Insurance policy input method and device, equipment and storage medium

Abstract

The invention belongs to the technical field of artificial intelligence, and discloses an insurance policy input method, device and equipment and a storage medium, and the method comprises the steps: accessing a preset insurance database, and constructing an insurance knowledge graph according to the preset insurance database; acquiring initial list recording information, and converting the initial list recording information into text graph structure information; processing the text graph structure information by using a graph neural network to generate intermediate recording information; and performing error correction on the intermediate recording information according to the insurance knowledge graph to obtain error information, and converting the intermediate recording information by adopting a sequence-to-sequence model according to the error information to generate target recording information. The insurance policy input method and device can be applied to use scenes of insurance policy input of finance, medical treatment and the like, and the accuracy and efficiency of insurance policy input are improved.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence, and in particular to a method, apparatus, device and storage medium for entering insurance policies. Background Art

[0002] With the rapid development of information technology, the billing process has gradually been moved online. This change has greatly improved the efficiency of sales representatives in issuing bills, shortened customer waiting time, and eliminated the need for customers to go to the site for processing, bringing significant operational optimization to the insurance and financial industries.

[0003] However, in the traditional online policy issuance process, for example, when entering insurance policies, sales staff need to manually input a large amount of key data such as customer information and policy details. This makes the traditional online policy issuance process not only time-consuming and labor-intensive, but also prone to input errors due to human factors, thereby increasing the complexity of subsequent business processing and may even cause claims disputes.

[0004] For example, in the financial field, many insurance companies have begun to implement electronic insurance policy systems. When purchasing insurance products, customers only need to complete information filling, underwriting, payment and other steps through the insurance company's official website or cooperative channels (such as banks). Then, the salesperson manually enters the customer's identity information, policy type, insurance period, insurance amount and other key data based on the information filled in by the customer. In the process of manual entry by the salesperson, input errors are likely to occur, resulting in the subsequent insurance company being unable to accurately identify customer information, or causing other claims disputes.

[0005] In summary, the existing online policy issuance process has the problems of low policy entry efficiency and low policy entry accuracy. Summary of the Invention

[0006] The purpose of the embodiments of the present application is to propose a policy entry method, device, equipment and storage medium, the main purpose of which is to improve the efficiency and accuracy of policy entry in the insurance policy issuance process.

[0007] First, in order to solve the above technical problems, the present application provides a method for entering insurance policies, which adopts the following technical solutions:

[0008] Accessing a preset insurance database and constructing an insurance knowledge graph based on the preset insurance database;

[0009] Obtaining initial order information, and converting the initial order information into text graph structure information;

[0010] Using a graph neural network to process the text graph structure information to generate intermediate order information;

[0011] The intermediate order information is corrected according to the insurance knowledge graph to obtain error information. Based on the error information, the intermediate order information is converted using a sequence-to-sequence model to generate target order information.

[0012] Secondly, in order to solve the above technical problems, the embodiments of the present application further provide a policy entry device, which adopts the following technical solution:

[0013] A graph construction module, used to access a preset insurance database and construct an insurance knowledge graph based on the preset insurance database;

[0014] The order information acquisition module is used to obtain initial order information and convert the initial order information into text graph structure information;

[0015] The model processing module uses a graph neural network to process the text graph structure information and generate intermediate order information;

[0016] An error information correction module is used to correct the intermediate order information according to the insurance knowledge graph to obtain error information. According to the error information, a sequence-to-sequence model is used to convert the intermediate order information to generate target order information.

[0017] On the third aspect, in order to solve the above-mentioned technical problems, an embodiment of the present application also provides a computer device, comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the insurance policy entry method as described above.

[0018] Fourthly, in order to solve the above-mentioned technical problems, an embodiment of the present application further provides a computer-readable storage medium storing a computer program, which implements the above-mentioned insurance policy entry method when executed by a processor.

[0019] Compared with the prior art, the embodiments of the present application have the following beneficial effects:

[0020] The insurance knowledge graph constructed based on the preset insurance database provides a basis for error correction of subsequent policy entry information, indirectly improving the accuracy of policy entry.

[0021] By converting the initial order information into text-graph structure information, it facilitates subsequent graph neural network recognition and improves the accuracy of insurance order recording; and through structured text-graph structure information, it facilitates rapid computer processing and improves the efficiency of the insurance order information process.

