《计算机应用》唯一官方网站 ›› 2023, Vol. 43 ›› Issue (5): 1454-1460.DOI: 10.11772/j.issn.1001-9081.2022040502
陈林颖1,2, 刘建华1,2(), 孙水华1,2, 郑智雄1,2, 林鸿辉1,2, 林杰1,2
收稿日期:
2022-04-19
修回日期:
2022-06-06
接受日期:
2022-06-09
发布日期:
2022-07-01
出版日期:
2023-05-10
通讯作者:
刘建华
作者简介:
陈林颖(1999—),女,福建莆田人,硕士研究生,CCF会员,主要研究方向:自然语言处理基金资助:
Linying CHEN1,2, Jianhua LIU1,2(), Shuihua SUN1,2, Zhixiong ZHENG1,2, Honghui LIN1,2, Jie LIN1,2
Received:
2022-04-19
Revised:
2022-06-06
Accepted:
2022-06-09
Online:
2022-07-01
Published:
2023-05-10
Contact:
Jianhua LIU
About author:
CHEN Linying, born in 1999, M. S. candidate. Her research interests include natural language processing.Supported by:
摘要:
面向方面的细粒度意见提取(AFOE)以意见对的形式从评论中提取方面词和意见词,或在此基础上再提取方面词的情感极性形成意见三元组。针对现有研究方法忽略了意见对与上下文相关性的问题,提出一种面向方面的自适应跨度特征的网格标记方案(ASF-GTS)模型。首先,利用BERT(Bidirectional Encode Representation from Transformers)模型获得句子的特征表示;然后,采用自适应跨度特征(ASF)方法加强意见对与局部上下文的联系;其次,通过网格标记方案(GTS)将意见对提取(OPE)转化为统一的网格标记任务;最后,使用特定的解码策略生成对应的意见对或意见三元组。在适用于意见元组提取任务的四个AFOE基准数据集上进行实验,结果表明,与GTS-BERT(Grid Tagging Scheme-BERT)模型相比,所提模型在意见对和意见三元组任务上的F1值分别提高了2.42%~7.30%和2.62%~6.61%。所提模型能够有效保留意见对与上下文的情感联系,更精确地提取意见对及其情感极性。
中图分类号:
陈林颖, 刘建华, 孙水华, 郑智雄, 林鸿辉, 林杰. 面向方面的自适应跨度特征的细粒度意见元组提取[J]. 计算机应用, 2023, 43(5): 1454-1460.
Linying CHEN, Jianhua LIU, Shuihua SUN, Zhixiong ZHENG, Honghui LIN, Jie LIN. Aspect-oriented fine-grained opinion tuple extraction with adaptive span features[J]. Journal of Computer Applications, 2023, 43(5): 1454-1460.
数据集 | 数据划分 | Sen | Asp | Opi | Pai | Tri |
---|---|---|---|---|---|---|
14Res | 训练集 | 1 259 | 2 064 | 2 098 | 2 356 | 2 356 |
验证集 | 315 | 487 | 506 | 580 | 580 | |
测试集 | 493 | 851 | 866 | 1 008 | 1 008 | |
14Lap | 训练集 | 899 | 1 257 | 1 270 | 1 452 | 1 452 |
验证集 | 225 | 332 | 313 | 383 | 383 | |
测试集 | 332 | 467 | 478 | 547 | 547 | |
15Res | 训练集 | 603 | 871 | 966 | 1 038 | 1 038 |
验证集 | 151 | 205 | 226 | 239 | 239 | |
测试集 | 325 | 436 | 469 | 493 | 493 | |
16Res | 训练集 | 863 | 1 213 | 1 329 | 1 421 | 1 421 |
验证集 | 216 | 298 | 331 | 348 | 348 | |
测试集 | 328 | 456 | 485 | 525 | 525 |
表1 AFOE数据集
Tab. 1 AFOE datasets
数据集 | 数据划分 | Sen | Asp | Opi | Pai | Tri |
---|---|---|---|---|---|---|
14Res | 训练集 | 1 259 | 2 064 | 2 098 | 2 356 | 2 356 |
验证集 | 315 | 487 | 506 | 580 | 580 | |
测试集 | 493 | 851 | 866 | 1 008 | 1 008 | |
14Lap | 训练集 | 899 | 1 257 | 1 270 | 1 452 | 1 452 |
验证集 | 225 | 332 | 313 | 383 | 383 | |
测试集 | 332 | 467 | 478 | 547 | 547 | |
15Res | 训练集 | 603 | 871 | 966 | 1 038 | 1 038 |
验证集 | 151 | 205 | 226 | 239 | 239 | |
测试集 | 325 | 436 | 469 | 493 | 493 | |
16Res | 训练集 | 863 | 1 213 | 1 329 | 1 421 | 1 421 |
验证集 | 216 | 298 | 331 | 348 | 348 | |
测试集 | 328 | 456 | 485 | 525 | 525 |
提取方式 | 模型 | 14Res | 14Lap | 15Res | 16Res | ||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|---|
P | R | F1 | P | R | F1 | P | R | F1 | P | R | F1 | ||
管道方式 | BiLSTM-ATT+IOG[ | 69.99 | 61.58 | 65.46 | 64.93 | 44.56 | 52.84 | 59.14 | 56.38 | 57.73 | 66.07 | 62.55 | 64.13 |
