融合句法信息的无触发词事件检测方法

doi:10.11772/j.issn.1001-9081.2021060928

《计算机应用》唯一官方网站 ›› 2021, Vol. 41 ›› Issue (12): 3534-3539.DOI: 10.11772/j.issn.1001-9081.2021060928

• 第十八届中国机器学习会议(CCML 2021) • 上一篇

融合句法信息的无触发词事件检测方法

汪翠¹^,², 张亚飞¹^,²(), 郭军军¹^,², 高盛祥¹^,², 余正涛¹^,²

^1.昆明理工大学信息工程与自动化学院，昆明 650504
^2.云南省人工智能重点实验室（昆明理工大学），昆明 650504

收稿日期:2021-05-12 修回日期:2021-06-17 接受日期:2021-07-04 发布日期:2021-12-28 出版日期:2021-12-10
通讯作者: 张亚飞
作者简介:汪翠（1997—），女，云南昭通人，硕士研究生，CCF会员，主要研究方向：自然语言处理、事件检测
郭军军（1987—），男，山西吕梁人，副教授，博士，主要研究方向：自然语言处理、机器翻译
高盛祥（1977—），女，云南大理人，副教授，博士，主要研究方向：自然语言处理、机器翻译
余正涛（1970—），男，云南曲靖人，教授，博士，主要研究方向：自然语言处理、机器翻译、信息检索。
基金资助:
国家自然科学基金资助项目(61762056);国家重点研发计划项目(2018YFC0830105);云南省高新技术产业专项(201606);云南省重大科技专项计划项目(202002AD080001-5);云南省基础研究计划项目(202001AS070014)

Event detection without trigger words incorporating syntactic information

Cui WANG¹^,², Yafei ZHANG¹^,²(), Junjun GUO¹^,², Shengxiang GAO¹^,², Zhengtao YU¹^,²

^1.Faculty of Information Engineering and Automation，Kunming University of Science and Technology，Kunming Yunnan 650504，China
^2.Yunnan Key Laboratory of Artificial Intelligence （Kunming University of Science and Technology），Kunming Yunnan 650504，China

Received:2021-05-12 Revised:2021-06-17 Accepted:2021-07-04 Online:2021-12-28 Published:2021-12-10
Contact: Yafei ZHANG
About author:WANG Cui， born in 1997， M. S. candidate. Her research interests include natural language processing， event detection.
GUO Junjun， Born in 1987， Ph. D.， associate professor. His research interests include natural language processing， machine translation.
GAO Shengxiang， born in 1977， Ph. D.， associate professor. Her research interests include natural language processing， machine translation.
YU Zhengtao， born in 1970， Ph. D.， professor. His research interests include natural language processing， machine translation， information retrieval.
Supported by:
the National Natural Science Foundation of China(61762056);the National Key Research and Development Program of China(2018YFC0830105);the Special Project of Yunnan High-Tech Industry(201606);the Yunnan Major Science and Technology Program(202002AD080001-5);the Yunnan Basic Research Program(202001AS070014)

摘要/Abstract

摘要：

事件检测（ED）是信息抽取领域中最重要的任务之一，旨在识别文本中特定事件类型的实例。现有的ED方法通常采用邻接矩阵来表示句法依存关系，然而邻接矩阵往往需要借助图卷积网络（GCN）进行编码来获取句法信息，由此增加了模型的复杂度。为此，提出了融合句法信息的无触发词事件检测方法。通过将依赖父词及其上下文转换为位置标记向量，并在模型源端以无参数的方式融入依赖子词的单词嵌入来加强上下文的语义表征，而不需要经过GCN进行编码；此外，针对触发词的标注费时费力的问题，设计了基于多头注意力机制的类型感知器，以对句子中潜在的触发词进行建模，实现无触发词的事件检测。为了验证所提方法的性能，在ACE2005数据集以及低资源越南语数据集上进行了实验。其中，在ACE2005数据集上与图变换网络事件检测（GTN-ED）方法相比，所提方法的F1值提升了3.7%；在越南语数据集上，与二分类的方法类型感知偏差注意机制神经网络（TBNNAM）相比，所提方法的F1值提升了9%。结果表明，通过在Transformer中融入句法信息能有效地连接句子中分散的事件信息来提高事件检测的准确性。

