Method for identifying a named entity based on policy value network and tree search

A named entity recognition and tree search technology, applied in the field of information processing, can solve problems such as labor-intensive

Active Publication Date: 2019-09-06
BEIJING UNIV OF POSTS & TELECOMM
View PDF4 Cites 14 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

The former, such as the maximum entropy Markov model (MEMM) and conditional random fie

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
  • Method for identifying a named entity based on policy value network and tree search
  • Method for identifying a named entity based on policy value network and tree search
  • Method for identifying a named entity based on policy value network and tree search

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0019] Next, embodiments of the present invention will be described in more detail.

[0020] First, the first part of the present invention is introduced: Markov decision process modeling.

[0021] Reinforcement learning is another important learning category in machine learning. It mainly uses a trial-and-error mechanism to interact with the environment, and finally achieves the goal of maximizing cumulative rewards. The Markov decision process of reinforcement learning mainly includes five elements: state, action, transition probability, reward and decay factor. figure 1 A schematic diagram of a Markov decision process is given. Suppose X={x 1 ,x 2 ,...,x M} is the word sequence that needs to be marked, Z={z 1 ,z 2 ,…,z M} corresponds to the real label sequence, where M is the length of the sequence, and the design process of the Markov decision process is as follows:

[0022] Step S1: State: Define the state (S) at time t as the following triplet.

[0023]

[00...

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 discloses a method for identifying a named entity based on a policy value network, and belongs to the field of information processing. The method comprises the following steps of firstly, modeling a labeling process of named entity recognition into a Markov decision process (MDP) so that a model for identifying the named entity based on enhanced learning is provided, namely MM-NER. MM-NER is the first to apply the Monte Carlo Tree Search (MCTS) enhanced MDP model to named entity recognition (sequence tag task).A strategy value network is designed on the basis of an MDP state definition to obtain label probability and label sequence accuracy evaluation, and an MCTS is used for simulation, so that a label sequence with more global consciousness is searched out. In the inferenceprocess, the strategy value network is directly used to ensure that the identification effect is basically consistent with the tree search strategy, and the time complexity is greatly reduced. An experimental result of the present invention on the CoNLL2003 named entity identification data set demonstrates the effectiveness of the MM-NER with the K-step exploration decision mechanism.

Description

technical field [0001] The invention relates to the field of information processing, in particular to a named entity recognition method based on strategy value network and tree search enhancement. Background technique [0002] Named entity recognition is widely used, and the main methods can be divided into methods based on statistics and methods based on neural networks. The former, such as maximum entropy Markov model (MEMM) and conditional random field (CRF), most of these methods require a lot of manpower to construct manual features. The latter is the current mainstream method, and the representative model has a bidirectional long short-term memory network (BLSTM). In order to further model the globality of the output label sequence and solve the problem of label bias, the model of bidirectional long short-term memory network plus conditional random field (BLSTM+CRF) has achieved good results. In recent years, the development of deep learning has greatly enhanced the ...

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
IPC IPC(8): G06F16/33G06F16/36
CPCG06F16/33G06F16/367
Inventor 高升劳雅迪李思徐雅静陈光徐君胡旻卉
Owner BEIJING UNIV OF POSTS & TELECOMM
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