Monitoring method of public opinion in minor languages using deep learning

Qianli Song; Hua Lai

doi:10.3788/IRLA20210298

Journals >Infrared and Laser Engineering >Volume 50 >Issue S2 >Page 20210298 > Article

Infrared and Laser Engineering
Vol. 50, Issue S2, 20210298 (2021)

Monitoring method of public opinion in minor languages using deep learning

Qianli Song^1、2 and Hua Lai^1、2

Author Affiliations

¹Faculty of Information Engineering and Automation, Kunming University of Science and Technology, Yunnan 650500, China

²Yunnan Key Laboratory of Artificial Intelligence, Kunming University of Science and Technology, Yunnan 650500, China

show less

DOI: 10.3788/IRLA20210298 Cite this Article

Qianli Song, Hua Lai. Monitoring method of public opinion in minor languages using deep learning[J]. Infrared and Laser Engineering, 2021, 50(S2): 20210298 Copy Citation Text

show less

Fig. 1. Flow chart of Chinese-Vietnamese news opinion sentence extraction incorporating the characteristics of shared topics

Download full size | View in the Article

Fig. 2. Opinion sentence extraction model incorporating the discriminative features of opinion sentences

Download full size | View in the Article

	Number of Chinese news articles	Number of Vietnamese news articles
Training set	450	450
Test set	25	25
Validation set	25	25

Table 1. Distribution of training set, test set, and verification set of Chinese and Vietnamese news corpus

View in the Article

Opinion sentence extraction model	P	R	F₁
LSTM + opinion sentence discriminative features	0.623	0.639	0.631
Bi-LSTM + discriminant features of opinion sentences	0.658	0.667	0.662
Transformer+discriminant features of opinion sentences	0.711	0.732	0.721

Table 2. Comparison of the effect of opinion sentence extraction under different models

View in the Article

Discriminant features of opinion sentences	P	R	F₁
None	0.638	0.650	0.644
Shared topic	0.686	0.695	0.690
Position	0.664	0.683	0.673
Emotion	0.699	0.707	0.703
Vietnamese theme features + location + emotion	0.676	0.709	0.650
Shared theme topic + location + emotion	0.711	0.732	0.721

Table 3. Comparison of the effect of opinion sentence extraction under different characteristic

View in the Article

	Chinese news	Vietnamese news
Text	越通社河内—12月21日和22日，越南有关部门、越南驻美国大使馆、越南国家航空公司同当地政府有关部门配合，将在美国滞留的近360名越南公民安全接回国。 ...今后，将在国外滞留的公民接回国工作将继续根据公民的愿望和国内疫情和隔离能力等情况展开。	Trong hai ngày 21-22/12, các cơ quan chức năng Việt Nam, các cơ quan đại diện Việt Nam tại Hoa Kỳ, hãng Hàng không Quốc gia Việt Nam đã phối hợp với các cơ quan chức năng sở tại đưa gần 360 công dân Việt Nam về nước an toàn....Thời gian tới, việc đưa công dân có hoàn cảnh đặc biệt khó khăn về nước sẽ được sắp xếp theo nguyện vọng của công dân, phù hợp với tình hình dịch bệnh và năng lực cách ly trong nước.
Subject headings	越南，疫情...	Việt Nam(越南), cách ly (隔离)...
No shared theme feature	Opinion sentence number is 1, 8 (8 sentences in total)	Opinion sentence number is 1 (9 sentences in total)
Incorporate shared theme features	Opinion sentence number is 1, 8 (8 sentences in total)	Opinion sentence number is 1, 9 (9 sentences in total)
Artificially annotated opinion sentence	Opinion sentence number is 1, 8 (8 sentences in total)	Opinion sentence number is 1, 9 (9 sentences in total)

Table 4. Comparison of the effect of opinion sentence extraction in Chinese-Vietnamese news cases

Qianli Song, Hua Lai. Monitoring method of public opinion in minor languages using deep learning[J]. Infrared and Laser Engineering, 2021, 50(S2): 20210298

Download Citation

Tools

Save the article for my favorites

Paper Information