Financial text event extraction method and device suitable for few samples and biased data

• Main Authors: LI SHUAN, NA CHONGNING
• Format: Patent
• Language: Chinese; English
• Published: 28.07.2023
• Subjects: CALCULATING; COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS; COMPUTING; COUNTING; DATA PROCESSING SYSTEMS OR METHODS, SPECIALLY ADAPTED FORADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL, SUPERVISORYOR FORECASTING PURPOSES; ELECTRIC DIGITAL DATA PROCESSING; PHYSICS; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE,COMMERCIAL, FINANCIAL, MANAGERIAL, SUPERVISORY OR FORECASTINGPURPOSES, NOT OTHERWISE PROVIDED FOR

The invention discloses a financial text event extraction method and device suitable for few samples and biased data, and aims to solve the problems that the number of event types in financial event tasks is large, the frequency of occurrence of different events is different, and few samples and biased data often exist. Data with few samples are enhanced from two levels of event subjects and event types, so that the influence of biased and few-sample type data on a model extraction result is effectively relieved; for the problem that the same event body in the description text appears for multiple times and corresponds to a plurality of event types and the problem that a plurality of event bodies in the description text correspond to the same event type, an extraction model based on two-stage distribution processing of event body prediction and event type prediction is constructed; the extraction model is trained by an enhanced training sample, and accurate extraction of financial text events under various co