DE-CNN+IOG[ | 67.70 | 69.41 | 68.55 | 59.59 | 51.68 | 55.35 | 59.18 | 60.08 | 58.04 | 62.97 | 66.22 | 64.55 | |
RINANTE+IOG[ | 70.16 | 65.47 | 67.74 | 61.76 | 53.11 | 57.10 | 63.24 | 55.57 | 59.16 | ||||
统一提取 | GTS-BERT[ | 75.95 | 70.81 | 73.29 | 66.15 | 63.11 | 64.60 | 66.40 | 68.71 | 67.53 | 72.25 | 77.41 | 74.74 |
本文模型 | ASF-GTS | 81.86 | 75.66 | 78.64 | 72.01 | 64.22 | 67.90 | 78.79 | 63.80 | 70.51 | 75.78 | 77.33 | 76.55 |
表2 OPE任务提取结果 ( %)
Tab. 2 Extraction results of OPE tasks
提取方式 | 模型 | 14Res | 14Lap | 15Res | 16Res | ||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|---|
P | R | F1 | P | R | F1 | P | R | F1 | P | R | F1 | ||
管道方式 | BiLSTM-ATT+IOG[ | 69.99 | 61.58 | 65.46 | 64.93 | 44.56 | 52.84 | 59.14 | 56.38 | 57.73 | 66.07 | 62.55 | 64.13 |
DE-CNN+IOG[ | 67.70 | 69.41 | 68.55 | 59.59 | 51.68 | 55.35 | 59.18 | 60.08 | 58.04 | 62.97 | 66.22 | 64.55 | |
RINANTE+IOG[ | 70.16 | 65.47 | 67.74 | 61.76 | 53.11 | 57.10 | 63.24 | 55.57 | 59.16 | ||||
统一提取 | GTS-BERT[ | 75.95 | 70.81 | 73.29 | 66.15 | 63.11 | 64.60 | 66.40 | 68.71 | 67.53 | 72.25 | 77.41 | 74.74 |
本文模型 | ASF-GTS | 81.86 | 75.66 | 78.64 | 72.01 | 64.22 | 67.90 | 78.79 | 63.80 | 70.51 | 75.78 | 77.33 | 76.55 |
模型 | 14Res | 14Lap | 15Res | 16Res | ||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|
P | R | F1 | P | R | F1 | P | R | F1 | P | R | F1 | |
Peng-unified-R+IOG | 58.89 | 60.41 | 59.64 | 48.62 | 45.52 | 47.02 | 51.70 | 46.04 | 48.71 | 59.25 | 58.09 | 58.67 |
IMN+IOG | 59.57 | 63.88 | 61.65 | 49.21 | 46.23 | 47.68 | 55.24 | 52.33 | 53.75 | |||
GTS-BERT | 70.92 | 69.49 | 70.20 | 57.52 | 51.91 | 54.58 | 59.29 | 58.07 | 58.67 | 63.95 | 70.85 | 67.22 |
ASF-GTS | 75.62 | 70.81 | 73.13 | 60.66 | 53.76 | 56.91 | 65.19 | 60.12 | 62.55 | 67.03 | 71.04 | 68.98 |
表3 OTE任务提取结果 ( %)
Tab. 3 OTE task extraction results
模型 | 14Res | 14Lap | 15Res | 16Res | ||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|
P | R | F1 | P | R | F1 | P | R | F1 | P | R | F1 | |
Peng-unified-R+IOG | 58.89 | 60.41 | 59.64 | 48.62 | 45.52 | 47.02 | 51.70 | 46.04 | 48.71 | 59.25 | 58.09 | 58.67 |
IMN+IOG | 59.57 | 63.88 | 61.65 | 49.21 | 46.23 | 47.68 | 55.24 | 52.33 | 53.75 | |||
GTS-BERT | 70.92 | 69.49 | 70.20 | 57.52 | 51.91 | 54.58 | 59.29 | 58.07 | 58.67 | 63.95 | 70.85 | 67.22 |
ASF-GTS | 75.62 | 70.81 | 73.13 | 60.66 | 53.76 | 56.91 | 65.19 | 60.12 | 62.55 | 67.03 | 71.04 | 68.98 |
模型 | 14Res | 14Lap | 15Res | 16Res | ||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|
P | R | F1 | P | R | F1 | P | R | F1 | P | R | F1 | |
GTS-BERT | 75.95 | 70.81 | 73.29 | 66.15 | 63.11 | 64.60 | 66.40 | 68.71 | 67.53 | 72.25 | 77.41 | 74.74 |
GTS-BERT+CDM | 73.05 | 73.64 | 73.34 | 68.51 | 59.08 | 63.45 | 68.89 | 63.40 | 66.02 | 68.18 | 72.39 | 70.22 |
ASF-GTS | 81.86 | 75.66 | 78.64 | 72.01 | 64.22 | 67.90 | 78.79 | 63.80 | 70.51 | 75.78 | 77.33 | 76.55 |