[0022] By using graph neural networks to generate intermediate order information, text graph structure information can be automatically processed, reducing manual intervention and error rates, and automatically extracting key information from the graph structure, thereby realizing automatic entry of insurance policies, improving the efficiency of policy entry, and increasing customer satisfaction. Graph neural networks can directly process graph structure data, and through information propagation and parameter updates between nodes, the accuracy of policy entry is improved.

[0023] The insurance knowledge graph can integrate data from different sources and structures. By performing error correction through the insurance knowledge graph, it can accurately identify errors in intermediate recording information, thereby improving the accuracy of error correction in the insurance knowledge graph and, in turn, improving the accuracy of insurance recording. Moreover, by combining the insurance knowledge graph with the sequence-to-sequence model, it is possible to achieve automated error correction of intermediate recording information, thereby improving the efficiency of policy entry. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] In order to more clearly illustrate the solutions in this application, a brief introduction will be given below to the drawings required for use in the description of the embodiments of this application. Obviously, the drawings described below are some embodiments of this application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0025] Figure 1 is an exemplary system architecture diagram to which the present application may be applied;

[0026] Figure 2 A flowchart of an embodiment of the insurance policy entry method according to the present application;

[0027] Figure 3 is a structural diagram of an embodiment of an insurance policy entry device according to the present application;

[0028] Figure 4 It is a structural diagram of an embodiment of a device according to the present application. DETAILED DESCRIPTION

[0029] Unless otherwise defined, all technical and scientific terms used herein have the same meanings as commonly understood by those skilled in the art to which this application belongs. The terms used in the specification of the application are for the purpose of describing specific embodiments only and are not intended to limit this application. The terms "including" and "having" and any variations thereof in the specification and claims of this application and the above-mentioned drawings are intended to cover non-exclusive inclusions. The terms "first", "second", etc. in the specification and claims of this application or the above-mentioned drawings are used to distinguish different objects, not to describe a specific order.

[0030] References herein to "embodiments" mean that a particular feature, structure, or characteristic described in connection with the embodiments may be included in at least one embodiment of the present application. The appearance of this phrase in various places in the specification does not necessarily refer to the same embodiment, nor does it constitute an independent or alternative embodiment that is mutually exclusive of other embodiments. It is understood, both explicitly and implicitly, by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0031] 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 accompanying drawings.

[0032] like Figure 1 As shown, system architecture 100 may include a terminal device 101, a network 102, and a server 103. Terminal device 101 may be a laptop computer 1011, a tablet computer 1012, or a mobile phone 1013. Network 102 is a medium for providing a communication link between terminal device 101 and server 103. Network 102 may include various connection types, such as wired or wireless communication links or fiber optic cables.

[0033] The user can use the terminal device 101 to interact with the server 103 via the network 102 to receive or send messages, etc. Various communication client applications can be installed on the terminal device 101, such as web browser applications, shopping applications, search applications, instant messaging tools, email clients, social platform software, etc.

[0034] The terminal device 101 can be various electronic devices with a display screen and supporting web browsing. In addition to the laptop computer 1011, tablet computer 1012 or mobile phone 1013, the terminal device 101 can also be an e-book reader, an MP3 player (Moving Picture Experts Group Audio Layer III), an MP4 (Moving Picture Experts Group Audio Layer IV) player, a laptop computer and a desktop computer, etc.

[0035] The server 103 may be a server that provides various services, such as a background server that provides support for web pages displayed on the terminal device 101 .

[0036] It should be noted that the insurance policy entry method provided in the embodiment of the present application is generally executed by a server / terminal device, and accordingly, the insurance policy entry device is generally set in the server / terminal device.

[0037] It should be understood that Figure 1The number of terminal devices, networks and servers in the embodiment is merely illustrative. Any number of terminal devices, networks and servers may be provided as required.

[0038] Continue to refer Figure 2 , shows a flow chart of an embodiment of the insurance policy entry method according to the present application. According to different requirements, the order of the steps in the flow chart can be changed, and some steps can be omitted. The insurance policy entry method provided in the embodiment of the present application can be applied to any scenario that requires insurance policy entry, and the insurance policy entry method can be applied to products in these scenarios. The insurance policy entry method includes the following steps:

[0039] Step S201: Access a preset insurance database and construct an insurance knowledge graph based on the preset insurance database.

[0040] In this embodiment, the insurance policy entry method is executed on the electronic device (eg Figure 1 The server / terminal device shown in the figure) can access the insurance database and the initial order information mentioned above through a wired connection or a wireless connection. It should be noted that the above-mentioned wireless connection method may include but is not limited to 3G / 4G / 5G connection, WiFi connection, Bluetooth connection, WiMAX connection, Zigbee connection, UWB (ultra wideband) connection, and other wireless connection methods currently known or to be developed in the future.

