Method and device for outputting information

Abstract

Embodiments of the present disclosure disclose methods and apparatuses for outputting information. A specific implementation of the method includes: obtaining the sentence to be recognized; performing word segmentation on the sentence to be recognized to obtain a word sequence; inputting the word sequence into a pre-trained multi-task element recognition model based on sequence labeling and entity word prediction, and outputting the recognition Entity words, entity categories and entity word positions, where the multi-task element recognition model includes a sequence tagging network and an entity word prediction network, the sequence tagging network is used to perform sequence tagging tasks, and the entity word prediction network is used to perform entity word prediction tasks , the BILSTM‑CRF network is fused with the entity word prediction network through the fusion module after the hidden layer of LSTM. This embodiment integrates domain dictionary knowledge into the element recognition model to improve the effect of element recognition.

Description

technical field [0001] Embodiments of the present disclosure relate to the field of computer technology, and in particular to methods and devices for outputting information. Background technique [0002] Domain element recognition technology is a natural language processing technology that solves the identification and labeling of elements (important entities) or key attributes contained in natural language texts in professional fields (medical, legal, financial, etc.). For example, in the text processing of medical records in medical scenes, it is hoped to extract the diseases, symptoms, operations, examinations and other elements that doctors are concerned about: the patient is coronary heart disease [disease], after coronary artery bypass grafting [operation], it should be considered that there may be atypical myocardial infarction [ disease], ECG [examination] and myocardial injury markers [examination] should be performed to exclude the diagnosis. [0003] Field elemen...

AI Technical Summary

Problems solved by technology

[0006] 1) The domain element recognition system using pure rules has weak generalization ability and can only identify elements covered by existing rules

[0007] 2) The establishment of the rule system requires a lot of energy from engineers. A complete domain rule system requires engineers to spend years and years in the field and analyze and solve system problems.

[0008] 3) Existing model-based domain element recognition systems require a large amount of manually labeled data

Labeling in the professional field is generally done by practitioners in the professional field, and the cost of labeling is very high, and the labeling cost is about 5-10 times that of ordinary labeling

Images