Document-level relation extraction via dual attention fusion and dynamic asymmetric loss
Abstract Document-level relation extraction (RE), which requires integrating and reasoning information to identify multiple possible relations among entities. However, previous research typically performed reasoning on heterogeneous graphs and set a global threshold for multiple relations classifica...
Saved in:
| Main Authors: | , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Springer
2024-11-01
|
| Series: | Complex & Intelligent Systems |
| Subjects: | |
| Online Access: | https://doi.org/10.1007/s40747-024-01632-8 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1832571200027492352 |
|---|---|
| author | Xiaoyao Ding Dongyan Ding Gang Zhou Jicang Lu Taojie Zhu |
| author_facet | Xiaoyao Ding Dongyan Ding Gang Zhou Jicang Lu Taojie Zhu |
| author_sort | Xiaoyao Ding |
| collection | DOAJ |
| description | Abstract Document-level relation extraction (RE), which requires integrating and reasoning information to identify multiple possible relations among entities. However, previous research typically performed reasoning on heterogeneous graphs and set a global threshold for multiple relations classification, regardless of interaction reasoning information among multiple relations and positive–negative samples imbalance on databases. This paper proposes a novel framework for Document-level RE with two techniques, dual attention fusion and dynamic asymmetric loss. Concretely, to obtain more interdependency feature learning, we construct entity pairs and contextual matrixes using multi-head axial attention and co-attention mechanism to learn the interaction among entity pairs deeply. To alleviate the hard-thresholds influence from positive–negative imbalance samples, we dynamically adjust weights to optimize the probabilities of different labels. We evaluate our model on two benchmark document-level RE datasets, DocRED and CDR. Experimental results show that our DASL (Dual Attention fusion and dynamic aSymmetric Loss) obtains superior performance on two public datasets, we further provide extensive experiments to analyze how dual attention fusion and dynamic asymmetric loss guide the model for better extracting multi-label relations among entities. |
| format | Article |
| id | doaj-art-d236cbfc53d34c67a13a168364e20e08 |
| institution | Kabale University |
| issn | 2199-4536 2198-6053 |
| language | English |
| publishDate | 2024-11-01 |
| publisher | Springer |
| record_format | Article |
| series | Complex & Intelligent Systems |
| spelling | doaj-art-d236cbfc53d34c67a13a168364e20e082025-02-02T12:49:11ZengSpringerComplex & Intelligent Systems2199-45362198-60532024-11-0111111210.1007/s40747-024-01632-8Document-level relation extraction via dual attention fusion and dynamic asymmetric lossXiaoyao Ding0Dongyan Ding1Gang Zhou2Jicang Lu3Taojie Zhu4Henan Open UniversityHenan Open UniversityPLA Strategic Support Force Information Engineering UniversityPLA Strategic Support Force Information Engineering UniversityPLA Strategic Support Force Information Engineering UniversityAbstract Document-level relation extraction (RE), which requires integrating and reasoning information to identify multiple possible relations among entities. However, previous research typically performed reasoning on heterogeneous graphs and set a global threshold for multiple relations classification, regardless of interaction reasoning information among multiple relations and positive–negative samples imbalance on databases. This paper proposes a novel framework for Document-level RE with two techniques, dual attention fusion and dynamic asymmetric loss. Concretely, to obtain more interdependency feature learning, we construct entity pairs and contextual matrixes using multi-head axial attention and co-attention mechanism to learn the interaction among entity pairs deeply. To alleviate the hard-thresholds influence from positive–negative imbalance samples, we dynamically adjust weights to optimize the probabilities of different labels. We evaluate our model on two benchmark document-level RE datasets, DocRED and CDR. Experimental results show that our DASL (Dual Attention fusion and dynamic aSymmetric Loss) obtains superior performance on two public datasets, we further provide extensive experiments to analyze how dual attention fusion and dynamic asymmetric loss guide the model for better extracting multi-label relations among entities.https://doi.org/10.1007/s40747-024-01632-8Relation extractionDocument-levelDual attention fusionDynamic asymmetric loss |
| spellingShingle | Xiaoyao Ding Dongyan Ding Gang Zhou Jicang Lu Taojie Zhu Document-level relation extraction via dual attention fusion and dynamic asymmetric loss Complex & Intelligent Systems Relation extraction Document-level Dual attention fusion Dynamic asymmetric loss |
| title | Document-level relation extraction via dual attention fusion and dynamic asymmetric loss |
| title_full | Document-level relation extraction via dual attention fusion and dynamic asymmetric loss |
| title_fullStr | Document-level relation extraction via dual attention fusion and dynamic asymmetric loss |
| title_full_unstemmed | Document-level relation extraction via dual attention fusion and dynamic asymmetric loss |
| title_short | Document-level relation extraction via dual attention fusion and dynamic asymmetric loss |
| title_sort | document level relation extraction via dual attention fusion and dynamic asymmetric loss |
| topic | Relation extraction Document-level Dual attention fusion Dynamic asymmetric loss |
| url | https://doi.org/10.1007/s40747-024-01632-8 |
| work_keys_str_mv | AT xiaoyaoding documentlevelrelationextractionviadualattentionfusionanddynamicasymmetricloss AT dongyanding documentlevelrelationextractionviadualattentionfusionanddynamicasymmetricloss AT gangzhou documentlevelrelationextractionviadualattentionfusionanddynamicasymmetricloss AT jicanglu documentlevelrelationextractionviadualattentionfusionanddynamicasymmetricloss AT taojiezhu documentlevelrelationextractionviadualattentionfusionanddynamicasymmetricloss |