Metric-Type Identification for Multi-Level Header Numerical Tables in Scientific Papers

Proceedings of the 16th Conference of the European Chapter of the Association for Computational Linguistics, pages 3062–3071 April 19 - 23, 2021. ©2021 Association for Computational Linguistics 3062 Metric-Type Identification for Multi-Level Header Numerical Tables in Scientific Papers Lya Hulliyyatus Suadaa1, Hidetaka Kamigaito1, Manabu Okumura1, Hiroya Takamura1,2 1Tokyo Institute of Technology 2National Institute of Advanced Industrial Science and Technology (AIST) lya@stis.ac.id {kamigaito,oku}@lr.pi.titech.ac.jp takamura@pi.titech.ac.jp Abstract Numerical tables are widely used to present ex- perimental results in scientific papers. For ta- ble understanding, a metric-type is essential to discriminate numbers in the tables. We intro- duce a new information extraction task, metric- type identification from multi-level header nu- merical tables, and provide a dataset extracted from scientific papers consisting of header ta- bles, captions, and metric-types. We then propose two joint-learning neural classifica- tion and generation schemes featuring pointer- generator-based and BERT-based models. Our results show that the joint models can handle both in-header and out-of-header metric-type identification problems. 1 Introduction Tables are powerful tools for presenting data effi- ciently in row and column views. In scientific pa- pers, numerical tables are commonly used to show experimental results for facilitating data analysis. Examples of numerical tables in scientific papers are shown in Figure 1. Tables have the ability to cover multiple cate- gories written in table headers by incorporating several header sets in a hierarchical view, called multi-level header tables. Scientific papers have strict guidelines about tables; for example, one states that a similar type of text is written in the same level of header. Figure 1a shows a multi-level header example in the column part, with task type (Task 1 and Task 2) in the first header-level and metric-type (Prec and Rec) in the second. The ta- ble also has a row header specifying the model type (Model A, Model B, Model C, and Model D). In the real-world, this header-type information is limited due to the unknown table scheme. However, we assume tables in scientific papers follow the rule of Models Task 1 Task 2 Prec Rec Prec Rec Model A 60 60 60 60 Model B 70 70 70 70 Model C 80 80 80 80 Model D 90 90 90 90 Table X. Model comparison in Task 1 and 2. (a) Metric-type in header Models Task 1 Task 2 Model A 60 65 Model B 70 75 Model C 80 85 Model D 90 95 Table Y. Model comparison in Task 1 and 2 (F-score). (b) No metric-type in header Figure 1: Example tables in scientific papers. Bold in- dicates their metric-type. categorizing a similar type of header name in the same header-level. To understand the numbers in the tables, metric- types are important for discriminating the num- bers. A comparison between numbers is applied for numbers in the same metric-type with different categories. For the table in Figure 1a, we cannot compare the number 60 for Model A in the first col- umn with 60 in the second one because they have a different metric-type: Prec and …

Content truncated to 500 words