表4 OPE任务适应性结果 ( %)
Tab. 4 OPE task adaptability results
模型 | 14Res | 14Lap | 15Res | 16Res | ||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|
P | R | F1 | P | R | F1 | P | R | F1 | P | R | F1 | |
GTS-BERT | 75.95 | 70.81 | 73.29 | 66.15 | 63.11 | 64.60 | 66.40 | 68.71 | 67.53 | 72.25 | 77.41 | 74.74 |
GTS-BERT+CDM | 73.05 | 73.64 | 73.34 | 68.51 | 59.08 | 63.45 | 68.89 | 63.40 | 66.02 | 68.18 | 72.39 | 70.22 |
ASF-GTS | 81.86 | 75.66 | 78.64 | 72.01 | 64.22 | 67.90 | 78.79 | 63.80 | 70.51 | 75.78 | 77.33 | 76.55 |
模型 | 不同样本预测的结果 | ||
---|---|---|---|
样本1: The avocado salad is a personal fave. | 样本2: Montparnasse's desserts — especially the silken creme brulee and paper — thin apple tart — are good enough on their own to make the restaurant worth the trip. | 样本3: menu-uneventful, small | |
Ground Truth (GT) | (avocado salad-fave-positive) | (desserts-good-positive) (crème brulee-silken-positive) (apple tart-thin-positive) | (menu-uneventful-negative) (menu-small-negative) |
GTS-BERT | (avocado salad-fave-positive) √ | (apple tart-good-positive) × (desserts-good-positive) √ (crème brulee-good-positive) × | (menu-uneventful-negative) × (menu-small-neutral) × |
GTS-BERT+CDM | (NULL-NULL-NULL) | (apple tart-good-positive) × (crème brulee-good-positive)× | (menu-uneventful-positive) × |
ASF-GTS | (avocado salad-fave-positive) √ | (apple tart-good-positive)× (desserts-good-positive) √ (crème brulee-good-positive) × | (menu-uneventful-negative) √ (menu-small-negative) √ |
表5 三个样本的预测结果
Tab. 5 Prediction results of three examples
模型 | 不同样本预测的结果 | ||
---|---|---|---|
样本1: The avocado salad is a personal fave. | 样本2: Montparnasse's desserts — especially the silken creme brulee and paper — thin apple tart — are good enough on their own to make the restaurant worth the trip. | 样本3: menu-uneventful, small | |
Ground Truth (GT) | (avocado salad-fave-positive) | (desserts-good-positive) (crème brulee-silken-positive) (apple tart-thin-positive) | (menu-uneventful-negative) (menu-small-negative) |
GTS-BERT | (avocado salad-fave-positive) √ | (apple tart-good-positive) × (desserts-good-positive) √ (crème brulee-good-positive) × | (menu-uneventful-negative) × (menu-small-neutral) × |
GTS-BERT+CDM | (NULL-NULL-NULL) | (apple tart-good-positive) × (crème brulee-good-positive)× | (menu-uneventful-positive) × |
ASF-GTS | (avocado salad-fave-positive) √ | (apple tart-good-positive)× (desserts-good-positive) √ (crème brulee-good-positive) × | (menu-uneventful-negative) √ (menu-small-negative) √ |
1 | WU Z, YING C C, ZHAO F, et al. Grid tagging scheme for aspect-oriented fine-grained opinion extraction[C]// Proceedings of the Findings of the Association for Computational Linguistics: EMNLP 2020. Stroudsburg, PA: ACL, 2020: 2576-2585. 10.18653/v1/2020.findings-emnlp.234 |
