非構造化データを活用する試み

―議会会議録と有価証券報告書を例として―

抄録

This paper describes shared tasks that use natural language processing techniques to link textual and tabular data, which are treated as unstructured data. Specifically, it describes the shared tasks of finding the basis for the figures and amounts contained in the tables in the minutes of local council meetings and annual securities reports. The local assembly minutes and annual security reports provide data that are expected to be used as primary information. However, publicly available data often consist of unstructured document formats. Therefore, we designed a budget argument mining task, which analyses the argument structure of the minutes, focusing on monetary expressions (amount of money). Furthermore, we designed the shared tasks of text-table relationship extraction and table data extraction. In this paper, we provide an overview of these shared tasks.