Underwriting method, system, device and program product

Abstract

The invention discloses an underwriting method, system and device and a program product. The method comprises the steps of obtaining an underwriting material; identifying a document structure of the underwriting material, and performing layout segmentation on the underwriting material based on the document structure; semantic analysis is conducted on the basis of a pre-training language model and the underwriting material obtained after layout segmentation, an analysis result is obtained, and the pre-training language model is obtained through training on the basis of historical underwriting information and used for conducting semantic analysis on the underwriting material; and comparing the analysis result with a preset underwriting rule to obtain an underwriting result. Therefore, the whole underwriting process does not need manual intervention, the auditing time is greatly shortened, and the underwriting process is improved from the manual hour level to the minute level. Besides, the method provided by the invention is not limited by human resources, a large number of underwriting requests can be processed, and a stable processing speed can be kept even under the condition that the business volume is increased sharply.

Description

Technical Field

[0001] The present application relates to the field of computer technology, and in particular to an underwriting method, system, device and program product. Background Art

[0002] In the underwriting process of the insurance industry, the traditional manual review method has always dominated. However, with the rapid development of insurance business and the increasing diversification of insurance products, the manual underwriting method has gradually exposed many limitations.

[0003] During the manual underwriting process, each policy needs to be carefully reviewed by the underwriter. This process is time-consuming, and the average review time is usually between 0.5 and 2 hours. When the business volume is large, the timeliness of manual underwriting cannot be guaranteed, which often leads to an extension of the underwriting cycle, affecting the customer's insurance experience and the insurance company's business efficiency. In addition, despite the rich experience of the auditors, the error rate of manual underwriting is still as high as 20% due to human factors. Common errors include but are not limited to missing information, misreading medical reports, and misunderstanding of underwriting rules. These errors may not only lead to incorrect underwriting conclusions, but may also cause subsequent claims disputes and increase the company's operational risks. At the same time, once an error is discovered, the correction process is often cumbersome and time-consuming, further affecting the overall work efficiency and service quality. Summary of the invention

[0004] Based on the above problems, the present application provides an underwriting method, system, device and program product.

[0005] The embodiments of the present application disclose the following technical solutions:

[0006] A first aspect of an embodiment of the present application provides an underwriting method, including:

[0007] Obtain underwriting materials;

[0008] Identifying the document structure of the underwriting materials, and performing layout segmentation on the underwriting materials based on the document structure;

[0009] Performing semantic analysis on the underwriting materials after layout segmentation based on a pre-trained language model to obtain an analysis result, wherein the pre-trained language model is a model trained based on historical underwriting information and used for performing semantic analysis on the underwriting materials;

[0010] The analysis results are compared with preset underwriting rules to obtain underwriting results.

[0011] In a possible implementation, the layout-segmented underwriting materials include a plurality of segmented regions, and the semantic analysis is performed based on the pre-trained language model and the layout-segmented underwriting materials to obtain analysis results, including:

[0012] Extracting text information that meets a preset criticality condition in at least one divided area;

[0013] Based on the text information of the preset critical condition, calling the corresponding semantic analysis template;

[0014] Based on the pre-trained language model and the semantic analysis template, semantic analysis is performed on the extracted text information to obtain an analysis result, and the semantic analysis template is used to guide the pre-trained language model to perform semantic analysis.

[0015] In a possible implementation, before comparing the analysis result with a preset underwriting rule to obtain an underwriting result, the method further includes:

[0016] Based on the text information of the preset critical conditions, the corresponding preset underwriting rules are called.

[0017] In a possible implementation, comparing the analysis result with a preset underwriting rule to obtain an underwriting result includes:

[0018] Adjusting the analysis result into a structured format corresponding to the preset underwriting rule;

[0019] The analysis results in a structured format are compared with the preset underwriting rules to obtain the underwriting results.

[0020] In a possible implementation, the method further includes:

[0021] When the uploaded underwriting materials do not meet the preset material requirements, a prompt message is generated, which includes the reason why the uploaded underwriting materials do not meet the preset material requirements, and is used to remind the uploader to provide underwriting materials that meet the preset material requirements.

[0022] In a possible implementation, the underwriting result includes text information in the underwriting materials that does not comply with preset underwriting rules, and the method further includes:

[0023] The text information in the underwriting materials after layout segmentation that does not comply with the preset underwriting rules is marked in a preset form.

