Financial text event extraction method based on trigger word morphological learning

Abstract

The invention realizes the method for extracting financial text events based on trigger word state learning through the method in the neural network field. The method includes three steps: financial text pre-training, event classification, and financial text event element extraction based on trigger word morphological learning; the implementation of the financial text pre-training step combines financial knowledge graphs to build a BERT pre-training model to input word sequences As the model input, combined with the neural network method, retraining is carried out on the existing financial text training set and financial knowledge map data, and the word representation and entity representation suitable for downstream event classification and event extraction are obtained, and then the word representation is used to make multiple labels. The classification task obtains the event detection result, and finally the word representation corresponding to each input word is calculated as the representation vector of the multi-label classification task to obtain the output result, thus forming a method that can automatically and accurately extract financial text events.

Description

technical field [0001] The invention relates to the field of neural networks, in particular to a method for extracting financial text events based on trigger word morphology learning. Background technique [0002] In today's information explosion era, quickly and efficiently capturing events from a large amount of text and mining the key information of events is an important problem that needs to be solved urgently. Therefore, the event extraction task came into being. However, for event extraction tasks, due to the difficulty of data labeling, domain data is more difficult to label, and practitioners in specific fields are required to label. Therefore, this adds difficulty to the task of domain event extraction. At present, there are relatively few studies on event extraction in the financial field, and few consider the voice relationship between trigger words and event elements. For a large number of domain vocabulary in the financial field, it is more difficult to under...

AI Technical Summary

Problems solved by technology

At present, there is no set of financial text event extraction methods and devices based on trigger word morphological learning to realize the method for extracting event elements in the financial field

Images