Named entity identification system and identification method based on deep network AS-LSTM

A BI-AS-LSTM, named entity recognition technology, applied in neural learning methods, biological neural network models, instruments, etc., can solve the problems of insufficient training set sample size, poor robustness of context and low iteration efficiency, etc. Improve iterative efficiency, increase robustness, predict accurate and stable results

Pending Publication Date: 2021-01-22
BEIJING ZHONGBIAO NETWORK TECH CO LTD
View PDF2 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0007] The purpose of the present invention is to improve the accuracy rate and recall rate of named entity recognition, and solve the problems of poor context robustness, high time cost, and high labor cost of the existing long-term short-term memory network (LSTM network) in actual use. Insufficient, to solve the problem of insufficient sample size and low iteration efficiency of the current deep model in the cold start training set, and provide a named entity recognition system and recognition method based on deep network AS-LSTM

Method used

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View more

Image

Smart Image Click on the blue labels to locate them in the text.
Viewing Examples
Smart Image
  • Named entity identification system and identification method based on deep network AS-LSTM
  • Named entity identification system and identification method based on deep network AS-LSTM
  • Named entity identification system and identification method based on deep network AS-LSTM

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0055] This embodiment discloses a named entity recognition system based on deep network AS-LSTM. In this embodiment, the named entity recognition system includes a network model BI-AS-LSTM-CRF, and the network model BI-AS-LSTM-CRF includes Text feature layer, context feature layer BI-AS-LSTM, CRF layer.

[0056] Generally speaking, in the traditional named entity recognition system, the BI-LSTM network in the named entity recognition network model BI-LSTM-CRF is composed of two LSTM networks spliced ​​to form a bidirectional LSTM network, such as figure 1 shown. Specifically, the LSTM network consists of a forget gate (such as Figure 2a ), the output gate (such as Figure 2b ), input gate (such as Figure 2c ) consists of three gates. The core of the network is the cell state, which is represented by a horizontal line running through the cell. The cell state is like a conveyor belt. It runs through the entire cell but has only a few branches, which ensures that informatio...

Embodiment 2

[0065]Since named entity recognition systems are mostly cold-started, they have the problem of low efficiency. At present, academia and the industry have tried to use some methods, the most common is to use pre-trained models for word embedding, such as ELMO, BERT, GPT-3 The pre-training model with a large amount of parameters is used as the generator of the upstream word vector, and then finetune is used to optimize the downstream tasks. However, for many research institutes, the computing resources and costs brought by this pre-training model are too large. The response speed of the service interface is too slow. For example, the NER model prediction speed of BERT under ordinary GPU calculation is about 500ms, which is very slow to meet daily use and services.

[0066] Therefore, this embodiment is improved on the basis of Embodiment 1, and the Random Replace training method is added to the named entity recognition system, and the Random Replace training method is combined wi...

Embodiment 3

[0070] Such as Figure 6 As shown, this embodiment discloses a named entity recognition method based on deep network AS-LSTM, which is applied to the above-mentioned named entity recognition system to recognize text, and the startup form of the named entity recognition system formed by deep network AS-LSTM is as follows Cold start. The named entity recognition method includes the following steps:

[0071] S1. Construction of the network model BI-AS-LSTM-CRF;

[0072] Specifically, the construction of the network model BI-AS-LSTM-CRF includes the extraction of feature information of the input text in the text, the output of the feature information to obtain the output sequence, the acquisition of the context features of the input text, and the context features marked by BIO for each in the input text. The position information of words in the text, and multiple steps such as obtaining entity labels.

[0073] S2. Determine the recognition target, and mark the recognition corpu...

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to view more

PUM

No PUM Login to view more

Abstract

The invention provides a named entity recognition system based on a deep network ASLSTM. The system comprises a network model BIASLSTMCRF, wherein the network model BIASLSTMCRF comprises a text feature layer, a context feature layer BIASLSTM, and a CRF layer, the context feature layer BIASLSTM comprises two ASLSTM deep networks, and the two ASLSTM deep networks are spliced to form a bidirectionalASLSTM network. According to the method, a novel ASLSTM deep network is designed in the named entity recognition system, the more stable and accurate cell state of the preamble and the postamble of the named entity in the input text can be obtained, and the network depends on self-learning, so context-related semantic representation can be learned, robustness for coping with the unrelated words ofthe precedent and the consequent can be improved, and errors of an identification system are reduced.

Description

technical field [0001] The invention belongs to the field of artificial intelligence natural language processing, and relates to named entity recognition technology in the field of natural language processing, in particular to a named entity recognition system and recognition method based on deep network AS-LSTM. Background technique [0002] With the development of artificial intelligence technology, machine learning has become one of the most commonly used methods of natural language processing. As a branch of machine learning, deep learning has achieved the best results in almost all sub-tasks of natural language processing, including dialogue systems, named entity recognition, language Translation and other tasks, among them, named entity recognition (NER) has also become the most common problem in the field of natural language processing. [0003] The deep learning network in Named Entity Recognition (NER) is currently a recognized method in the industry and academia. ...

Claims

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to view more

Application Information

Patent Timeline
no application Login to view more
Patent Type & Authority Applications(China)
IPC IPC(8): G06F40/295G06N3/04G06N3/08
CPCG06F40/295G06N3/049G06N3/08G06N3/045
Inventor 王国鸿
Owner BEIJING ZHONGBIAO NETWORK TECH CO LTD
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Try Eureka
PatSnap group products

Browse by: Latest US Patents, China's latest patents, Technical Efficacy Thesaurus, Application Domain, Technology Topic.

© 2024 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap