[Objective] This study proposes a new text summarization model for biomedicine research, aiming to improve the quality of their abstracts. [Methods] First, we obtained the important contents of the biomedical texts with extractive abstracting technology. Then, we combined the important contents with related knowledge base to extract the key terms and their corresponding concepts. Third, we integrated these contents and concepts to the neural network abstrcting model as background knowledge for the attention mechanism. With the help of domain knowledge, the proposed model can not only focus on the important information from the texts, but also reduce the noises occurring due to the introduction of external information. [Results] We examined the proposed model with three biomedical data sets. The average ROUGE of the proposed model’s PG-meta reached 31.06, which was 1.51 higher than the average ROUGE of the original PG model. [Limitations] We did not investigate the impacts of different knowledge acquiring methods on the effectiveness of our model. [Conclusions] The proposed model can better learn the in-depth meaning of biomedical documents and improve the quality of their abstracts.
Glaucoma is a leading cause of blindness within the United States and the leading cause of blindness among African-Americans. Measurement of intraocular pressure only is no longer considered adequate for screening. Recognition of risk factors and examination of the optic nerve are key strategies to identify individuals at risk. Medical and surgical treatment of glaucoma have ······
核心 内容
control of iop regulation resides within the aqueous outflow system of the eye ( grant , 1958 ) and iop regulation becomes abnormal in glaucoma.<q>iop is the only treatable risk factor.<q>the intrinsic outflow system abnormality inglaucoma is unknown but is described as poag······
术语 概念对
Glaucoma:Eye disease IOP:Intraocular pressure POAG:Glaucoma, Primary Open Angle
Table 1 知识获取与对应摘要的关系实例
数据集
数据集 大小
原文平均 长度(词)
摘要平均 长度(词)
原文术语概念对平均 数量(个)
标准摘要术语概念对平均 数量(个)
Full-Abs
训练集
3 200
25 825
1 072
#
10
验证集
400
24 348
903
#
9
测试集
400
24 484
936
#
9
Abs-Ti
训练集
3 514
1 477
111
9
3
验证集
439
1 439
109
9
3
测试集
439
1 466
113
9
3
BioAbsTi
训练集
24 631
1 574
118
109
4
验证集
8 210
1 584
118
109
4
测试集
8 210
1 562
117
10
4
Table 2 实验数据集统计结果
Fig.3 不同d值下的实验结果
Full-Abs
Abs-Ti
BioAbsTi
模型
R-1
R-2
R-L
R-1
R-2
R-L
R-1
R-2
R-L
AVG
Lead
27.93
15.38
24.58
23.36
13.79
24.58
31.79
16.55
28.67
22.96
TextRank
27.82
14.64
24.83
24.83
13.56
20.93
33.52
17.29
30.23
23.07
PG
35.95
20.42
29.87
33.25
18.36
31.45
36.58
25.56
34.52
29.55
Keywords-PG
36.23
20.89
30.52
33.75
19.06
31.83
36.93
26.22
34.85
30.03
PG-meta(All)
36.15
20.85
30.28
33.54
18.75
31.79
36.88
25.96
34.58
29.86
BERT+聚类
32.85
18.87
28.85
27.53
14.22
23.93
35.25
21.52
32.85
26.20
BERTSum
33.93
19.86
29.18
28.70
14.56
24.34
37.60
22.81
33.65
27.18
PG-meta
37.05
21.96
33.58
34.82
20.21
32.97
37.58
26.26
35.19
31.06
Table 3 参与比较的基线模型与PG-meta模型在三个数据集上的实验结果比较
类别
文本内容
原文
······. A recent study reported that cardiac lymphatic endothelial cells (LECs) stem from venous and non-venous origins in mice. Here, we identified Isl1-expressing progenitors as a potential non-venous origin of cardiac LECs. Genetic lineage tracing with Isl1-Cre reporter mice suggested a possible contribution from the Isl1-expressing pharyngeal mesoderm constituting the second heart field to lymphatic vessels around the cardiac outflow tract as well as to those in the facial skin and the lymph sac. Isl1(+) lineage-specific deletion of Prox1 resulted in disrupted LYVE1(+) vessel structures, indicating a Prox1-dependent mechanism in this contribution. ······
Here, we identified Isl1-expressing progenitors as a potential non-venous origin of cardiac LECs. 译文:在这里,我们确定Isl1-expressing的祖细胞是心脏LECs的潜在非静脉起源。
PG模型的摘要结果
The non-venous cell lineage can help the development of cardiac lymphatic vessels. 译文:非静脉细胞谱系可以帮助心脏淋巴管的发育。
PG-meta模型的 摘要结果
The non-venous cell lineage of Isl1-expressing promotes the development of cardiac lymphatic vessels. 译文: Isl1-expression的非静脉细胞谱系促进心脏淋巴管的发育。
Table 4 三种模型自动生成的摘要结果对比
数据集
指标
PG
PG-meta (LEAD)
PG-meta(TR)
PG-meta(BS)
Full-Abs
R-1
35.95
36.89
36.97
37.05
R-2
20.42
21.88
21.97
21.96
R-L
29.87
32.95
33.56
33.58
Abs-Ti
R-1
33.25
34.67
34.85
34.82
R-2
18.36
19.66
19.35
20.21
R-L
31.45
32.58
32.73
32.97
BioAbsTi
R-1
36.58
37.42
37.55
37.58
R-2
25.56
25.93
26.13
26.26
R-L
34.52
34.97
35.16
35.19
AVG
29.55
30.77
30.91
31.06
Table 5 不同重要内容抽取方法下的实验结果
数据集
指标
PG-meta(term)
PG-meta(con)
PG-meta(t-c)
Full-Abs
R-1
36.53
36.75
37.05
R-2
21.85
21.72
21.96
R-L
33.24
33.35
33.58
Abs-Ti
R-1
34.63
34.79
34.82
R-2
20.19
20.25
20.21
R-L
32.73
32.79
32.97
BioAbsTi
R-1
37.46
37.59
37.58
R-2
26.07
26.12
26.26
R-L
35.07
35.03
35.19
AVG
30.86
30.93
31.06
Table 6 不同知识关联粒度下的实验结果
Fig.4 不同知识融合方式实验结果
[1]
Mishra R, Bian J T, Fiszman M, et al. Text Summarization in the Biomedical Domain: A Systematic Review of Recent Research[J]. Journal of Biomedical Informatics, 2014, 52: 457-467.
doi: 10.1016/j.jbi.2014.06.009
pmid: 25016293
[2]
See A, Liu P J, Manning C D. Get to the Point: Summarization with Pointer-Generator Networks[C]// Proceedings of the 55th Annual Meeting of the Association for Computational Linguistics. 2017: 1073-1083.
(Yu Shanshan, Su Jindian, Li Pengfei. Improved TextRank-Based Method for Automatic Summarization[J]. Computer Science, 2016, 43(6): 240-247.)
doi: 10.11896/j.issn.1002-137X.2016.06.048
[5]
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. 2019: 4171-4186.
[6]
Liu Y. Fine-Tune BERT for Extractive Summarization[OL]. arXiv Preprint, arXiv: 1903.10318.
[7]
Hermann K M, Kočiský T, Grefenstette E, et al. Teaching Machines to Read and Comprehend[OL]. arXiv Preprint, arXiv: 1506.03340.
[8]
Zhou L, Hovy E. Template-Filtered Headline Summarization[C]// Proceedings of the ACL-04 Workshop:Text Summarization Branches Out. 2004: 56-60.
(Shi Lei, Ruan Xuanmin, Wei Ruibin, et al. Abstractive Summarization Based on Sequence to Sequence Models: A Review[J]. Journal of the China Society for Scientific and Technical Information, 2019, 38(10): 1102-1116.)
[10]
Rush A M, Chopra S, Weston J. A Neural Attention Model for Abstractive Sentence Summarization[C]// Proceedings of the 2015 Conference on Empirical Methods in Natural Language Processing. 2015: 379-389.
[11]
Chen T, Xu R F, He Y L, et al. Improving Sentiment Analysis via Sentence Type Classification Using BiLSTM-CRF and CNN[J]. Expert Systems with Applications, 2017, 72: 221-230.
doi: 10.1016/j.eswa.2016.10.065
[12]
Gehring J, Auli M, Grangier D, et al. Convolutional Sequence to Sequence Learning[C]// Proceedings of the 34th International Conference on Machine Learning. 2017: 1243-1252.
[13]
Cai T, Shen M J, Peng H L, et al. Improving Transformer with Sequential Context Representations for Abstractive Text Summarization[C]// Proceedings of the 8th CCF International Conference on Natural Language Processing and Chinese Computing. 2019: 512-524.
(Luo Pengcheng, Wang Yibo, Wang Jimin. Automatic Discipline Classification for Scientific Papers Based on a Deep Pre-Training Language Model[J]. Journal of the China Society for Scientific and Technical Information, 2020, 39(10): 1046-1059.)
[15]
Bhatia N, Jaiswal A. Automatic Text Summarization and It's Methods—A Review[C]// Proceedings of the 6th International Conference-Cloud System and Big Data Engineering(Confluence). IEEE, 2016: 65-72.
[16]
Nallapati R, Zhou B, Dos Santos C, et al. Abstractive Text Summarization Using Sequence-to-Sequence RNNs and Beyond[C]// Proceedings of the 20th SIGNLL Conference on Computational Natural Language Learning. 2016: 280-290.
