Date of Birth Information Extraction from the Internet by Deep Learning Approach

The problem of ambiguity, incompleteness, and high expense of information extraction by human could be alleviated by using computer to extract such information, however, basic pattern-matching approach still has limited performance especially against a language with no punctuation or clear mark like Thai. This work proposes a Deep-learning approach based on BERT-th and Keras. Comparison with manually maintained websites and pattern matching approach, this paper shows that the deep learning approach can achieve better recall and precision especially for Thai language.