[0024] In a possible implementation, the method further includes:

[0025] Acquire query information of the user, where the query information is used to represent the user's questions and / or needs regarding the underwriting process;

[0026] Reply information is generated based on the query information and fed back to the user, and the reply information is used to assist the user in submitting underwriting materials.

[0027] In a first aspect, an embodiment of the present application provides an underwriting system.

[0028] An acquisition unit, used to acquire underwriting materials;

[0029] An identification unit, used to identify the document structure of the underwriting material, and perform layout segmentation on the underwriting material based on the document structure;

[0030] An analysis unit, configured to perform semantic analysis on the underwriting materials after layout segmentation based on a pre-trained language model to obtain an analysis result, wherein the pre-trained language model is a model trained based on historical underwriting information and used for performing semantic analysis on the underwriting materials;

[0031] The comparison unit is used to compare the analysis result with the preset underwriting rules to obtain the underwriting result.

[0032] A third aspect of an embodiment of the present application provides an electronic device, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the underwriting method as described in the first aspect above is implemented.

[0033] A fourth aspect of an embodiment of the present application provides a computer program product. When the computer program product runs on a computer, the computer executes the underwriting method as described in the first aspect above.

[0034] A fifth aspect of an embodiment of the present application provides a computer-readable storage medium, in which instructions are stored. When the instructions are executed on a terminal device, the terminal device executes the underwriting method as described in the first aspect above.

[0035] Compared with the prior art, this application has the following beneficial effects:

[0036] After obtaining the underwriting materials, the document structure is automatically identified and the layout is segmented, and then the pre-trained language model is used to perform semantic analysis on the materials. In addition, based on the pre-trained language model trained with historical underwriting information, it is possible to perform in-depth semantic analysis of the underwriting materials, which not only reduces the reliance on manual experience and expertise, but also ensures the consistency and accuracy of the review. As a result, the entire underwriting process does not require manual intervention, greatly shortens the review time, and improves the underwriting process from manual hours to minutes. In addition, the method provided in this application is not limited by human resources, can handle a large number of underwriting requests, and can maintain a stable processing speed even in the case of a surge in business volume. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0038] Figure 1 A flowchart of the first underwriting method provided in an embodiment of the present application;

[0039] Figure 2 A flowchart of a second underwriting method provided in an embodiment of the present application;

[0040] Figure 3 A structural diagram of the underwriting system provided for an embodiment of the present application. DETAILED DESCRIPTION

[0041] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0042] To facilitate understanding of the technical solutions provided by the embodiments of the present application, the professional terms involved in the embodiments of the present application will be explained below.

[0043] Underwriting is a key link in the insurance business, which refers to the process by which insurance companies assess the risks of potential insured persons or objects before accepting insurance applications. Its main purpose is to determine whether to insure and under what conditions, including setting premiums, determining insurance amounts and terms, etc.

[0044] RAG stands for Retrieval-Augmented Generation in English. Retrieval-Augmented Generation is a technical framework that combines information retrieval (Retrieval) and natural language generation (Generation). It improves the quality and accuracy of generated text by retrieving relevant information from a large number of external knowledge bases and incorporating this information into the generation model.

[0045] To facilitate understanding of the technical solutions provided by the embodiments of the present application, the background technology involved in the embodiments of the present application will be described below.

[0046] As mentioned above, when customers upload medical reports, due to the variety of materials, they often need to interact with insurance company employees in multiple rounds to complete the upload of correct materials. This may lead to time conflicts and reduce communication efficiency. Manual review of each underwriting application takes about 0.5 to 2 hours, which cannot break through manpower limitations and has poor timeliness. The underwriting process requires high professional knowledge and complex rules, and manual review requires certain experience. In addition, the error rate of manual underwriting is as high as 20%, which affects the accuracy of the underwriting results. Due to the long-tail problem of specialized diseases, that is, each individual involves various personalized conditions, the iteration cycle of product underwriting rules is very long, and a lot of underwriting experience is needed to master the rules and make corrections.

