon Expansion for Focus Time Estimation of News Articles

Many facts change over time, which is a fundamental aspect of our physical environment. In the case pandemic articles, user not interested in creation date document, but and cause last pandemic. Fake news can be better combated by having document with temporal focus. Currently, neither sequence events nor focus considered when obtaining documents. Despite limited number aspects available datasets, it difficult to test evaluate conclusions model. The goal this work develop article retrieval model based on co-training advance research semi-supervised learning. A mapping dataset performed using 1) evolving time 2) method coincidence contexts for learning low-dimensional continuous vectors neural contrast embedding models generating time-based query sequential articles facilitate understanding vectors. diverse used effectiveness proposed method. With lexicon expansion, result developed achieve 89%. than previous baselines traditional machine improvements 12.65% 4.7%, respectively.