[0041] In this embodiment, the preset insurance database refers to a database that stores a large amount of insurance-related data and information, including insurance product information (such as the name, type, coverage, premium, insured amount and other detailed information of the insurance product), customer information (such as customer name, age, gender, contact information, occupation, and customer's insurance purchase history, claim records, etc.), claim data (including detailed information of all claim cases, such as claim type, claim amount, etc.), insurance terms and laws and regulations (such as the terms and contents of various insurance products, as well as relevant insurance laws and regulations and regulatory policies), industry reports and statistical data (including market analysis reports and statistical data of the insurance industry, etc.). After accessing the preset insurance database, the insurance data information of the insurance database is obtained, and an insurance knowledge graph is constructed based on the insurance data information.

[0042] In one embodiment, constructing an insurance knowledge graph based on the preset insurance database includes:

[0043] Obtaining insurance data information from the preset insurance database, performing entity recognition on the insurance data information, and obtaining at least one graph entity;

[0044] Analyzing the semantic relationship of each of the graph entities, and performing relationship extraction on at least one of the graph entities according to the semantic relationship to obtain a graph relationship;

[0045] The insurance knowledge graph is constructed by taking the graph entities as nodes and the graph relationships as edges.

[0046] In this embodiment, the required insurance-related insurance data information (such as insurance product information, insurance claim information, etc.) is extracted from a preset insurance database, and the extracted insurance data information is preprocessed. The preprocessing includes but is not limited to data cleaning of the insurance data information, removing duplicate, erroneous and invalid data, filling missing values and converting data formats, etc.; according to the preset entity table, the entity type to be identified is clearly defined, such as insurance products, claim procedures, etc., and then natural language processing technology (such as named entity recognition NER) is used to perform entity recognition on the insurance data information, and the entities to be identified are extracted from the insurance data information. The extracted entities are then cleaned to remove duplicate or erroneous entities to obtain at least one graph entity; according to the preset relationship table, the semantic relationship between the graph entities is defined, and according to the preset relationship table, the relationship between the graph entities is extracted to obtain the graph relationship, and finally the graph entities are used as nodes of the insurance knowledge graph, and the graph relationship is used as the edge of the insurance knowledge graph to construct the insurance knowledge graph.

[0047] In this embodiment, the insurance knowledge graph constructed based on the preset insurance database can provide a basis for correcting errors in subsequent policy entry information, thereby indirectly improving the accuracy of policy entry.

[0048] Step S202: Acquire initial order information, and convert the initial order information into text graph structure information.

[0049] In this embodiment, the above-mentioned initial order recording information refers to the salesperson's voice description or text description of the insurance order recording information, including the customer's basic information (such as name, contact information, address, etc.), detailed information of the insurance product (such as insurance product type, insurance product name, premium, etc.), and related additional information (such as insurance time, effective date, etc.); after receiving the initial order recording information uploaded by the salesperson, the initial order recording information is pre-processed, wherein the pre-processing operation includes word segmentation, part-of-speech tagging, and named entity recognition of the text, and the named entity recognition includes but is not limited to identifying names of people, places, product names, etc. in the text, and then the initial order recording information after the pre-processing operation is subjected to grammatical relationship analysis, and the initial order recording information is converted into text graph structure information according to the analyzed grammatical relationship.

[0050] In one embodiment, after obtaining the initial order information, the method further includes:

[0051] Identifying the type of the initial order information;

[0052] When it is identified that the type of the initial recording information is a voice type, the initial recording information is converted into a text form using voice conversion technology.

[0053] In this embodiment, after receiving the initial order information uploaded by the salesperson, the type of the initial order information is identified by identifying the file extension of the initial order information. When the extension of the initial order information is identified as .wav, etc., the type of the initial order information is judged to be an audio type, and voice conversion technology (such as Speech-to-Text API, etc.) is used to convert the initial order information into text form.

[0054] In this embodiment, by using a variety of forms to input the initial order information, the convenience and flexibility of the salesperson's input are improved, the efficiency of subsequent insurance order information recording is improved, and the salesperson's user experience is improved.