[0047] In order to solve the above problems, the embodiment of the present application automatically identifies the document structure and performs layout segmentation after obtaining the underwriting materials, and then uses the pre-trained language model to perform semantic analysis on the materials. The entire process does not require human intervention, which greatly shortens the review time and improves the underwriting process from manual hours to minutes. In addition, the online underwriting system can respond to customer questions and needs around the clock, without being restricted by working hours and human resources, significantly improving the timeliness and continuity of services. Based on the pre-trained language model, it is possible to perform in-depth semantic analysis of underwriting materials, automatically extract key information and understand its logical relationships. This not only reduces the reliance on manual experience and expertise, but also ensures the consistency and accuracy of the review.

[0048] It should be noted that the underwriting method, system, device and medium provided in this application can be applied to the field of computer technology. The above is only an example and does not limit the application field of the underwriting method, system, device and medium provided in this application. In addition, the embodiments of the present application may not limit the execution subject of the underwriting. For example, the underwriting method of the embodiment of the present application can be applied to data processing devices such as terminal devices or servers. Among them, the terminal device can be an electronic device such as a computer, a personal digital assistant (PDA). The server can be a stand-alone server, a cloud server, or a cluster server composed of multiple servers.

[0049] In order to make the purpose, technical solution and advantages of the embodiments of the present application clearer, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0050] The following is an example of an underwriting method provided by the present application. Figure 1 ,Should Figure 1 A flowchart of a first underwriting method provided in an embodiment of the present application, the method comprising:

[0051] S101. Obtain underwriting materials.

[0052] The system receives medical reports and other related materials in various formats uploaded by customers. These materials can be images (such as scanned copies), PDF documents, spreadsheets, etc.

[0053] Support multiple submission methods, including but not limited to online upload, email or mobile application upload. Perform preliminary verification when receiving materials to check whether the files meet basic requirements (such as file size, format, etc.) and provide immediate feedback to users.

[0054] When the uploaded underwriting materials do not meet the preset material requirements, a prompt message is generated, which includes the reason why the uploaded underwriting materials do not meet the preset material requirements, and is used to remind the uploader to provide underwriting materials that meet the preset material requirements.

[0055] S102: Identify the document structure of the underwriting materials, and layout and segment the underwriting materials based on the document structure.

[0056] According to the document structure identified in the previous step, the underwriting materials are laid out and segmented into logical units. For example, the personal information section, health status section, medical history section, etc. in the application form are segmented separately. Layout segmentation helps to decompose the document into smaller, easier-to-process units, such as paragraphs, tables, pictures, etc., to prepare for subsequent semantic analysis, making subsequent text processing more accurate and efficient.

[0057] In actual application, optical character recognition (OCR) can be used to convert documents in image form into editable and searchable text data. For documents containing tables, a specialized algorithm is used to parse the table content to ensure that the logical relationship of the data within the table is correctly understood. The layout analysis algorithm is used to further refine the recognition of the document structure, especially for complex structures such as nested tables or charts, and perform more refined area divisions. Not only the content of each part is recognized, but also the correlation between them is understood, for example, which paragraphs belong to the diagnosis description, which tables contain the examination results, etc.

[0058] S103: Perform semantic analysis on the underwriting materials after layout segmentation based on the pre-trained language model to obtain analysis results.

[0059] The pre-trained language model is a model trained based on historical underwriting information and is used to perform semantic analysis on underwriting materials. The following is an explanation of the training process of the model:

[0060] The model uses a large amount of historical underwriting information as training data, which includes but is not limited to past medical reports, pathology diagnoses, imaging results, and corresponding underwriting conclusions and claim records.

[0061] Before training, all text materials will undergo detailed preprocessing steps, such as denoising, standardization, word segmentation, and removal of stop words, to ensure that the model can learn useful features from high-quality data.

[0062] Given the professionalism and particularity of the insurance and medical fields, special emphasis is placed on domain adaptability during model training. By introducing domain-specific vocabularies, terminology libraries, and expertise, the model has higher accuracy and understanding when processing related texts.

[0063] The training adopts supervised learning, that is, each input sample (such as a complete underwriting application) is equipped with a clear label or target output (such as the final underwriting conclusion). This helps the model learn how to make correct judgments based on the provided materials.

[0064] The model can also combine image recognition and other non-text data parsing capabilities to achieve a more comprehensive understanding. For example, for a case that includes an X-ray scan, the model can process both the text description and the image content to provide a more accurate analysis result.

