CN-121999500-A - Invoice processing method and device, electronic equipment and storage medium

CN121999500ACN 121999500 ACN121999500 ACN 121999500ACN-121999500-A

Abstract

The application provides an invoice processing method, an invoice processing device, electronic equipment and a storage medium, and relates to the technical field of data processing. The method comprises the steps of obtaining an invoice image of an invoice, carrying out text recognition on the invoice image to obtain key invoice fields in the invoice image, matching the key invoice fields with stored data in a database through a semantic matching model to obtain a matching result, triggering and executing an invoice processing result instruction corresponding to the matching result to realize full-flow automation of invoice processing, reducing manual intervention requirements, and improving efficiency and accuracy of enterprise purchasing and financial management.

Inventors

LU YITONG
WANG XIN
FU GUOHUI
DANG YANWEI
LI HONGYU
QI WEI

Assignees

昆仑数智科技有限责任公司
中国石油天然气集团有限公司

Dates

Publication Date: 20260508
Application Date: 20251226

Claims (10)

1. A method of invoice processing, the method comprising: acquiring an invoice image of an invoice; Performing text recognition on the invoice image to obtain a key invoice field in the invoice image; Matching the key invoice field with the stored data in the database through a semantic matching model to obtain a matching result; Triggering and executing the invoice processing result instruction corresponding to the matching result.
2. The invoice processing method according to claim 1, wherein said triggering and executing the invoice processing result instruction corresponding to the matching result includes: When the matching result is that the matching is successful, obtaining target storage data matched with the key invoice field; Performing anomaly detection on the key invoice field and the target storage data based on a preset anomaly detection rule to obtain an anomaly detection result; when the abnormality detection result is no abnormality, the invoice processing result instruction is a financial processing instruction corresponding to the invoice; executing the financial processing instructions; when the matching result is a matching failure or the matching result is a matching success and the abnormality detection result is abnormal, the invoice processing result instruction is a model optimization instruction; and dynamically optimizing the semantic matching model based on the key invoice field and the matching result through the model optimization instruction.
3. The invoice processing method according to claim 2, wherein said dynamically optimizing said semantic matching model based on said key invoice fields and said matching results comprises: obtaining a rechecking field obtained after rechecking the key invoice field; Matching the rechecking field with the stored data in the database through the semantic matching model to obtain a secondary matching result; And when the secondary matching result is a matching failure or the secondary matching result is a matching success and the corresponding secondary abnormality detection result is abnormal, dynamically optimizing the semantic matching model based on the rechecking field, the matching result and the secondary matching result to obtain an optimized semantic matching model.
4. The invoice processing method according to claim 3, wherein said dynamically optimizing said semantic matching model based on said review field, said matching result, and said secondary matching result, to obtain an optimized semantic matching model, comprises: Obtaining a problem field and an error rate of the problem field based on the rechecking field, the matching result and the secondary matching result; Obtaining a weight release proportion of the problem field based on the error rate; Acquiring the business type of the invoice; Acquiring weight distribution proportion of other weights in the semantic matching model based on the service types; And carrying out field weight optimization on the semantic matching model based on the weight release proportion and the weight distribution proportion to obtain an optimized semantic matching model.
5. The invoice processing method according to any one of claims 1 to 4, wherein said text recognition of the invoice image results in a key invoice field in the invoice image, comprising: identifying a text in the invoice image; And carrying out semantic analysis on the text by a natural language processing method to obtain key invoice fields in the invoice image.
6. The invoice processing method according to any one of claims 1 to 4, wherein the matching the key invoice field with the stored data in the database by the semantic matching model to obtain a matching result includes: obtaining the similarity between the key invoice field and the storage data in the database through a semantic matching model, wherein the storage data comprises bill information corresponding to a purchase order and a warehouse entry; Taking the bill information with the similarity meeting the similarity condition as initial matching data; calculating cosine similarity between the key invoice field and each initial matching data to obtain a cosine similarity result; Taking initial matching data corresponding to the maximum value in the cosine similarity result as target storage data to obtain a matching result which is successful in matching, wherein the maximum value is larger than a set threshold value; and when the stored data matched with the key invoice field does not exist in the database, obtaining a matching result which is a matching failure.
7. The invoice processing method according to claim 6, wherein after said receipt information whose similarity satisfies a similarity condition is used as initial matching data, further comprising: obtaining an invoice status of the invoice based on the key invoice field; acquiring a risk verification result of a provider corresponding to the initial matching data; and terminating invoice processing when the invoice state does not meet a preset state condition and/or the risk verification result is that verification fails.
8. An invoice processing device, comprising: the acquisition unit is used for acquiring an invoice image of an invoice; the identification unit is used for carrying out character identification on the invoice image to obtain key invoice fields in the invoice image; The matching unit is used for matching the key invoice field with the stored data in the database through a semantic matching model to obtain a matching result; and the processing unit is used for triggering and executing the invoice processing result instruction corresponding to the matching result.
9. An electronic device, comprising: One or more processors; storage means for storing one or more programs which when executed by the one or more processors cause the electronic device to implement the invoice processing method as claimed in any one of claims 1 to 7.
10. A computer-readable storage medium having stored therein computer-executable instructions which, when executed by a processor, are adapted to carry out the invoice processing method as claimed in any one of claims 1 to 7.