关键词: 事件检测, 句法信息, 无参数, 无触发词, 类型感知器

Abstract:

Event Detection （ED） is one of the most important tasks in the field of information extraction， aiming to identify instances of specific event types in text. Existing ED methods usually use adjacency matrix to express syntactic dependencies， however， the adjacency matrix often needs to be encoded with Graph Convolutional Network （GCN） to obtain syntactic information， which increases the complexity of the model. Therefore， an event detection method without trigger words incorporating syntactic information was proposed. After converting the dependent parent word and its context into a position marker vector， the word embedding of dependent sub-word was incorporated at the source end of the model in a parameter-free manner to strengthen the semantic representation of the context， without the need of GCN for encoding. In addition， for the time-consuming and laborious labeling of trigger words， a type perceptron based on the multi-head attention mechanism was designed， which was able to model the potential trigger words in the sentence to complete the event detection without trigger words. In order to verify the performance of the proposed method， experiments were conducted on the ACE2005 dataset and the low-resource Vietnamese dataset. Compared with the Event Detection Using Graph Transformer Network （GTN-ED） method， the F1-score of the proposed method was increased by 3.7% on the ACE2005 dataset； compared with the binary classification method Type-aware Bias Neural Network with Attention Mechanisms （TBNNAM）， the F1-score of the proposed method was increased by 9% on the Vietnamese dataset. The results show that the integration of syntactic information into Transformer can effectively connect the scattered event information in the sentence to improve the accuracy of event detection.

Key words: Event Detection (ED), syntactic information, parameter-free, no trigger word, type perceptron

中图分类号:

TP391

汪翠, 张亚飞, 郭军军, 高盛祥, 余正涛. 融合句法信息的无触发词事件检测方法[J]. 计算机应用, 2021, 41(12): 3534-3539.

Cui WANG, Yafei ZHANG, Junjun GUO, Shengxiang GAO, Zhengtao YU. Event detection without trigger words incorporating syntactic information[J]. Journal of Computer Applications, 2021, 41(12): 3534-3539.

图/表 10

图1 融合句法信息的无触发词事件检测模型的基本架构

Fig. 1 Basic structure of no trigger word event detection model incorporating syntactic information

图2 句法信息处理实例

Fig. 2 Examples of syntactic information processing

图3 依赖缩放自注意力（PASCAL）结构

Fig. 3 Structure of Parent-Scaled Self-Attention （PASCAL）

表1 事件句s的二分类举例

Tab. 1 Example of binary classifications of sentence s

实例	标签
〈s， t₁〉	1
〈s， t₂〉	0
〈s， t₃〉	1

图4 越南语数据集标注案例（无触发词）

Fig. 4 Examples of labeling of Vietnamese dataset （no trigger words）

表2 模型超参数

Tab. 2 Model hyperparameters

参数	值
词向量/类型向量维数	1 024
Transformer层数	12
多头注意力机制头数（Transformer）	8
多头注意力机制头数（类型感知器）	8
Batch大小	32
句子最大长度（ACE）	50
句子最大长度（越南语）	100
Dropout	0.3
Adam权重衰减率	1E-4
学习率	0.003

表3 事件检测实验结果 ( %)

Tab. 3 Experimental results of event detection

模型	P	R	F₁
DMCNN	75.7	66.0	70.5
JRNN	66.0	73.0	69.3
HNN	84.6	64.9	73.4
TBNNAM	76.2	64.5	69.9
MOGANED	79.5	72.3	75.7
Bi-LSTM+GCN	75.5	72.4	73.9
MH-GCN	77.5	75.2	76.3
GTN-ED	80.9	73.2	76.8
本文模型	78.5	82.7	80.5

表4 越南语数据集上的实验结果 ( %)

Tab. 4 Experimental results on Vietnamese dataset

模型	P	R	F₁
TBNNAM	55.8	51.1	53.8
本文模型（越南语）	63.7	62.0	62.8

表5 不设置PASCAL以及在不同Transformer层中设置PASCAL的实验结果 ( %)

