Abstract by Chih-Han Tsai
Applying Machine Learning to Chinese Jiapu
Jia Pu, the ancient Chinese genealogical records contain precious information about culture and family spanning hundreds, and sometimes thousands of years. The many different formats and handwriting present unique challenges inherent in extracting family history information from Jia Pu. We have created a semi-automated Asian Records Transcription System (ARTS) to extract genealogical information from these records and make them more accessible to the world. To further automate this work, we are using ARTS to extract training data, so that we can train a neural network to localize and recognize critical information in the records, and finally making the data searchable through the internet.