Budget Argument Mining Dataset Using Japanese Minutes from the National Diet and Local Assemblies

Proceedings of the 13th Conference on Language Resources and Evaluation (LREC 2022), pages 6131–6138 Marseille, 20-25 June 2022 © European Language Resources Association (ELRA), licensed under CC-BY-NC-4.0 6131 Budget Argument Mining Dataset Using Japanese Minutes from the National Diet and Local Assemblies Yasutomo Kimura, Hokuto Ototake, Minoru Sasaki Otaru University of Commerce, Fukuoka University, Ibaraki university 3-5-21 Midori, Otaru-shi, Hokkaido, Japan 8-19-1 Nanakuma, Jonan-ku, Fukuoka-shi, Fukuoka, Japan 4-12-1 Nakanarusawa, Hitachi-shi, Ibaraki, Japan kimura@res.otaru-uc.ac.jp, ototake@fukuoka-u.ac.jp, minoru.sasaki.01@vc.ibaraki.ac.jp Abstract Budget argument mining attempts to identify argumentative components related to a budget item, and then classifies these argumentative components, given budget information and minutes. We describe the construction of the dataset for budget argument mining, a subtask of QA Lab-PoliInfo-3 in NTCIR-16. Budget argument mining analyses the argument structure of the minutes, focusing on monetary ex- pressions (amount of money). In this task, given sufficient budget information (budget item, budget amount etc.), relevant argumentative components in the minutes are identified and argument labels (claim, premise and other) are assigned their components. In this paper, we describe the design of the data format, the annotation procedure, and release information of budget argument mining dataset, to link budget information to minutes. Keywords: Japanese assembly minutes, budget, argument mining 1. Introduction Politics has the important role of creating a budget that determines how government funds will be spent, con- sidering income and expenditures. National and lo- cal budget deliberations are held in the National Diet and Local assemblies. The national budget is drafted by the cabinet and discussed in the National Diet be- fore becoming the official budget. The budgets of local governments are proposed by the governors or mayors, and are discussed and approved in the assembly. How- ever, most people will have difficulty understanding the background of the proposed budget, as well as the dis- cussions that lead to the final budget. In the field of natural language processing, argu- ment mining has been studied to elucidate the dis- cussion process of discussion1．Argument mining attempts to identify argumentative structures from natural language text. The identification of argu- mentation structures involves several sub-tasks such as separating arguments, classifying argument com- ponents into claims and premises, and identify- ing argumentative relations. Argument mining is a task that analyzes argument structures, which as- signs labels (claims and premise) to discourse com- ponents such as sentences and clauses in essays (Kuribayashi et al., 2019)(Stab and Gurevych, 2017)． Budget information for national and local governments is available on the web. For example, the Global Open Data Index provides the most comprehensive snapshot available on the state of open government data publi- cation2 although the index did not apply any text min- ing tasks related to natural language processing. The study on the Malta Government Budget is an exam- 1http://acl2016tutorial.arg.tech/ 2https://index.okfn.org/dataset/ ple of studies on government budgets using text min- ing. Malta Government Budget 2018 dataset comprises over 500 online posts in English, and/or in the Mal- tese less-resourced language, gathered from social me- dia platforms (Cortis and Davis, 2019). However …

Content truncated to 500 words