Full Text:
<33>
CLC number:
On-line Access: 2024-12-16
Received: 2024-03-31
Revision Accepted: 2024-09-22
Crosschecked: 0000-00-00
Cited: 0
Clicked: 41
MAL: multilevel active learning with BERT for Chinese affective structure analysis
| |||||||||||||
Shufeng XIONG, Guipei ZHANG, Xiaobo FAN, Wenjie TIAN, Lei XI, Hebing LIU, Haiping SI. MAL: multilevel active learning with BERT for Chinese affective structure analysis[J]. Frontiers of Information Technology & Electronic Engineering, 1998, -1(-1): .
@article{title="MAL: multilevel active learning with BERT for Chinese affective structure analysis",
author="Shufeng XIONG, Guipei ZHANG, Xiaobo FAN, Wenjie TIAN, Lei XI, Hebing LIU, Haiping SI",
journal="Frontiers of Information Technology & Electronic Engineering",
volume="-1",
number="-1",
pages="",
year="1998",
publisher="Zhejiang University Press & Springer",
doi="10.1631/FITEE.2400242"
}
%0 Journal Article
%T MAL: multilevel active learning with BERT for Chinese affective structure analysis
%A Shufeng XIONG
%A Guipei ZHANG
%A Xiaobo FAN
%A Wenjie TIAN
%A Lei XI
%A Hebing LIU
%A Haiping SI
%J Journal of Zhejiang University SCIENCE C
%V -1
%N -1
%P
%@ 2095-9184
%D 1998
%I Zhejiang University Press & Springer
%DOI 10.1631/FITEE.2400242
TY - JOUR
T1 - MAL: multilevel active learning with BERT for Chinese affective structure analysis
A1 - Shufeng XIONG
A1 - Guipei ZHANG
A1 - Xiaobo FAN
A1 - Wenjie TIAN
A1 - Lei XI
A1 - Hebing LIU
A1 - Haiping SI
J0 - Journal of Zhejiang University Science C
VL - -1
IS - -1
SP -
EP -
%@ 2095-9184
Y1 - 1998
PB - Zhejiang University Press & Springer
ER -
DOI - 10.1631/FITEE.2400242
Abstract: Chinese textual affective structure analysis is a sequence labeling task that often relies on supervised deep learning methods. However, acquiring a large annotated dataset for training can be expensive and time-consuming. active learning offers a solution by selecting the most valuable samples to reduce labeling costs. Previous approaches have focused on uncertainty or diversity but faced challenges such as biased models or selecting insignificant samples. To address these issues, this paper introduces multilevel active learning (MAL), which leverages the power of deep textual information at both the sentence and word levels, taking into account the complex structure of the Chinese language. By integrating the sentence-level features extracted from BERT embeddings and the word-level probability distributions obtained through a CRF model, MAL comprehensively captures the affective structure of Chinese text. Experimental results demonstrate that MAL significantly reduces annotation costs by approximately 70% and achieves more consistent performance compared to baseline strategies.
Open peer comments: Debate/Discuss/Question/Opinion
<1>