Due to the small size of the annotated corpora and the sparsity of the event trigger words, the event coreference resolver cannot capture enough event semantics, especially the trigger semantics, to identify coreferen...Due to the small size of the annotated corpora and the sparsity of the event trigger words, the event coreference resolver cannot capture enough event semantics, especially the trigger semantics, to identify coreferential event mentions. To address the above issues, this paper proposes a trigger semantics augmentation mechanism to boost event coreference resolution. First, this mechanism performs a trigger-oriented masking strategy to pre-train a BERT (Bidirectional Encoder Representations from Transformers)-based encoder (Trigger-BERT), which is fine-tuned on a large-scale unlabeled dataset Gigaword. Second, it combines the event semantic relations from the Trigger-BERT encoder with the event interactions from the soft-attention mechanism to resolve event coreference. Experimental results on both the KBP2016 and KBP2017 datasets show that our proposed model outperforms several state-of-the-art baselines.展开更多
As a subtask of information extraction (IE), which aims to extract structured information from texts, event extraction is to recognize event trigger mentions of a predefined event type and their arguments. In general,...As a subtask of information extraction (IE), which aims to extract structured information from texts, event extraction is to recognize event trigger mentions of a predefined event type and their arguments. In general, event extraction can be divided into two subtasks: trigger extraction and argument extraction. Currently, the frequent existences of unannotated trigger mentions and poor-context trigger mentions impose critical challenges in Chinese trigger extraction. This paper proposes a novel three-layer joint model to integrate three components in trigger extraction, i.e., trigger identification, event type determination, and event subtype determination. In this way, different kinds of evidence on distinct pseudo samples can be well captured to eliminate the harmful effects of those un-annotated trigger mentions. In addition, this paper introduces various types of linguistically driven constraints on the trigger and argument semantics into the joint model to recover those poor-context trigger mentions. The experimental results show that our joint model significantly outperforms the state-of-the-art Chinese trigger extraction and Chinese event extraction as a whole.展开更多
基金supported by the National Natural Science Foundation of China under Grant Nos.61836007 and 61772354.
文摘Due to the small size of the annotated corpora and the sparsity of the event trigger words, the event coreference resolver cannot capture enough event semantics, especially the trigger semantics, to identify coreferential event mentions. To address the above issues, this paper proposes a trigger semantics augmentation mechanism to boost event coreference resolution. First, this mechanism performs a trigger-oriented masking strategy to pre-train a BERT (Bidirectional Encoder Representations from Transformers)-based encoder (Trigger-BERT), which is fine-tuned on a large-scale unlabeled dataset Gigaword. Second, it combines the event semantic relations from the Trigger-BERT encoder with the event interactions from the soft-attention mechanism to resolve event coreference. Experimental results on both the KBP2016 and KBP2017 datasets show that our proposed model outperforms several state-of-the-art baselines.
文摘As a subtask of information extraction (IE), which aims to extract structured information from texts, event extraction is to recognize event trigger mentions of a predefined event type and their arguments. In general, event extraction can be divided into two subtasks: trigger extraction and argument extraction. Currently, the frequent existences of unannotated trigger mentions and poor-context trigger mentions impose critical challenges in Chinese trigger extraction. This paper proposes a novel three-layer joint model to integrate three components in trigger extraction, i.e., trigger identification, event type determination, and event subtype determination. In this way, different kinds of evidence on distinct pseudo samples can be well captured to eliminate the harmful effects of those un-annotated trigger mentions. In addition, this paper introduces various types of linguistically driven constraints on the trigger and argument semantics into the joint model to recover those poor-context trigger mentions. The experimental results show that our joint model significantly outperforms the state-of-the-art Chinese trigger extraction and Chinese event extraction as a whole.