[0055] In one embodiment, converting the initial order information into text graph structure information includes:

[0056] Performing text preprocessing on the initial order information, wherein the text preprocessing includes word segmentation, stop word removal, and part-of-speech tagging;

[0057] Based on the initial order information after the text preprocessing operation, the initial order information is identified using a named entity recognition technology to obtain multiple text entities;

[0058] Relationship identification is performed on the plurality of text entities using a preset natural language processing tool to obtain entity relationships among the plurality of text entities;

[0059] The text graph structure information is constructed by using the multiple text entities as nodes of the text graph structure and the multiple entity relationships as edges.

[0060] In this embodiment, a text preprocessing operation is performed on the initial order entry information. The text preprocessing operation includes word segmentation, stop word removal, and词性标注 (pos tagging). Specifically, the text preprocessing operation uses a word segmentation tool (such as jieba) to perform word segmentation on the initial order entry information, splitting the continuous initial order entry information into independent words. Then, according to a preset stop word list, stop words in the initial order entry information, such as "de", "le", etc., are filtered out. Finally, a pos tagging tool is used to perform pos tagging on the word-segmented initial order entry information, completing the text preprocessing operation on the initial order entry information. Based on the initial order entry information after the text preprocessing operation, a named entity recognition tool (such as the NER module of jieba) is used to identify multiple text entities (such as person names, insurance product names, etc.) in the initial order entry information. After identifying the multiple text entities in the initial order entry information, through a preset natural language processing tool, such as a dependency parser (such as the Stanford Parser in NLTK), a deep learning model, etc., relationship recognition is performed on the multiple text entities. According to the parameters configured for the natural language processing tool, the entity relationships (such as syntactic relationships, semantic relationships) between the multiple text entities are identified. After obtaining the multiple text entities and the relationships between the multiple text entities, the text entities are used as nodes of the text graph structure information, and the relationships between the text entities are used as edges to construct the text graph structure information.

[0061] In this embodiment, by performing entity recognition and entity relationship recognition on the initial order entry information and constructing text graph structure information based on the recognized entities and entity relationships, the clarity of the initial insurance order entry information can be improved, providing a basis for subsequent insurance order entry information input and improving the accuracy of the policy entry process.

[0062] It should be noted that the Chinese term "词性标注" in the original text is not translated as there is no English equivalent provided in the rules. You may need to correct it according to the actual English term for "词性标注" in a real scenario.In the financial field, an implementable example A is that a salesperson receives information from a customer named Zhang San on an insurance company's online platform: "My name is Zhang San, I'm 35 years old, my phone number is 138xxxx8888, and I want to buy a critical illness insurance policy with a coverage of 2 million yuan." The salesperson enters Zhang San's information into the system as initial order information through oral or text input, and performs text preprocessing on the initial order information: I (r) / name (v) / Zhang San (PER) / , (w) / this year (TIME) / 35 years old (m) / , (w) / contact number (n) / is (v) / 138xxxx8888 (m) / . (w) / I (r) / want (v) / to buy (v) / a (m) / critical illness insurance policy (nz) / , (w) / coverage (n) / 2 million yuan (m) / . (w)". Based on the pre-processed initial order information, named entity recognition technology is used to identify the text entities corresponding to the initial order information: "Zhang San" is the person's name, the age is 35 years old, "138xxxx8888" is the phone number, the quantity is one, "Major Illness Insurance" is the insurance product name, and the insured amount is 2 million yuan. Natural language processing tools are used to analyze the relationship between each entity, such as the relationship between Zhang San and 35 years old. The text entities are used as nodes and the relationships between text entities as edges to convert the initial order information into text graph structure information.

[0063] In this embodiment, by converting the initial record information into text graph structure information, it is facilitated by the subsequent graph neural network recognition, thereby improving the accuracy of the insurance record; and the structured text graph structure information facilitates rapid computer processing, thereby improving the efficiency of the insurance record information process.

[0064] Step S203: Use a graph neural network to process the text graph structure information to generate intermediate order information.

[0065] In this embodiment, the above-mentioned graph neural network refers to a deep learning model GNN (Graph Neural Network), which focuses on processing graph structure data. The graph neural network includes an input layer, a multi-layer graph convolution layer and an output layer. The intermediate order information is standardized order information generated based on the initial order information; the text graph structure information is processed by the graph neural network to generate standardized intermediate order information, and the intermediate order information is stored in a preset intermediate database for subsequent extraction and processing.

[0066] In one embodiment, the processing of the text graph structure information using a graph neural network to generate intermediate order information includes:

[0067] Identify the nodes of the text graph structure information, extract the feature vector of each node, and obtain an initial feature vector corresponding to each node;

[0068] Inputting all the initial feature vectors into the graph neural network through the input layer of the graph neural network;

[0069] Based on each of the initial feature vectors, collecting neighbor node information of the initial feature vector through the graph convolution layer of the graph neural network, and fusing the initial feature vector with the neighbor node information to obtain a target feature vector;

[0070] The target feature vector is converted into structured data through the output layer of the graph neural network to obtain the intermediate order information.

