[May-20]Predicting Judicial Decisions on Post-divorce Child Custody by Machine Learning with NLP

Institute of Information Systems and Applications


Prof. Daw-Wei Wang

Physics Department, General Education Center, National Tsing-Hua University


Predicting Judicial Decisions on Post-divorce Child Custody by Machine Learning with NLP


13:30-15:00 Wednesday 20-May-2020


Delta 105

QR code:




Hosted by:

Prof. Cheng-Hsin Hsu


The doctrine of the "best interests of the child" is the primary consideration for family court judges in making post-divorce custody decisions. However, a precise understanding of the "best interests" standard remains elusive. For example, Article 1055-1 of Civil Code requires judges in Taiwan to consider many qualitative factors such as the wish of the child, the character, occupation, financial condition and lifestyle of the parents etc, and hence these descriptive or unstructured statements in human language make it difficult for traditional computation algorithms to predict or to simulate these decisions properly. In this talk, we show how to label sentences in court verdicts as a corpus and apply Natural Language Processing (NLP) techniques to train a machine learning model for such a task. Using the favorable and/or unfavorable statements of each parent, our model could provide a pretty high accuracy up to 93.23% for binary decision of custody (awarded to father or mother), demonstrating the potential to predict court decisions solely based on human language as the input information. We further extend our model to include cases with custody awarded to both parents. The overall accuracy for such ternary decision becomes 81.53%, due to the interference between cases with custody awarded to mother and to both parents. Finally, our model can be also applied for the dispute resolution, where the two parties could get prediction results for child custody immediately based on the information they provide without filing a formal lawsuit. An on-line version of PSCC is available for test in the following website:  https://custodyprediction.herokuapp.com/.


Daw-Wei Wang is a professor in Physics Department and a joint appointment professor in General Education Center, National Tsing-Hua University (NTHU). He also serves as Vice Director of Physics Division of National Center for Theoretical Sciences and Director of Counseling Center in NTHU. He has been awarded Daniel Tsui Fellowship (Hong Kong U.) and Ta-You Wu Memorial Award (MoST) for his research work in theoretical physics. In recent years, his major interest is to apply Artificial Intelligence for interdisciplinary research, including physics, astronomy, neuron sciences, laws and counselling. More information can be found in his research website: http://www.phys.nthu.edu.tw/~aicmt/

All faculties and students are welcome to join.