I visited the University of Cambridge for a month to conduct a joint project “Extracting argument structure from legal documents” through WRHI project and aid.

Thanks to WRHI aid, I could conduct the joint research with Professor Simone Teufel at Computer Laboratory, the University of Cambridge for a month. Prof. Simone Teufel is a specialist of automatic summarization and an inventor of “Argument Zoning” which is a rhetorical analysis framework for the scientific articles. We are now trying to adapt and extend it for legal documents to generate more precise summarization which can preserve the argumentative structure from the original documents. We conducted several machine learning experiments for automatic extraction of the argumentative structure during the visiting. Especially, we used the deep learning architecture to implement the models for Argument Zoning classification. Although we, unfortunately, could not observe the major improvement with the models from our former models with handcrafted features, it was good to know that the handcrafted features perform is competitive. The results we got during the visiting will be carefully examined, and we aim to publish it with more experiments in some international conferences.

Moreover, thanks to WRHI project, I could make a short trip to the University of Edinburgh and the University of Bristol to exchange opinions and build a relationship for future joint researches.

I visited Professor Clare Grover at the University of Edinburgh. Thanks to her introduction, I could have a meeting with Arlene Casey and Clare Llewellyn. Arlene Casey studies on the part of scientific articles with Argument Zoning based approaches and Clare Llewellyn studies on topic extraction from user submissions on SNS websites. We discussed appropriate machine learning approaches when the only a small amount of data is available. We agreed to keep in touch to exchange useful information for topic extraction and argument zoning.

At the University of Bristol, I visited Dr. Oliver Ray. His group now studies automatic extraction of Ratio Decidendi (the most critical part in the judgment) from UK judgment documents and his topic and interest is closely related to ours. We shared our ideas and exchanged opinions. We found we have the same point of view and motivations which was the benefit from face to face communication. As a result, we agreed that we could conduct the joint research project especially for the argument extraction from the judgment documents. I will keep in touch with Oliver Ray and his group for future joint research.

Doctoral student, Hiroaki Yamada
Tokunaga Lab
Department of Computer 
Science School of Computing