US 12,223,268 B2
Method and apparatus for event extraction and extraction model training, device and medium
Bingqian Wang, Beijing (CN); Shaoxun Su, Beijing (CN); and Tianxin Liang, Beijing (CN)
Assigned to BOE TECHNOLOGY GROUP CO., LTD., Beijing (CN)
Appl. No. 17/419,836
Filed by BOE TECHNOLOGY GROUP CO., LTD., Beijing (CN)
PCT Filed Sep. 28, 2020, PCT No. PCT/CN2020/118374
§ 371(c)(1), (2) Date Jun. 30, 2021,
PCT Pub. No. WO2022/061877, PCT Pub. Date Mar. 31, 2022.
Prior Publication US 2022/0318506 A1, Oct. 6, 2022
Int. Cl. G06F 40/284 (2020.01); G06F 40/126 (2020.01); G06F 40/30 (2020.01)
CPC G06F 40/284 (2020.01) [G06F 40/126 (2020.01); G06F 40/30 (2020.01)] 18 Claims
OG exemplary drawing
 
1. A method for event extraction, comprising:
processing an object text using a preset extraction model to determine event information of the object text,
wherein the event information comprises an event element, and an event type and a role corresponding to the event element; and the preset extraction model comprises a classification layer an output layer, and an event extraction layer, and the processing of the text using the presets extraction model, comprising:
determining a token attribute of a token in the object text by the classification layer, wherein the token attribute comprises whether the token is a start token of the event element, and whether the token is an end token of the event element;
determining the event element according to the token attribute of the token and determining the event type and the role corresponding to the event element, by the output layer; and
determining a type probability of each event type by the event extraction layer, wherein the type probability of the event type represents a correlation between the object text and the event type,
the determining of the token attribute of the token in the object text comprises:
determining an initial probability that the token in the object text is the start token of the event element of one event type and one role;
determining, according to the initial probability and the type probability of the event type, whether the token is the start token of the event element of the one event type and the one role;
determining an initial probability that the token in the object text is the end token of the event element of one other event type and one other role; and
determining, according to the initial probability and the type probability of the event type, whether the token is the end token of the event element of the one other event type and the one other role.