2 | LIU B. Sentiment Analysis and Opinion Mining, SLHLT[M]. Cham: Springer, 2012: 1-167. 10.2200/s00416ed1v01y201204hlt016 |
3 | PANG B, LEE L. Opinion mining and sentiment analysis[J]. Foundations and Trends in Information Retrieval, 2008, 2(1/2): 1-135. 10.1561/1500000011 |
4 | WANG W Y, PAN S J, DAHLMEIER D, et al. Recursive neural conditional random fields for aspect-based sentiment analysis[C]// Proceedings of the 2016 Conference on Empirical Methods in Natural Language Processing. Stroudsburg, PA: ACL, 2016: 616-626. 10.18653/v1/d16-1059 |
5 | YU J F, JIANG J, XIA R. Global inference for aspect and opinion terms co-extraction based on multi-task neural networks[J]. IEEE/ACM Transactions on Audio, Speech, and Language Processing, 2019, 27(1): 168-177. 10.1109/taslp.2018.2875170 |
6 | DAI H L, SONG Y Q. Neural aspect and opinion term extraction with mined rules as weak supervision[C]// Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics. Stroudsburg, PA: ACL, 2019: 5268-5277. 10.18653/v1/p19-1520 |
7 | LI Y C, WANG F, ZHANG W J, et al. A more fine-grained aspect-sentiment-opinion triplet extraction task[EB/OL]. (2021-08-29) [2022-05-29].. |
8 | DEVLIN J, CHANG M W, LEE K, et al. BERT: pre-training of deep bidirectional transformers for language understanding[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: ACL, 2019: 4171-4186. 10.18653/v1/n18-2 |
9 | MUKHERJEE R, NAYAK T, BUTALA Y, et al. PASTE: a tagging-free decoding framework using pointer networks for aspect sentiment triplet extraction[C]// Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing. Stroudsburg, PA: ACL, 2021: 9279-9291. 10.18653/v1/2021.emnlp-main.731 |
10 | 夏鸿斌,李强,肖奕飞. 用于方面情感三元组抽取的词对关系学习方法[J]. 模式识别与人工智能, 2022, 35(3):262-270. 10.16451/j.cnki.issn1003-6059.202203006 |
XIA H B, LI Q, XIAO Y F. Word-pair relation learning method for aspect sentiment triplet extraction[J]. Pattern Recognition and Artificial Intelligence, 2022, 35(3):262-270. 10.16451/j.cnki.issn1003-6059.202203006 | |
11 | ZENG B Q, YANG H, XU R Y, et al. LCF: a local context focus mechanism for aspect-based sentiment classification[J]. Applied Sciences, 2019, 9(16): No.3389. 10.3390/app9163389 |
12 | YANG H, ZENG B, YANG J H, et al. A multi-task learning model for Chinese-oriented aspect polarity classification and aspect term extraction[J]. Neurocomputing, 2021, 419: 344-356. 10.1016/j.neucom.2020.08.001 |
13 | MIKOLOV T, SUTSKEVER I, CHEN K, et al. Distributed representations of words and phrases and their compositionality[C]// Proceedings of the 26th International Conference on Neural Information Processing Systems — Volume 2. Red Hook, NY: Curran Associates Inc., 2013: 3111-3119. |
14 | PENNINGTON J, SOCHER R, MANNING C D. GloVe: global vectors for word representation[C]// Proceedings of the 2014 Conference on Empirical Methods in Natural Language Processing. Stroudsburg, PA: ACL, 2014: 1532-1543. 10.3115/v1/d14-1162 |