Tab. 5 Experimental results of not setting PASCAL and setting PASCAL in different Transformer layers

模型	P	R	F₁
-1_PASCAL	76.1	79.5	77.7
1_PASCAL	78.5	82.7	80.5
2_PASCAL	78.4	80.6	79.4
3_PASCAL	76.5	79.3	77.8
4_PASCAL	74.5	78.4	76.4
5_PASCAL	73.8	75.1	74.4

图5 在12层Transformer的第m层中融入句法信息的F1值

Fig. 5 F1 score incorporating syntactic information into m-layer of the 12-layer Transformer

参考文献 15

1	CHEN Y B， XU L H， LIU K， et al. Event extraction via dynamic multi-pooling convolutional neural networks［C］// Proceedings of the 53rd Annual Meeting of the Association for Computational Linguistics/ the 7th International Joint Conference on Natural Language Processing （Volume 1： Long Papers）. Stroudsburg， PA： Association for Computational Linguistics， 2015：167-176. 10.3115/v1/p15-1017
2	NGUYEN T H， CHO K， GRISHMAN R ，et al. Joint event extraction via recurrent neural networks［C］// Proceedings of the 2016 Conference of the North American Chapter of the Association for Computational Linguistics： Human Language Technologies. Stroudsburg， PA： Association for Computational Linguistics， 2016：300-309. 10.18653/v1/n16-1034
3	FENG X C， HUANG L F， TANG D Y， et al. A language-independent neural network for event detection［C］ // Proceedings of the 54th Annual Meeting of the Association for Computational Linguistics （Volume 2： Short Papers）. Stroudsburg， PA： Association for Computational Linguistics， 2016：66-71. 10.18653/v1/p16-2011
4	YAN H R， JIN X L， MENG X B， et al. Event detection with multi-order graph convolution and aggregated attention［C］// Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing /the 9th International Joint Conference on Natural Language Processing. Stroudsburg， PA： Association for Computational Linguistics， 2019：5766-5770. 10.18653/v1/d19-1582
5	陈佳丽，洪宇，王捷，等. 利用门控机制融合依存与语义信息的事件检测方法［J］. 中文信息学报， 2020， 34（8）：51-60. 10.3969/j.issn.1003-0077.2020.08.007
	CHEN J L， HONG Y， WANG J， et al. Combination of dependency and semantic information via gated mechanism for event detection［J］. Journal of Chinese Information Processing， 2020， 34（8）：51-60. 10.3969/j.issn.1003-0077.2020.08.007
6	LI L， JIN L， ZHANG Z Q， et al. Graph convolution over multiple latent context-aware graph structures for event detection［J］. IEEE Access， 2020， 8： 171435-171446. 10.1109/access.2020.3024872
7	BUGLIARELLO E， OKAZAKI N. Enhancing machine translation with-dependency aware self-attention［C］// Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics. Stroudsburg， PA： Association for Computational Linguistics， 2020： 1618-1627. 10.18653/v1/2020.acl-main.147
8	LI X， NGUYEN T H， CAO K， et al. Improving event detection with abstract meaning representation［C］// Proceedings of the 1st Workshop on Computing News Storylines. Stroudsburg， PA： Association for Computational Linguistics， 2015：11-15. 10.18653/v1/w15-4502
9	DUTTA S， MA L， SAHA T K， et al. GTN-ED： event detection using graph transformer networks［C］// Proceedings of the 15th Workshop on Graph-Based Natural Language Processing. Stroudsburg， PA： Association for Computational Linguistics， 2021：132-137. 10.18653/v1/2021.textgraphs-1.13
10	YUN S， JEONG M， KIM R， et al. Graph transformer networks［C/OL］// Proceedings of the 33rd Conference on Neural Information Processing Systems. ［2021-02-20］..
11	LIU S L， LI Y， ZHOU X P， et al. Event detection without triggers［C］// Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics： Human Language Technologies， Volume 1 （Long and Short Papers）. Stroudsburg， PA： Association for Computational Linguistics， 2019：735-744.
12	LIU J， CHEN Y B， LIU K， et al. Event detection via gated multilingual attention mechanism［C］// Proceedings of the 32nd AAAI Conference on Artificial Intelligence. Palo Alto， CA： AAAI Press， 2018：4865-4872. 10.1609/aaai.v33i01.33016754
13	WANG X Z， HAN X， LIU Z Y， et al. Adversarial training for weakly supervised event detection［C］// Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics： Human Language Technologies， Volume 1 （Long and Short Papers）. Stroudsburg， PA： Association for Computational Linguistics， 2019：998-1008.
14	HUANG L F， JI H， CHO K H， et al. Zero-shot transfer learning for event extraction［C］// Proceedings of the 56th Annual Meeting of the Association for Computational Linguistics. Stroudsburg， PA： Association for Computational Linguistics， 2018：2160-2170. 10.18653/v1/p18-1201
15	RAGANATO A， TIEDEMANN J. An analysis of encoder representations in transformer based machine translation［C］// Proceedings of the 2018 EMNLP Workshop BlackboxNLP： Analyzing and Interpreting Neural Networks for NLP. Stroudsburg， PA： Association for Computational Linguistics， 2018：287-297. 10.18653/v1/w18-5431

