Please wait a minute...
Advanced Search
数据分析与知识发现  2021, Vol. 5 Issue (7): 101-110     https://doi.org/10.11925/infotech.2096-3467.2020.1216
  研究论文 本期目录 | 过刊浏览 | 高级检索 |
基于不确定性损失函数和任务层级注意力机制的多任务谣言检测研究*
杨晗迅,周德群,马静(),罗永聪
南京航空航天大学经济与管理学院 南京 211100
Detecting Rumors with Uncertain Loss and Task-level Attention Mechanism
Yang Hanxun,Zhou Dequn,Ma Jing(),Luo Yongcong
College of Economics and Management, Nanjing University of Aeronautics and Astronautics, Nanjing 211100, China
全文: PDF (996 KB)   HTML ( 14
输出: BibTeX | EndNote (RIS)      
摘要 

【目的】 通过引入不确定性损失函数和层级注意力机制,解决多任务谣言检测研究中主观设定主任务和辅助任务问题。【方法】 融合谣言勘探、立场检测和谣言检测任务的领域信息,构建改进的任务层级注意力机制模型。同时,首次在多任务谣言检测研究中,引入同方差不确定性损失函数,替代传统损失函数。最后使用PHEME数据集,将改进模型与传统多分类模型进行对比。【结果】 所提模型相比于目前最优模型,在Pheme4数据集中,Macro-F值提升4.2个百分点;在Pheme5数据集中,Macro-F值提升7.6个百分点。【局限】 只在Pheme数据集进行实验测试,对于其他谣言检测数据集未测试。【结论】 该模型在不划分主任务和辅助任务的情况下,仍可得到理想解。

服务
把本文推荐给朋友
加入引用管理器
E-mail Alert
RSS
作者相关文章
杨晗迅
周德群
马静
罗永聪
关键词 不确定性损失函数多任务学习谣言检测注意力机制    
Abstract

[Objective] This paper proposes a new model with the help of uncertainty loss function and task-level attention mechanism, aiming to address the issue of setting main and auxiliary tasks in rumor detection. [Methods] First, we integrated the domain knowledge of rumor exploration, stance classification, and rumor detectioin. Then, we constructed a modified model with task-level attention mechanism. Third, we used uncertainty loss function to explore the weight relationshaip of each task and obtain better detection results. Finally, we examined our model’s performance with the Pheme4 and Pheme5 datasets. [Results] Compared to the exisiting models, the Macro-F of our model increased by 4.2 and 7.6 percentage points with Pheme4 and Pheme5. [Limitations] We only examined our model with the Pheme dataset. [Conclusions] The proposed method could effective detect rumors without dividing the main and auxiliary tasks.