[0071] In this embodiment, all nodes are identified from the text graph structure information, and each of the nodes is converted into a feature vector through vector conversion technology (such as word2Vec, etc.) to obtain the initial feature vector corresponding to each node; based on all the initial feature vectors, all the initial feature vectors are combined into a matrix to obtain an input matrix; the input matrix is input into the graph neural network through the input layer of the graph neural network, wherein the input layer is a simple linear transformation layer for mapping the input matrix to the internal representation space of the graph neural network; for the initial feature vector corresponding to each node, the information of the neighboring nodes is collected through the graph convolution layer, wherein the neighboring node refers to the node that has a connection relationship with the node in the input matrix, and the collected information of the neighboring nodes is fused with the initial feature vector, wherein the fusion method includes but is not limited to weighted average, summation, etc., and the initial feature vector corresponding to each node is processed by stacking multiple graph convolution layers to finally obtain the target feature vector, and the intermediate record information is output through the output layer of the graph neural network.

[0072] In this embodiment, the graph neural network can effectively capture the complex relationship of text graph structural information, and the initial feature vector of each node is fused with the neighbor node information, so that the extracted features are more comprehensive and accurate, thereby improving the accuracy of information extraction; and the output layer of the graph neural network converts the target feature vector into structured data to obtain intermediate order information, making subsequent processing more convenient and efficient.

[0073] In one embodiment, after processing the text graph structure information using a graph neural network to generate intermediate order information, the method further includes:

[0074] Parsing the intermediate order information to obtain the main insurance type and insured person characteristics;

[0075] Based on the insurance knowledge graph, screening out a first set of supplementary insurances related to the main insurance type according to the main insurance type;

[0076] Filtering the first set of supplementary insurance policies according to the insured's characteristics to obtain a second set of supplementary insurance policies;

[0077] According to the amount factor of each supplementary insurance in the second supplementary insurance set, the second supplementary insurance set is sorted, and a supplementary insurance set with a preset ranking is selected from the sorted second supplementary insurance set as the target supplementary insurance set, and the target supplementary insurance set is pushed.

[0078] In this embodiment, the intermediate order information is first parsed to extract key information such as the main insurance type and the insured's characteristics. The main insurance type refers to the insurance type in the intermediate order information, such as auto insurance, health insurance, life insurance, etc. The insured's characteristics refer to age, gender, health status, and occupation. First, based on the insurance knowledge graph constructed in step S201, the main insurance type is used as the starting node, and according to the relationship established in the insurance knowledge graph, the first supplementary insurance set is screened out. On the basis of the first supplementary insurance set, further screening is performed according to the insured's characteristics. For example, if the main insurance type is medical insurance, then for older insureds, Cancer insurance, critical illness insurance, etc. can be screened out. For insured persons in high-risk occupations, accidental injury insurance can be recommended. For example, if the gender is female, female-specific disease insurance can be screened out to obtain the first supplementary insurance set; since the amount factor is an important aspect that customers consider when choosing supplementary insurance, the second supplementary insurance set is sorted according to the amount factor (such as premium, compensation amount, etc.) to obtain the sorted second supplementary insurance set, and the top N supplementary insurances are selected from the sorted second supplementary insurance set as the target supplementary insurance set, and the target supplementary insurance is pushed. After the customer selects the target supplementary insurance, the target supplementary insurance is added to the intermediate recording information.

[0079] In this embodiment, by parsing the intermediate order recording message and obtaining the main insurance type and the insured's characteristics, personalized supplementary insurance can be pushed according to the specific situation of the insured, thereby improving the accuracy of the supplementary insurance push; by using the insurance knowledge graph for push, supplementary insurance related to the main insurance type can be quickly screened out, reducing the time for manual screening and improving the efficiency of supplementary insurance recommendations.

[0080] Continuing with the above-mentioned feasible example A, the above-mentioned text structured information is processed by a graph neural network to generate intermediate order information. The intermediate order information can be stored in a structured data format:

[0081]

[0082]

[0083] Based on the above data, the target supplementary insurance is screened out. When the user selects the target supplementary insurance, the target supplementary insurance is added to the intermediate order information.

