MIG at the NTCIR-15 FinNum-2 Task: Use the transfer learning and feature engineering for numeral attachme... | Publication

Title	MIG at the NTCIR-15 FinNum-2 Task: Use the transfer learning and feature engineering for numeral attachment task
Creator	劉昭麟 Liu, Chao-Lin Chen, Yu-Yu
Contributor	資科系
Key Words	Numeral attachment ; financial social media ; transfer learning ; feature engineering
Date	2020-12
Date Issued	22-Sep-2021 10:39:33 (UTC+8)
Summary	In the FinNum-2 task, the goal is to judge whether the specified numeral is related to the given stock symbol in a financial tweet. We employ a transfer-learning mechanism and the Google BERT embeddings so that we only need to collect and annotate a small amount of data to train the classifiers for the task. In addition, our classifiers consider some intuitive but useful syntactic features, e.g., the positions of words in the tweets. Experimental results indicate that these new features boost the prediction quality, and we achieved better than 68% in the tests in the formal run.
Relation	NTCIR-15 Proceedings, NII, Japan, pp.79‒82
Type	conference

dc.contributor	資科系
dc.creator (作者)	劉昭麟
dc.creator (作者)	Liu, Chao-Lin
dc.creator (作者)	Chen, Yu-Yu
dc.date (日期)	2020-12
dc.date.accessioned	22-Sep-2021 10:39:33 (UTC+8)	-
dc.date.available	22-Sep-2021 10:39:33 (UTC+8)	-
dc.date.issued (上傳時間)	22-Sep-2021 10:39:33 (UTC+8)	-
dc.identifier.uri (URI)	http://nccur.lib.nccu.edu.tw/handle/140.119/137222	-
dc.description.abstract (摘要)	In the FinNum-2 task, the goal is to judge whether the specified numeral is related to the given stock symbol in a financial tweet. We employ a transfer-learning mechanism and the Google BERT embeddings so that we only need to collect and annotate a small amount of data to train the classifiers for the task. In addition, our classifiers consider some intuitive but useful syntactic features, e.g., the positions of words in the tweets. Experimental results indicate that these new features boost the prediction quality, and we achieved better than 68% in the tests in the formal run.
dc.format.extent	554805 bytes	-
dc.format.mimetype	application/pdf	-
dc.relation (關聯)	NTCIR-15 Proceedings, NII, Japan, pp.79‒82
dc.subject (關鍵詞)	Numeral attachment ; financial social media ; transfer learning ; feature engineering
dc.title (題名)	MIG at the NTCIR-15 FinNum-2 Task: Use the transfer learning and feature engineering for numeral attachment task
dc.type (資料類型)	conference