15 | VASWANI A, SHAZEER N, PARMAR N, et al. Attention is all you need[C]// Proceedings of the 31st International Conference of Neural Information Processing Systems. Red Hook, NY: Curran Associates Inc., 2017: 6000-6010. |
16 | PONTIKI M, GALANIS D, PAPAGEORGIOU H, et al. Semeval-2015 Task 12: aspect based sentiment analysis[C]// Proceedings of the 9th International Workshop on Semantic Evaluation. Stroudsburg, PA: ACL, 2015: 486-495. 10.18653/v1/s15-2082 |
17 | FAN Z F, WU Z, DAI X Y, et al. Target-oriented opinion words extraction with target-fused neural sequence labeling[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: ACL, 2019: 2509-2518. 10.18653/v1/n19-1259 |
18 | HOCHREITER S, SCHMIDHUBER J. Long short-term memory[J]. Neural Computation, 1997, 9(8): 1735-1780. 10.1162/neco.1997.9.8.1735 |
19 | XU H, LIU B, SHU L, et al. Double embeddings and CNN-based sequence labeling for aspect extraction[C]// Proceedings of the 56th Annual Meeting of the Association for Computational Linguistics (Volume 2: Short Papers). Stroudsburg, PA: ACL, 2018: 592-598. 10.18653/v1/p18-2094 |
20 | PENG H Y, XU L, BING L D, et al. Knowing what, how and why: a near complete solution for aspect-based sentiment analysis[C]// Proceedings of the 34th AAAI Conference on Artificial Intelligence. Palo Alto, CA: AAAI Press, 2020: 8600-8607. 10.1609/aaai.v34i05.6383 |
21 | HEY R D, LEE W S, NG H T, et al. An interactive multi-task learning network for end-to-end aspect-based sentiment analysis[C] // Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics. Stroudsburg, PA: ACL, 2019:504-515. 10.18653/v1/p19-1048 |
[1] | 汪锦云 向阳. 基于关键词图表示的文本语义去重算法 [J]. 《计算机应用》唯一官方网站, 0, (): 0-0. |
[2] | 林于翔 吴运兵 阴爱英 廖祥文 . 基于语义相关性分析的多模态摘要模型 [J]. 《计算机应用》唯一官方网站, 0, (): 0-0. |
[3] | 方澄, 李贝, 韩萍, 吴琼. 基于语法依存图的中文微博细粒度情感分类[J]. 《计算机应用》唯一官方网站, 2023, 43(4): 1056-1061. |
[4] | 王惠茹, 李秀红, 李哲, 马春明, 任泽裕, 杨丹. 多模态预训练模型综述[J]. 《计算机应用》唯一官方网站, 2023, 43(4): 991-1004. |
[5] | 何子仪 杨燕 张熠玲. 深度融合多视图聚类网络[J]. 《计算机应用》唯一官方网站, 0, (): 0-0. |
[6] | 王雨, 袁玉波, 过弋, 张嘉杰. 情感增强的对话文本情绪识别模型[J]. 《计算机应用》唯一官方网站, 2023, 43(3): 706-712. |
[7] | 尹春勇, 周立文. 基于再编码的无监督时间序列异常检测模型[J]. 《计算机应用》唯一官方网站, 2023, 43(3): 804-811. |
[8] | 王啸飞 鲍胜利 陈炯环. 基于潜在因子模型在子空间上的缺失值注意力聚类算法[J]. 《计算机应用》唯一官方网站, 0, (): 0-0. |
[9] | 于碧辉 蔡兴业 魏靖烜. 基于提示学习的小样本文本分类方法[J]. 《计算机应用》唯一官方网站, 0, (): 0-0. |
[10] | 吴明月 周栋 赵文玉 屈薇. 基于流形学习的句向量优化[J]. 《计算机应用》唯一官方网站, 0, (): 0-0. |
[11] | 王炫力 靳小龙 侯中妮 廖华明 张瑾. 基于森林的实体关系联合抽取模型[J]. 《计算机应用》唯一官方网站, 0, (): 0-0. |
[12] | 林呈宇, 王雷, 薛聪. 标签语义增强的弱监督文本分类模型[J]. 《计算机应用》唯一官方网站, 2023, 43(2): 335-342. |
[13] | 王奇, 雷航, 王旭鹏. 姿态干扰下的深度人脸验证[J]. 《计算机应用》唯一官方网站, 2023, 43(2): 595-600. |
[14] | 宗传玉, 宪超, 夏秀峰. 实例簇驱动的图结构聚类参数计算算法[J]. 《计算机应用》唯一官方网站, 2023, 43(2): 398-406. |
[15] | 胡婕, 陈晓茜, 张龑. 基于池化和特征组合增强BERT的答案选择模型[J]. 《计算机应用》唯一官方网站, 2023, 43(2): 365-373. |
阅读次数 | ||||||
全文 |
|
|||||
摘要 |
|
|||||