[0065] Once trained, the pre-trained language model can be applied to newly received underwriting materials to perform deep semantic analysis. It can not only identify and extract key information, but also understand the logical relationship between this information, providing a solid foundation for subsequent rule matching and decision support.

[0066] In actual applications, the model will be used with customized semantic analysis templates to guide the semantic analysis process, ensuring that no important information is missed and that the analysis strategy can be flexibly adjusted for different types of underwriting materials. As new underwriting cases continue to accumulate, the model can maintain its performance through incremental learning or periodic retraining to ensure that it is always in the best condition and adapts to changing market demands and technological advances.

[0067] S104: Compare the analysis result with the preset underwriting rules to obtain the underwriting result.

[0068] This step compares the results of semantic analysis with the insurance company's preset underwriting rules. The preset underwriting rules may involve multiple evaluation criteria such as age, health status, financial status, etc. The comparison process may involve complex logical judgments and risk assessments, but ultimately outputs a clear underwriting result, such as underwriting, rejection, or premium underwriting.

[0069] In a possible implementation, the layout-segmented underwriting materials include a plurality of segmented regions, and the semantic analysis is performed based on the pre-trained language model and the layout-segmented underwriting materials to obtain analysis results, including:

[0070] Step A1: extract text information that meets preset key conditions in at least one divided area.

[0071] Traverse all the segmented areas and identify the text information that contains the preset key conditions. The text information that meets the preset key conditions may include but is not limited to: specific keywords or phrases (such as "medical history", "income status", "family status", etc.). Specific data types (such as dates, numbers, addresses, etc.). Specific text formats (such as a column in a table). By setting these key conditions, the system can accurately filter out information directly related to the underwriting decision from the underwriting materials.

[0072] Step A2: Based on the text information of the preset key conditions, a corresponding semantic analysis template is called.

[0073] After extracting text information that meets the key conditions, the corresponding semantic analysis template can be called according to the type or content of this information. These templates may be specially designed for different key conditions to provide more accurate and in-depth semantic analysis. For example, for text information containing "medical history", a template specifically used to analyze health conditions can be called; and for text information containing "income status", a template used to evaluate financial status can be called.

[0074] Step A3: Based on the pre-trained language model and the semantic analysis template, semantic analysis is performed on the extracted text information to obtain an analysis result. The semantic analysis template is used to guide the pre-trained language model to perform semantic analysis.

[0075] Use pre-trained language models (such as BERT, GPT, etc.) and the previously called semantic analysis template to perform deep semantic analysis on the extracted text information. The pre-trained language model can capture the contextual relationships, implicit information, and potential risk points in the text, while the semantic analysis template can guide the model to understand this information more accurately. Through this step of analysis, an analysis result can be generated that includes key information extraction, risk level assessment, health status overview, etc. This analysis result will serve as the basis for subsequent comparison with the preset underwriting rules.

[0076] By dividing the underwriting materials into multiple segments and performing detailed semantic analysis on these segments, the key information in the underwriting materials can be captured more accurately, improving the accuracy and efficiency of underwriting decisions. At the same time, by calling different semantic analysis templates, it is more suitable for different types of underwriting materials, further enhancing the flexibility and scalability of the method.

[0077] In a possible implementation, before comparing the analysis result with a preset underwriting rule to obtain an underwriting result, the method further includes:

[0078] Based on the text information of the preset critical conditions, the corresponding preset underwriting rules are called.

[0079] After extracting text information that meets the preset key conditions, not only can the corresponding semantic analysis template be called for semantic analysis, but the preset underwriting rules related to these key conditions can also be directly called. These preset underwriting rules are formulated based on the insurance company's business logic, risk assessment standards and regulatory requirements, and are used to guide the formulation of underwriting decisions.

[0080] For example, if the extracted text information contains a specific disease name, the underwriting rules related to the disease can be called. These rules can include factors such as the severity of the disease, treatment conditions, and prognosis, and corresponding underwriting suggestions or restrictions can be given accordingly. After calling the corresponding preset underwriting rules, the results of the semantic analysis can be compared with these rules. This process includes matching and evaluation at multiple levels, including verification of key information, comparison of risk levels, and compliance checks.

[0081] In a possible implementation, comparing the analysis result with a preset underwriting rule to obtain an underwriting result includes:

[0082] The analysis results are adjusted to a structured format corresponding to the preset underwriting rules; the analysis results in the structured format are compared with the preset underwriting rules to obtain the underwriting results.

