CDMG:Combing DeBERTa-v2 Model with Globalpointer for Chinese Text Level Event Extraction

Extracting event information from professional texts is a prerequisite for intelligence decision-making system or intelligence personnel to conduct intelligence analysis and decision-making. In the past, event extraction from professional texts mainly relied on human reading and event extraction. Since entering the era of big data, millions of professional texts of interest to intelligence personnel are generated every day. And it's basically impossible to extract information about events from these texts by relying on people alone. In this case. We first use the pre-trained Chinese DeBERTa-v2 model for text coding and word segmentation. Secondly, gobalpointer technology is used to predict and extract event arguments, and Adam optimizer is used for optimization. Finally, a Chinese text level event extraction framework for the professional field is constructed by dividing argument into events through complete subgraph recognition technology. In this way, the event information of the text in the professional field can be extracted, so as to improve the application efficiency of text intelligence in the professional field and promote the application of artificial intelligence technology in the professional intelligence field.