[17]
Gu J, Lu Z, Li H, et al. Incorporating Copying Mechanism in Sequence-to-Sequence Learning[C]// Proceedings of the 54th Annual Meeting of the Association for Computational Linguistics. 2016: 1631-1640.
[18]
Tu Z, Lu Z, Liu Y, et al. Modeling Coverage for Neural Machine Translation[C]// Proceedings of the 54th Annual Meeting of the Association for Computational Linguistics. 2016: 76-85.
[19]
Jiang X P, Hu P, Hou L W, et al. Improving Pointer-Generator Network with Keywords Information for Chinese Abstractive Summarization[C]// Proceedings of the 7th CCF International Conference on Natural Language Processing and Chinese Computing. 2018: 464-474.
[20]
Nasr-Azadani M, Ghadiri N, Davoodijam E. Graph-Based Biomedical Text Summarization: An Itemset Mining and Sentence Clustering Approach[J]. Journal of Biomedical Informatics, 2018, 84: 42-58.
doi: S1532-0464(18)30111-4
pmid: 29906584
[21]
Yoo I, Hu X H, Song I Y. A Coherent Graph-Based Semantic Clustering and Summarization Approach for Biomedical Literature and a New Summarization Evaluation Method[J]. BMC Bioinformatics, 2007, 8(S9): S4.
[22]
Afzal M, Alam F, Malik K M, et al. Clinical Context-Aware Biomedical Text Summarization Using Deep Neural Network: Model Development and Validation[J]. Journal of Medical Internet Research, 2020, 22(10): e19810.
doi: 10.2196/19810
[23]
Moradi M, Dorffner G, Samwald M. Deep Contextualized Embeddings for Quantifying the Informative Content in Biomedical Text Summarization[J]. Computer Methods and Programs in Biomedicine, 2020, 184: 105117.
doi: 10.1016/j.cmpb.2019.105117
[24]
Kondadadi R, Manchanda S, Ngo J, et al.Optum at MEDIQA 2021: Abstractive Summarization of Radiology Reports Using Simple BART Finetuning[C]// Proceedings of the 20th Workshop on Biomedical Language Processing. 2021: 280-284.
[25]
Mahajan D, Tsou C H, Liang J J. IBM Research at MEDIQA 2021: Toward Improving Factual Correctness of Radiology Report Abstractive Summarization[C]// Proceedings of the 20th Workshop on Biomedical Language Processing. 2021: 302-310.
[26]
Sotudeh S, Goharian N, Filice R. Attend to Medical Ontologies: Content Selection for Clinical Abstractive Summarization[C]// Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics. 2020: 1899-1905.
[27]
Bhattacharya S, Ha-Thuc V, Srinivasan P. MeSH: A Window into Full Text for Document Summarization[J]. Bioinformatics, 2011, 27(13): i120-i128.
doi: 10.1093/bioinformatics/btr223
[28]
Plaza L, Díaz A, Gervás P. A Semantic Graph-Based Approach to Biomedical Summarisation[J]. Artificial Intelligence in Medicine, 2011, 53(1): 1-14.
doi: 10.1016/j.artmed.2011.06.005
pmid: 21752612
[29]
Bodenreider O. The Unified Medical Language System(UMLS): Integrating Biomedical Terminology[J]. Nucleic Acids Research, 2004, 32(S1): D267-D270.
doi: 10.1093/nar/gkh061
[30]
MacAvaney S, Sotudeh S, Cohan A, et al. Ontology-Aware Clinical Abstractive Summarization[C]// Proceedings of the 42nd International ACM SIGIR Conference on Research and Development in Information Retrieval. 2019: 1013-1016.
[31]
Zhang Y, Ding D Y, Qian T, et al. Learning to Summarize Radiology Findings[C]// Proceedings of the 9th International Workshop on Health Text Mining and Information Analysis. 2018: 204-213.
[32]
Mohan S, Li D. MedMentions: A Large Biomedical Corpus Annotated with UMLS Concepts[OL]. arXiv Preprint, arXiv: 1902.09476.
[33]
Du Y P, Li Q X, Wang L L, et al. Biomedical-Domain Pre-Trained Language Model for Extractive Summarization[J]. Knowledge-Based Systems, 2020, 199: 105964.
doi: 10.1016/j.knosys.2020.105964
[34]
Lin C Y. ROUGE: A Package for Automatic Evaluation of Summaries[C]// Proceedings of the 2004 Workshop on Text Summarization Branches Out. 2004: 74-81.
[35]
Lin C Y, Hovy E. Automatic Evaluation of Summaries Using N-Gram Co-Occurrence Statistics[C]// Proceedings of the 2003 Conference of the North American Chapter of the Association for Computational Linguistics on Human Language Technology. 2003: 71-78.