Key wordsUncertain Loss    Multi-task Learning    Rumor Detection    Attention Mechanism
收稿日期: 2020-12-06      出版日期: 2021-04-09
ZTFLH:  TP393  
基金资助:*南京航空航天大学前瞻性发展策略研究基金项目(NW2020001);国家社会科学基金重点项目(20ZDA092);南京航空航天大学研究生创新基地(实验室)开放基金项目(kfjj20200901)
通讯作者: 马静,ORCID: 0000-0001-8472-2518     E-mail: majing5525@126.com
引用本文:   
杨晗迅, 周德群, 马静, 罗永聪. 基于不确定性损失函数和任务层级注意力机制的多任务谣言检测研究*[J]. 数据分析与知识发现, 2021, 5(7): 101-110.
Yang Hanxun, Zhou Dequn, Ma Jing, Luo Yongcong. Detecting Rumors with Uncertain Loss and Task-level Attention Mechanism. Data Analysis and Knowledge Discovery, 2021, 5(7): 101-110.
链接本文:  
https://manu44.magtech.com.cn/Jwk_infotech_wk3/CN/10.11925/infotech.2096-3467.2020.1216      或      https://manu44.magtech.com.cn/Jwk_infotech_wk3/CN/Y2021/V5/I7/101
Fig.1  谣言识别流程
Fig.2  多任务模型
Fig.3  Branch_LSTM模型
Fig.4  基于任务层次注意力机制图
事件名称 事件文本量 回复文本量 疑似谣言 非疑似谣言 谣言 非谣言 无法判定
Charlie Hebdo 2 079 38 268 458 1621 193 116 149
Sydney Siege 1 221 23 996 522 699 382 86 54
Ferguson 1 143 24 175 284 859 10 8 266
Ottawa Shooting 890 12 284 470 420 329 72 69
Germanwings-crash 469 4 489 238 231 94 111 33
Putin Missing 238 835 126 112 0 9 117
Prince Toronto 233 902 229 4 0 222 7
Gurlitt 138 179 61 77 59 0 2
Elbola Essien 14 226 14 0 0 14 0
合计 6 425 105 354 2 402 4 023 1 067 638 697
Table 1  PHEME数据集
Fig.5  具有三个分支的谣言文本与评论的树状结构
操作系统配置 参数或版本
CPU Xeon(R) Gold 5218 CPU
GPU NVIDIA T4(16GB)
Python 3.6
TensorFlow 1.1.31
Keras 2.3.1
内存 1TB
Table 2  实验环境
实验 算法 acc Macro-F
实验一 Majority(True) 0.591 0.247
NileTMRG* 0.444 0.205
Branch-LSTM 0.466 0.362
MTL3 0.462 0.322
ES-ATT-MTL3 0.395 0.263
Task-ATT-MTL3 0.494 0.333
Un-Task-ATT-MTL3 0.425 0.364
实验二 Majority(True) 0.511 0.226
NileTMRG* 0.438 0.339
Branch-LSTM 0.454 0.336
MTL3 0.492 0.396
ES-ATT-MTL3 0.459 0.280
Task-ATT-MTL3 0.505 0.372
Un-Task-ATT-MTL3 0.467 0.472
实验三 Majority(True) 0.444 0.205
NileTMRG* 0.360 0.297
Branch-LSTM 0.314 0.259
MTL3 0.405 0.405
ES-ATT-MTL3 0.356 0.240
Task-ATT-MTL3 0.418 0.347
Un-Task-ATT-MTL3 0.385 0.393
Table 3  谣言检测任务实验结果
事件 acc Macro-F 谣言类
平均
F1
非谣言类平均F1 无法确定类平均F1
Charlie Hebdo 0.292 0.213 0.147 0.131 0.362
Sydney Siege 0.339 0.204 0.400 0.162 0.501
Ferguson 0.697 0.268 0 0.004 0.801
Ottawa Shooting 0.675 0.306 0.806 0.107 0.010
Germanwings-crash 0.356 0.245 0.310 0.355 0.091
Table 4  单事件结果
[1] Zubiaga A, Aker A, Bontcheva K, et al. Detection and Resolution of Rumours in Social Media: A Survey[J]. ACM Computing Surveys (CSUR), 2018, 51(2):1-36.
[2] 陈燕方, 李志宇, 梁循, 等. 在线社会网络谣言检测综述[J]. 计算机学报, 2018, 41(7):1648-1676.
[2] (Chen Yanfang, Li Zhiyu, Liang Xun, et al. Review on Rumor Detection of Online Social Networks[J]. Chinese Journal of Computers, 2018, 41(7):1648-1676.)
[3] Qazvinian V, Rosengren E, Radev D, et al. Rumor Has It: Identifying Misinformation in Microblogs[C]// Proceedings of the 2011 Conference on Empirical Methods in Natural Language Processing. 2011: 1589-1599.