[0083] Structured data can be directly mapped to preset underwriting rules, allowing computers to quickly and accurately perform automated rule matching without the need for manual inspection. Through structured processing, each data point is ensured to be correctly identified and classified, reducing errors caused by inconsistent data formats or missing information.

[0084] Structured data allows the system to more accurately compare specific terms in the underwriting rules. For example, key factors such as the diagnosis of a specific disease, the patient's age, and treatment history can be directly matched one-to-one with the corresponding conditions in the rules.

[0085] When all analysis results are organized into a structured form, each decision point can be traced back to a specific part of the original document, increasing the transparency of the system. Auditors or customers can view specific analysis results and their corresponding underwriting rules through a visual interface, which enhances trust. In addition, structured data is easy to store and retrieve, which facilitates manual review or audit in the future, ensuring the compliance and fairness of the underwriting process.

[0086] Assume that a customer uploads an application for special insurance for breast nodules. After optical character recognition (OCR) processing and deep semantic analysis by the pre-trained language model, the extracted information can be organized into the following structured format:

[0087] Basic information of the patient:

[0088] Name: Zhang San

[0089] Age: 45

[0090] Gender: Female

[0091] Pathological examination results:

[0092] Cytological diagnosis: benign tumor

[0093] Imaging diagnosis: no abnormal lymphadenopathy

[0094] Past medical history: No other major illness record

[0095] This structured information can be directly used to compare underwriting rules. For example, based on the patient's age (45 years old) and cytological diagnosis results (benign tumor), the system can quickly determine whether the insurance conditions are met. If the imaging diagnosis results show abnormalities, further review or the customer may be required to provide additional medical reports.

[0096] In a possible implementation, the underwriting result includes text information in the underwriting materials that does not comply with preset underwriting rules, and the method further includes:

[0097] The text information in the underwriting materials after layout segmentation that does not comply with the preset underwriting rules is marked in a preset form.

[0098] In order to more intuitively display the text information that does not conform to the rules, it can be marked in a preset form. This mark can be highlighted, underlined, bolded, color-differentiated or any other form that can attract attention. The specific form of the mark can be determined according to the needs of the insurance company, the design of the user interface and the technical implementation of the system.

[0099] For example, if there is a sentence in the underwriting materials that is denied coverage because it contains a disease name that does not comply with the rules, a red border or highlight can be added around this sentence to alert the reviewer.

[0100] In actual application scenarios, the parsing results and underwriting conclusions of the documents uploaded by customers can be directly located to the original text and presented in a visual form. This can not only reduce the error rate, but also greatly improve the interpretability and facilitate manual backtracking and re-inspection.

[0101] Therefore, the embodiment of the present application provides a more detailed and intuitive underwriting result. By marking text information that does not comply with the rules, auditors can locate the problem more quickly, thereby improving underwriting efficiency. At the same time, this also provides insurance companies with a more transparent and fair underwriting method, which helps to enhance customer trust.

[0102] In actual application scenarios, the method provided in the embodiments of the present application also includes a process of answering users' questions and inquiries.

[0103] Obtaining the user's query information, wherein the query information is used to represent the user's questions and / or needs regarding the underwriting process; generating reply information fed back to the user based on the query information, wherein the reply information is used to assist the user in submitting underwriting materials.

[0104] That is, the embodiment of the present application provides 7x24 hours of all-weather service based on the digital labor dialogue tool, guides customers to upload correct medical reports and imaging materials, and improves interaction efficiency and user experience.

[0105] In one possible implementation, based on the accumulated historical insurance policy claims records, potential patterns that have not been discovered in the underwriting rules are continuously summarized and concluded to update the underwriting rules.

[0106] The underwriting process provided by this application is explained from the perspective of functional modularization. Figure 2 , Figure 2 A flowchart of the second underwriting method provided in an embodiment of the present application.