[1]	李衍志, 范勇, 高琳. 基于形态流的石油钻井水流异常检测[J]. 计算机应用, 2021, 41(6): 1842-1848.
[2]	邱保志, 程栾. 基于拉普拉斯中心性和密度峰值的无参数聚类算法[J]. 计算机应用, 2018, 38(9): 2511-2514.
[3]	毛莺池, 齐海, 接青, 王龙宝. M-TAEDA:多变量水质参数时序数据异常事件检测算法[J]. 计算机应用, 2017, 37(1): 138-144.
[4]	潘磊, 周欢, 王明辉. 适用于密集人群的异常事件实时检测方法[J]. 计算机应用, 2016, 36(6): 1719-1723.
[5]	林玉娥, 陈静逸, 许光宇, 梁兴柱. 完备的无参数近邻保持及最大化非近邻算法[J]. 计算机应用, 2015, 35(8): 2244-2248.
[6]	费绍栋, 杨玉珍, 刘培玉, 王健. 融合情感过滤的突发事件检测方法[J]. 计算机应用, 2015, 35(5): 1320-1323.
[7]	毛莺池, 接青, 陈豪. 河网中具有时空关系的异常事件在线检测[J]. 计算机应用, 2015, 35(11): 3106-3111.
[8]	冯永韩楠贾东风. 云计算环境下基于代表点增量层次密度聚类的微博事件检测及跟踪[J]. 计算机应用, 2013, 33(12): 3559-3562.
[9]	陈波毛剑琳乔冠华戴宁. 改进的基于统计学的滑动窗口无参数的累积和算法[J]. 计算机应用, 2013, 33(01): 88-91.
[10]	任梅詹永照潘道远孙佳瑶. 基于概率超图的视频事件语义检测[J]. 计算机应用, 2012, 32(11): 3014-3017.
[11]	汤一平胡飞虎. 基于全方位计算机视觉的盗窃事件检测[J]. 计算机应用, 2010, 30(1): 36-40.
[12]	汪勤黄山张洪斌杨权张建军. 基于视频图像处理的交通事件检测系统[J]. 计算机应用, 2008, 28(7): 1886-1889.
[13]	薛晓飞张永奎任晓东. 基于新闻要素的新事件检测方法研究[J]. 计算机应用, 2008, 28(11): 2975-2977.
[14]	党长青张景辉沈志远. 基于Boosting算法和RBF神经网络的交通事件检测[J]. 计算机应用, 2007, (12): 3105-3107.
[15]	赵宇海，王国仁，印莹. 一种用于基因表达数据的无参数聚类算法[J]. 计算机应用, 2005, 25(06): 1388-1391.

融合句法信息的无触发词事件检测方法

Event detection without trigger words incorporating syntactic information

RichHTML

PDF

可视化

摘要/Abstract

引用本文

使用本文

图/表 10

参考文献 15

相关文章 15

编辑推荐

Metrics