[4] Liang G, He W B, Xu C, et al. Rumor Identification in Microblogging Systems Based on Users’ Behavior[J]. IEEE Transactions on Computational Social Systems, 2015, 2(3):99-108.
doi: 10.1109/TCSS.2016.2517458
[5] Kwon S, Cha M, Jung K, et al. Prominent Features of Rumor Propagation in Online Social Media[C]// Proceedings of the 13th International Conference on Data Mining. IEEE, 2013: 1103-1108.
[6] Kochkina E, Liakata M, Zubiaga A. All-in-One: Multi-task Learning for Rumour Verification[OL]. arXiv Preprint, arXiv: 1806.03713.
[7] Sejeong K, Meeyoung C, Kyomin J, et al. Rumor Detection over Varying Time Windows[J]. PLoS One, 2017, 12(1):e0168344.
doi: 10.1371/journal.pone.0168344
[8] Yang F, Liu Y, Yu X H, et al. Automatic Detection of Rumor on Sina Weibo[C]// Proceedings of the ACM SIGKDD Workshop on Mining Data Semantics. 2012: 13.
[9] Chang C, Zhang Y H, Szabo C, et al. Extreme User and Political Rumor Detection on Twitter[C]// Proceedings of International Conference on Advanced Data Mining and Applications. 2016: 751-763.
[10] Ma J, Gao W, Mitra P, et al. Detecting Rumors from Microblogs with Recurrent Neural Networks[C]// Proceedings of the 25th International Joint Conference on Artificial Intelligence. 2016: 3818-3824.
[11] Chen T, Li X, Yin H Z, et al. Call Attention to Rumors: Deep Attention Based Recurrent Neural Networks for Early Rumor Detection[C]// Proceedings of Pacific-Asia Conference on Knowledge Discovery and Data Mining. Springer, Cham, 2018: 40-52.
[12] Collobert R, Weston J. A Unified Architecture for Natural Language Processing: Deep Neural Networks with Multitask Learning[C]// Proceedings of the 25th International Conference on Machine Learning. 2008: 160-167.
[13] Liu P F, Qiu X P, Huang X J. Recurrent Neural Network for Text Classification with Multi-task Learning[OL]. arXiv Preprint, arXiv: 1605.05101.
[14] Dong D X, Wu H, He W, et al. Multi-task Learning for Multiple Language Translation[C]// Proceedings of the 53rd Annual Meeting of the Association for Computational Linguistics and the 7th International Joint Conference on Natural Language Processing (Volume 1: Long Papers). 2015: 1723-1732.
[15] Firat O, Cho K, Bengio Y. Multi-way, Multilingual Neural Machine Translation with a Shared Attention Mechanism[OL]. arXiv Preprint, arXiv: 1601.01073.
[16] Ma J, Gao W, Wong K F. Detect Rumor and Stance Jointly by Neural Multi-task Learning[C]// Companion Proceedings of the Web Conference 2018. 2018:585-593.
[17] Li Q Z, Zhang Q, Si L. Rumor Detection by Exploiting User Credibility Information, Attention and Multi-task Learning[C]// Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics. 2019: 1173-1179.
[18] Sener O, Koltun V. Multi-task Learning as Multi-objective Optimization[C]// Proceedings of the 32nd Conference on Neural Information Processing Systems. 2018: 527-538.
