融合多粒度特征的缅甸语文本图像识别方法(Burmese Language Recognition Method Fused with Multi-Granularity Features)

Enyu He (何恩宇), Rui Chen (陈蕊), Cunli Mao (毛存礼), Yuxin Huang (黄于欣), Shengxaing Gao (高盛祥), Zhengtao Yu (余正涛)


Abstract
“缅甸语属于东南亚低资源语言,缅甸语文本图像识别对开展缅甸语机器翻译等任务具有重要意义。由于缅甸语属于典型的字符组合型语言,一个感受野内存在多个字符嵌套,现有缅甸语识别方法主要是从字符粒度进行识别,在解码时会出现某些字符未能正确识别而导致局部乱码。考虑到缅甸语存在特殊的字符组合规则,本文提出了一种融合多粒度特征的缅甸语文本图像识别方法,将较细粒度的字符粒度和较粗粒度的字符簇粒度进行序列建模,然后将两种粒度特征序列进行融合后利用解码器进行解码。实验结果表明,该方法能够有效缓解识别结果乱码的现象,并且在人工构建的数据集上相比“VGG16+BiLSTM+Transformer”的基线模型识别准确率提高2.4%,达到97.35%。 "
Anthology ID:
2023.ccl-1.11
Volume:
Proceedings of the 22nd Chinese National Conference on Computational Linguistics
Month:
August
Year:
2023
Address:
Harbin, China
Editors:
Maosong Sun, Bing Qin, Xipeng Qiu, Jing Jiang, Xianpei Han
Venue:
CCL
SIG:
Publisher:
Chinese Information Processing Society of China
Note:
Pages:
124–134
Language:
Chinese
URL:
https://aclanthology.org/2023.ccl-1.11
DOI:
Bibkey:
Cite (ACL):
Enyu He, Rui Chen, Cunli Mao, Yuxin Huang, Shengxaing Gao, and Zhengtao Yu. 2023. 融合多粒度特征的缅甸语文本图像识别方法(Burmese Language Recognition Method Fused with Multi-Granularity Features). In Proceedings of the 22nd Chinese National Conference on Computational Linguistics, pages 124–134, Harbin, China. Chinese Information Processing Society of China.
Cite (Informal):
融合多粒度特征的缅甸语文本图像识别方法(Burmese Language Recognition Method Fused with Multi-Granularity Features) (He et al., CCL 2023)
Copy Citation:
PDF:
https://aclanthology.org/2023.ccl-1.11.pdf