[0107] Among them, step 1 represents that when applying for underwriting, if the customer is familiar with the underwriting process and related medical reports, he can directly submit the underwriting materials. Steps 2 and 3 represent that if the customer has questions about the insurance conditions or underwriting process, he can communicate his questions and needs online 24 hours a day through the digital labor of the intelligent communication module, and complete the submission of underwriting materials with the assistance of the digital labor;

[0108] Steps ④, ⑥ and ⑨ represent that after the underwriting materials are submitted, the intelligent review module can parse, identify and process the submitted materials (including pictures, documents and tables, etc.), and give the underwriting conclusion and basis based on semantic understanding, combined with prior insurance knowledge and the retrieval of the current underwriting rules for patients with diseases; in steps ⑤ and ④, if the uploaded materials cannot meet the information required by the underwriting rules, the intelligent review module can identify the missing information and the corresponding medical report, activate the intelligent communication module at the same time, and remind the customer to upload the supplementary materials in the dialog box. After the materials are uploaded, repeat steps ④ and ⑥ to complete the underwriting review;

[0109] Steps ⑦ and ⑧ represent that the intelligent underwriting rule iteration module collects and accumulates past underwriting conclusions and insurance policy claim records, regularly (for example, one to two months) summarizes and concludes the rules that have not been discovered in the current underwriting rules, and updates the underwriting rules.

[0110] The intelligent communication module is used to respond to customers' underwriting questions and needs in real time, and to assist customers in uploading complete underwriting materials; the intelligent review module is used to identify, parse and deeply understand the materials uploaded by customers, and to provide underwriting conclusions based on prior insurance knowledge and underwriting rules; the intelligent underwriting rule iteration module is used to continuously summarize and conclude potential patterns that have not been discovered in the underwriting rules based on the accumulated historical insurance policy claims records, and to update the underwriting rules.

[0111] In summary, the embodiment of the present application utilizes 7x24 hours digital labor to respond to customer questions and needs at any time throughout the day, and efficiently guide customers to complete the uploading of underwriting materials.

[0112] By combining a large model with a retrieval-augmented generation (RAG) engine, we can achieve efficient recognition, parsing, and deep semantic understanding of materials uploaded by customers (such as images, documents, tables, etc.), and provide underwriting conclusions while ensuring that the conclusions are well-founded.

[0113] Using an open source RAG framework such as txtai or RAGFlow, the processing process and results can be white-boxed. That is, all information extracted from customer uploaded documents can be directly traced back to the original text and can be displayed through a visual interface. This approach helps reduce the error rate, enhance the transparency and explainability of decision-making, and facilitate subsequent manual review or re-inspection.

[0114] Introducing a multi-channel recall mechanism during the retrieval process, that is, obtaining relevant information from multiple sources at the same time, and then improving the quality of the retrieval results through a re-ranking algorithm. This ensures the accuracy of data retrieval, thereby supporting the generation of higher-quality underwriting conclusions.

[0115] The above are some specific implementations of the underwriting method provided in the embodiment of the present application. Based on this, the present application also provides a corresponding underwriting system. The system provided in the embodiment of the present application will be introduced from the perspective of functional modularization. Figure 3 A structural diagram of an underwriting system provided in an embodiment of the present application.

[0116] The system comprises:

[0117] An acquisition unit 110 is used to acquire underwriting materials;

[0118] An identification unit 111 is used to identify the document structure of the underwriting material and perform layout segmentation on the underwriting material based on the document structure;

[0119] An analysis unit 112 is used to perform semantic analysis on the underwriting materials after layout segmentation based on a pre-trained language model to obtain an analysis result, wherein the pre-trained language model is a model trained based on historical underwriting information and used for performing semantic analysis on the underwriting materials;

[0120] The comparison unit 113 is used to compare the analysis result with the preset underwriting rules to obtain the underwriting result.

[0121] The embodiments of the present application also provide corresponding devices and computer storage media for implementing the test method solutions provided in the embodiments of the present application.

[0122] The device includes a memory and a processor, the memory is used to store instructions or codes, and the processor is used to execute the instructions or codes so that the device executes the test method described in any embodiment of the present application.

[0123] The computer storage medium stores codes, and when the codes are executed, a device executing the codes implements the test method described in any embodiment of the present application.

[0124] It should be noted that the various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments, and the same or similar parts between the various embodiments can be referred to each other. For the system or device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part description.

[0125] It should be understood that in the present application, "at least one (item)" means one or more, and "plurality" means two or more. "And / or" is used to describe the association relationship of associated objects, indicating that three relationships may exist. For example, "A and / or B" can mean: only A exists, only B exists, and A and B exist at the same time, where A and B can be singular or plural. The character " / " generally indicates that the objects associated before and after are in an "or" relationship. "At least one of the following" or similar expressions refers to any combination of these items, including any combination of single or plural items. For example, at least one of a, b or c can mean: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, c can be single or multiple.