[19] Kendall A, Gal Y, Cipolla R. Multi-task Learning Using Uncertainty to Weigh Losses for Scene Geometry and Semantics[C]// Proceedings of the 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition. 2018: 7482-7491.
[20] Kendall A, Gal Y. What Uncertainties do We Need in Bayesian Deep Learning for Computer Vision?[C]// Proceedings of the 31st International Conference on Neural Information Processing Systems. 2017: 5580-5590.
[21] Gal Y. Uncertainty in Deep Learning[D]. University of Cambridge, 2016.
[22] Staudemeyer R C, Morris E R. Understanding LSTM - a Tutorial into Long Short-Term Memory Recurrent Neural Networks[OL]. arXiv Preprint, arXiv: 1909. 09586.
[23] Fabbri M, Moro G. Dow Jones Trading with Deep Learning: The Unreasonable Effectiveness of Recurrent Neural Networks[C]// Proceedings of the 7th International Conference on Data Science, Technology and Applications. 2018: 142-153.
[24] Kochkina E, Liakata M, Augenstein I. Turing at Semeval-2017 Task 8: Sequential Approach to Rumour Stance Classification with Branch-LSTM[OL]. arXiv Preprint, arXiv: 1704.07221.
[25] Mikolov T, Chen K, Corrado G, et al. Efficient Estimation of Word Representations in Vector Space[OL]. arXiv Preprint, arXiv: 1301.3781.
[26] Enayet O, El-Beltagy S R. NileTMRG at SemEval-2017 Task 8: Determining Rumour and Veracity Support for Rumours on Twitter[C]// Proceedings of the 11th International Workshop on Semantic Evaluation (SemEval-2017). 2017: 470-474.
[1] 范涛,王昊,吴鹏. 基于图卷积神经网络和依存句法分析的网民负面情感分析研究*[J]. 数据分析与知识发现, 2021, 5(9): 97-106.
[2] 谢豪,毛进,李纲. 基于多层语义融合的图文信息情感分类研究*[J]. 数据分析与知识发现, 2021, 5(6): 103-114.
[3] 尹鹏博,潘伟民,张海军,陈德刚. 基于BERT-BiGA模型的标题党新闻识别研究*[J]. 数据分析与知识发现, 2021, 5(6): 126-134.
[4] 余本功,朱晓洁,张子薇. 基于多层次特征提取的胶囊网络文本分类研究*[J]. 数据分析与知识发现, 2021, 5(6): 93-102.
[5] 韩普,张展鹏,张明淘,顾亮. 基于多特征融合的中文疾病名称归一化研究*[J]. 数据分析与知识发现, 2021, 5(5): 83-94.
[6] 段建勇,魏晓鹏,王昊. 基于多角度共同匹配的多项选择机器阅读理解模型 *[J]. 数据分析与知识发现, 2021, 5(4): 134-141.
[7] 王雨竹,谢珺,陈波,续欣莹. 基于跨模态上下文感知注意力的多模态情感分析 *[J]. 数据分析与知识发现, 2021, 5(4): 49-59.
[8] 蒋翠清,王香香,王钊. 基于消费者关注度的汽车销量预测方法研究*[J]. 数据分析与知识发现, 2021, 5(1): 128-139.
[9] 尹浩然,曹金璇,曹鲁喆,王国栋. 扩充语义维度的BiGRU-AM突发事件要素识别研究*[J]. 数据分析与知识发现, 2020, 4(9): 91-99.
[10] 黄露,周恩国,李岱峰. 融合特定任务信息注意力机制的文本表示学习模型*[J]. 数据分析与知识发现, 2020, 4(9): 111-122.
[11] 石磊,王毅,成颖,魏瑞斌. 自然语言处理中的注意力机制研究综述*[J]. 数据分析与知识发现, 2020, 4(5): 1-14.
[12] 薛福亮,刘丽芳. 一种基于CRF与ATAE-LSTM的细粒度情感分析方法*[J]. 数据分析与知识发现, 2020, 4(2/3): 207-213.
[13] 祁瑞华,简悦,郭旭,关菁华,杨明昕. 融合特征与注意力的跨领域产品评论情感分析*[J]. 数据分析与知识发现, 2020, 4(12): 85-94.
[14] 徐彤彤,孙华志,马春梅,姜丽芬,刘逸琛. 基于双向长效注意力特征表达的少样本文本分类模型研究*[J]. 数据分析与知识发现, 2020, 4(10): 113-123.
[15] 姚俊良,乐小虬. 科技查新查新点语义匹配方法研究[J]. 数据分析与知识发现, 2019, 3(6): 50-56.
Viewed
Full text


Abstract

Cited

  Shared   
  Discussed   
版权所有 © 2015 《数据分析与知识发现》编辑部
地址:北京市海淀区中关村北四环西路33号 邮编:100190
电话/传真:(010)82626611-6626,82624938
E-mail:jishu@mail.las.ac.cn