Search

Document-level event extraction aims to extract event-related information from an unstructured document composed of multiple sentences. Existing approaches are not effective due to the challenge of event arguments that are scattered across multi-sentences and they pay more attention to the coreference relationship between entity mentions. However, it is an extremely common phenomenon that there are a large number of crossing sentences pronouns that referring to entity mentions. These pronouns also contain rich semantic information related to events in the document. Therefore, there is still a challenge that how to effectively construct the mention–pronoun coreference relationship and better learn the rich semantic entities representations for DEE. Aiming at the above problems, we propose a novel document-level multi-task learning approach based on coreference-aware dynamic heterogeneous graph network for event extraction, named DMCGEE. Specifically, first, an information enhancement extractor module is constructed to effectively capture multi-types of semantic association information for mentions representations. Second, a mention–pronoun coreference resolution method is proposed to capture mention–pronoun coreference resolution pairs, and a coreference-aware dynamic heterogeneous graph network is constructed to help sentences and mentions representations to focus on the effective global related information, thereby improving the performance of DMCGEE. Experiments show that DMCGEE outperforms the state-of-the-art. [ABSTRACT FROM AUTHOR]

Document-level financial event argument extraction aims to extract a set of structured financial information related to particular financial events from a financial document. This task is challenging because there is complex fine-grained semantic information between financial event arguments, such as cross-document, inner-sentence and context semantic information. Existing approaches are still unable to capture these fine-grained financial semantic features well, leading to the low performance defects. Different from these existing approaches, we propose an end-to-end fine-grained document-level financial event argument extraction approach (FGD-FEAE), which defines event argument extraction as a sequence tagging task. In the encoder, a novel multiple granularity attention layer and a long-short term memory network (LSTM) extension layer are proposed and designed to effectively capture these fine-grained financial semantic information. In the decoder, to avoid the financial semantic confusion problem, a conditional random field (CRF) layer is constructed to jointly tag the financial event arguments. Finally, FGD-FEAE is tested on the two benchmark financial datasets, Chinese financial annotation dataset (ChiFinAnn) and English financial press releases (EFPR), and our own Chinese financial project dataset (ChiFD). These datasets contain a large amount of financial corpus information and complete financial event arguments and provide a solid foundation for verifying the general applicability of the proposed FGD-FEAE. Experimental results show that FGD-FEAE can achieve the better performance than the state-of-the-art approaches, and the performance is improved by 4.69, 5.01 and 3.78 on the three actual financial datasets, respectively. [ABSTRACT FROM AUTHOR]