Description

Invoice processing method and device, electronic equipment and storage medium Technical Field The present application relates to data processing, and in particular, to an invoice processing method, an invoice processing device, an electronic device, and a storage medium. Background In the purchasing management flow of enterprise resource planning (ERP, ENTERPRISE RESOURCE PLANNING) systems, the processing of purchasing invoices is a core link of financial and supply chain management. The enterprise purchasing department needs to sign a purchasing Order (PO, purchase Order) with the supplier, the warehouse department generates a warehouse entry Order (GRN, goods Received Note) according to the actual arrival condition, and the financial department needs to match the invoice actually received based on the PO and the GRN to complete consistency check of three orders (PO, GRN, invoice). The existing invoice processing flow has the problems of excessive manual intervention and regular solidification, so that a large amount of labor cost is consumed, financial risk is increased due to high error rate, and enterprise operation efficiency and compliance are seriously affected. Disclosure of Invention The embodiment of the application provides an invoice processing method, an invoice processing device, electronic equipment and a storage medium, which are used for realizing full-flow automation of invoice processing, reducing manual intervention requirements and remarkably improving the efficiency and accuracy of enterprise purchasing and financial management. In a first aspect, an embodiment of the present application provides an invoice processing method, including acquiring an invoice image of an invoice; performing text recognition on the invoice image to obtain a key invoice field in the invoice image; Matching the key invoice field with the stored data in the database through a semantic matching model to obtain a matching result; Triggering and executing invoice processing result instructions corresponding to the matching results. In one possible implementation manner, triggering and executing invoice processing result instructions corresponding to the matching result includes: When the matching result is that the matching is successful, obtaining target storage data matched with the key invoice field; Performing anomaly detection on the key invoice field and the target storage data based on a preset anomaly detection rule to obtain an anomaly detection result; When the abnormality detection result is that no abnormality exists, the invoice processing result instruction is a financial processing instruction corresponding to the invoice; executing financial processing instructions; when the matching result is that the matching is failed, or the matching result is that the matching is successful and the abnormality detection result is that the abnormality exists, the invoice processing result instruction is a model optimization instruction; and dynamically optimizing the semantic matching model based on the key invoice field and the matching result through a model optimization instruction. In one possible implementation, dynamically optimizing the semantic matching model based on key invoice fields and matching results includes: Obtaining a rechecking field obtained after rechecking the key invoice field; Matching the rechecking field with the stored data in the database through a semantic matching model to obtain a secondary matching result; and when the secondary matching result is failed in matching or the secondary matching result is successful in matching and the corresponding secondary abnormality detection result is abnormal, dynamically optimizing the semantic matching model based on the rechecking field, the matching result and the secondary matching result to obtain an optimized semantic matching model. In one possible implementation manner, dynamically optimizing the semantic matching model based on the review field, the matching result and the secondary matching result to obtain an optimized semantic matching model, including: Obtaining a problem field and the error rate of the problem field based on the rechecking field, the matching result and the secondary matching result; Obtaining a weight release proportion of the problem field based on the error rate; Acquiring the business type of the invoice; acquiring weight distribution proportion of other weights in the semantic matching model based on the service types; And carrying out field weight optimization on the semantic matching model based on the weight release proportion and the weight distribution proportion to obtain an optimized semantic matching model. In one possible implementation manner, performing text recognition on the invoice image to obtain key invoice fields in the invoice image, including: identifying text in the invoice image; And carrying out semantic analysis on the text by a natural language processing method to obtain key invo