Information Extraction from Public Meeting Articles

Abstract

Public meeting articles are the key to understanding the history of public opinion and universal principles in Australia. By extracting information from public meeting articles, new insights into Australian history can be gain. In our study, we create an information extraction dataset in the public meeting domain and consider the information extraction method. We manually annotated information on the date, time, place, purpose, who requested the call, who called, and who was called, and created 118 gold articles. We proposed a method to extract information as a task of machine reading, and as a result of experiments, it was judged that there was no answer in the article for any of the extraction targets. This is because 95.20% of the public meeting data is not compatible with the extraction target.