Novel insurance policy identification model size compression method
A technology for identifying models and compression methods, applied in neural learning methods, character and pattern recognition, biological neural network models, etc., which can solve customer claims, underwriting, pre-insurance inconvenience, model cannot run smoothly, and reduce customer user experience and other problems, to achieve the effect of reducing the model size, increasing the calculation speed, and reducing the amount of calculation
Pending Publication Date: 2021-10-15
众淼创新科技(青岛)股份有限公司
View PDF0 Cites 0 Cited by
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
Among them, there are less than 190 million digital insurance policies, and the vast majority of insurance policies are still paper policies, which have not been digitized, causing inconvenience for customers in many links such as claim settlement, underwriting, and pre-underwriting.
However, due to the CPU performance limitation of the mobile terminal, the model cannot run smoothly on the mobile terminal, which reduces the customer's user experience
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 moreImage
Smart Image Click on the blue labels to locate them in the text.
Smart ImageViewing Examples
Examples
Experimental program
Comparison scheme
Effect test
Embodiment Construction
[0051] Exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be embodied in various forms and should not be limited by the embodiments set forth herein. Rather, these embodiments are provided for more thorough understanding of the present disclosure and to fully convey the scope of the present disclosure to those skilled in the art.
[0052] The present invention proposes a novel policy recognition model size compression method QSlim. On the premise of ensuring the same accuracy, the calculation speed is increased by 187%, and the volume is reduced by 26%.
[0053] The first core innovation of the present invention is based on the calculation process characteristics of the policy OCR neural network, converts the float type into the int type, and adds an error fun...
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
Login to View More Abstract
According to the novel insurance policy identification model size compression method provided by the invention, by analyzing OCR identification results of 200,000 insurance policies, the calculation speed is improved by 187% on the premise of ensuring that the precision is not changed. The novel insurance policy identification model size compression method QSlim mainly comprises two parts: firstly, on the basis of calculation process characteristics of an insurance policy OCR neural network, a float type is converted into an int type, and an error function bias is added to the tail of each layer, so that the operand is greatly reduced; and secondly, based on the structure characteristics of the insurance policy OCR neural network, a network structure fitting method is invented, networks with similar structures are found out, the fitting method is used for reducing the grids with the similar structures, and the model volume is greatly reduced.
Description
technical field [0001] The invention relates to the technical field of insurance policy recognition (OCR) model size compression technology, in particular to a novel policy recognition model size compression method QSlim (QuanZhangGui Slim). Background technique [0002] Policy recognition refers to a technical solution for converting paper policy into unstructured plain text through OCR technology, which can be invoked by other business processes in the insurance industry. With the rapid development of my country's economy and the substantial increase in the level of national income, the number of insurance policies held by the public is also soaring rapidly, which puts forward higher requirements for the digital management level of the insurance industry. As of 2020, there are a total of 1.67 billion insurance policies in my country, with an average of 1.19 policies per capita. Among them, there are less than 190 million digital insurance policies, and the vast majority o...
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
Login to View More Patent Type & Authority Applications(China)
IPC IPC(8): G06K9/20G06K9/62G06Q40/08G06N3/04G06N3/08
CPCG06Q40/08G06N3/04G06N3/08G06F18/22
Inventor 王合平
Owner 众淼创新科技(青岛)股份有限公司