Financial Numeral Classification Model Based on BERT
| Autor: | Maofu Liu (刘茂福), Junyi Xiang (向军毅), Ruibin Mao (毛瑞彬), Yukun Zhang, Wei Wang |
|---|---|
| Rok vydání: | 2019 |
| Předmět: | |
| Zdroj: | NII Testbeds and Community for Information Access Research ISBN: 9783030368043 NTCIR |
| DOI: | 10.1007/978-3-030-36805-0_15 |
| Popis: | Numerals contain rich semantic information in financial documents, and they play significant roles in financial data analysis and financial decision making. This paper proposes a model based on the Bidirectional Encoder Representations from Transformers (BERT) to identify the category and subcategory of a numeral in financial documents. Our model holds the obvious advantages in the fine-grained numeral understanding and achieves good performance in the FinNum task at NTCIR-14. The FinNum task is to classify the numerals in financial tweets into seven categories, and further extend these categories into seventeen subcategories. In our proposed model, we first analyze the obtained financial data from the FinNum task and enhance data for some subcategories by entity replacement. And then, we adopt our fine-tuning BERT model to finish the task. As a supplement, some popular traditional and deep learning models have been selected for comparative experiments, and the experimental results show that our model has achieved the state-of-the-art performances. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|