[0126] It should be understood that the terms "center", "longitudinal", "lateral", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside" and the like indicate orientations or positional relationships based on the orientations or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the referred device or element must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be understood as a limitation on the present invention.

[0127] It should be noted that, unless otherwise clearly specified and limited, the terms "installed", "connected" and "connected" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection or an indirect connection through an intermediate medium, or it can be the internal communication of two components. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0128] It should also be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the presence of other identical elements in the process, method, article or device including the elements.

[0129] The steps of the method or algorithm described in conjunction with the embodiments disclosed herein may be implemented directly using hardware, a software module executed by a processor, or a combination of the two. The software module may be placed in a random access memory (RAM), a memory, a read-only memory (ROM), an electrically programmable ROM, an electrically erasable programmable ROM, a register, a hard disk, a removable disk, a CD-ROM, or any other form of storage medium known in the art.

[0130] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present application. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application will not be limited to the embodiments shown herein, but will conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. An underwriting method, characterized in that: include: Obtain underwriting materials; Identifying the document structure of the underwriting materials, and performing layout segmentation on the underwriting materials based on the document structure; Performing semantic analysis on the underwriting materials after layout segmentation based on a pre-trained language model to obtain an analysis result, wherein the pre-trained language model is a model trained based on historical underwriting information and used for performing semantic analysis on the underwriting materials; The analysis results are compared with preset underwriting rules to obtain underwriting results.

2. The method according to claim 1, characterized in that The layout-segmented underwriting materials include a plurality of segmented regions, and the semantic analysis is performed based on the pre-trained language model and the layout-segmented underwriting materials to obtain analysis results, including: Extracting text information that meets a preset criticality condition in at least one divided area; Based on the text information of the preset critical condition, calling the corresponding semantic analysis template; Based on the pre-trained language model and the semantic analysis template, semantic analysis is performed on the extracted text information to obtain an analysis result, and the semantic analysis template is used to guide the pre-trained language model to perform semantic analysis.

3. The method according to claim 2, characterized in that Before comparing the analysis result with the preset underwriting rules to obtain the underwriting result, the method further includes: Based on the text information of the preset critical conditions, the corresponding preset underwriting rules are called.

4. The method according to claim 3, characterized in that The step of comparing the analysis result with the preset underwriting rules to obtain the underwriting result includes: Adjusting the analysis result into a structured format corresponding to the preset underwriting rule; The analysis results in a structured format are compared with the preset underwriting rules to obtain the underwriting results.

5. The method according to claim 1, characterized in that The method further comprises: When the uploaded underwriting materials do not meet the preset material requirements, a prompt message is generated, which includes the reason why the uploaded underwriting materials do not meet the preset material requirements, and is used to remind the uploader to provide underwriting materials that meet the preset material requirements.

6. The method according to claim 1, characterized in that The underwriting result includes text information in the underwriting materials that does not comply with preset underwriting rules, and the method further includes: The text information in the underwriting materials after layout segmentation that does not comply with the preset underwriting rules is marked in a preset form.

7. The method according to claim 1, characterized in that The method further comprises: Acquire query information of the user, where the query information is used to represent the user's questions and / or needs regarding the underwriting process; Reply information is generated based on the query information and fed back to the user, and the reply information is used to assist the user in submitting underwriting materials.

8. An underwriting system, characterized in that: include: An acquisition unit, used to acquire underwriting materials; An identification unit, used to identify the document structure of the underwriting material, and perform layout segmentation on the underwriting material based on the document structure; An analysis unit, configured to perform semantic analysis on the underwriting materials after layout segmentation based on a pre-trained language model to obtain an analysis result, wherein the pre-trained language model is a model trained based on historical underwriting information and used for performing semantic analysis on the underwriting materials; The comparison unit is used to compare the analysis result with the preset underwriting rules to obtain the underwriting result.

9. An electronic device, characterized in that: include: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the underwriting method according to any one of claims 1 to 7 is implemented.

10. A computer program product, characterized in that When the computer program product runs on a computer, the computer executes the underwriting method according to any one of claims 1 to 7.

Images