[0084] In this embodiment, by using graph neural networks to generate intermediate recording information, text graph structure information can be automatically processed, manual intervention and error rate can be reduced, key information in the graph structure can be automatically extracted, and automatic entry of insurance policies can be realized, thereby improving the efficiency of insurance policy entry and improving customer satisfaction; graph neural networks can directly process graph structure data, and the accuracy of insurance policy entry can be improved through information propagation and parameter updates between nodes.

[0085] Step S204: Correct the intermediate order information according to the insurance knowledge graph to obtain error information. Based on the error information, convert the intermediate order information using a sequence-to-sequence model to generate target order information.

[0086] In this embodiment, the intermediate order information is corrected according to the insurance knowledge graph, and the intermediate order information is input into the insurance knowledge graph. The nodes and edges in the graph are used to verify the accuracy of the intermediate order information. When errors or omissions are found in the intermediate order information, such as the insurance product selected by the insured does not comply with the regulations, the insurance amount filled in does not match the insurance amount specified in the insurance product, etc., the intermediate order information is marked with a type to obtain error information; based on the error information, the sequence-to-sequence model is used to replace the intermediate order information. When the replacement is completed, the target order information is generated, and the recorded order information is fed back to the salesperson for final confirmation. After confirmation, the system generates a formal insurance policy document based on the entered data and sends it to the customer for signing and archiving. The above-mentioned sequence-to-sequence (Seq2Seq) model is a deep model for processing sequence data. It can convert an input sequence of variable length into an output sequence of variable length. In this embodiment, the sequence-to-sequence model is used to convert the erroneous input sequence into a correct output sequence.

[0087] In one embodiment, before correcting the intermediate order information according to the insurance knowledge graph, the method further includes:

[0088] Correcting the intermediate record information using a dictionary to obtain the intermediate record information after dictionary correction;

[0089] The intermediate record information after the dictionary error correction is calculated using a random forest model to obtain the probability of an erroneous sentence;

[0090] When the probability of the erroneous statement exceeds a preset error threshold, the erroneous statement is marked to obtain the error information.

[0091] In this embodiment, the intermediate record information is preliminarily corrected through the dictionary, and the spelling and grammar of the intermediate record information are corrected by using the dictionary, and the correct spelling information in the dictionary is used to replace it to obtain the intermediate record information after the dictionary correction. The intermediate record information after the dictionary correction is calculated using the random forest model to calculate the probability of incorrect sentences in the intermediate record information after the dictionary correction. Specifically, the intermediate record information is text-vectorized and the Bag of Words model is used. The method converts the intermediate record information into numerical features using methods such as word frequency (such as word frequency, part of speech, etc.), TF-IDF (term frequency-inverse document frequency), or word embedding (such as Word2Vec, BERT, etc.), extracts key features from the numerical features, and the key features include but are not limited to lexical features (such as word frequency, part of speech, etc.), grammatical features (such as sentence structure, dependency relationship, etc.), and sets the parameters of the random forest model, such as the number of decision trees, maximum depth, minimum number of samples, etc.; inputs the key features into a pre-trained random forest model, calculates the error probability of the sentence corresponding to the key features based on the input key features, and when the probability of the erroneous sentence is greater than a preset error threshold, marks the erroneous sentence with a type to obtain the error information.

[0092] In this embodiment, the error information obtained by the random forest model and the error information obtained by the insurance knowledge graph are both converted through a sequence-to-sequence model.

[0093] In this embodiment, the dual error correction of the dictionary and the random forest model can improve the comprehensiveness of the error correction and thus improve the accuracy of the insurance policy entry.

[0094] In one embodiment, converting the intermediate order information using a sequence-to-sequence model based on the error information to generate target order information includes:

[0095] Identify the error type and the position of the error information in the intermediate order information according to the type mark of the error information;

[0096] Encoding the error information through the encoder of the sequence-to-sequence model to obtain an error information code;

[0097] According to the position of the error information in the intermediate recording information, context information is obtained, and according to the context information and the error type, the error information is decoded by the sequence-to-sequence decoder to generate the target recording information.

[0098] In this embodiment, the error type of the error information and the position of the error information in the intermediate recording information are determined according to the above-mentioned type mark when identifying the error information, and the error information is input into the sequence-to-sequence model. The error information is encoded by the encoder of the sequence-to-sequence model and encoded into a vector representation. Then, according to the position of the error information and the predefined length, the text before and after the error information is extracted from the intermediate recording information as context information, and the error coding information and the context information are passed to the decoder. The decoder predicts the character or word sequence for generating the target recording information based on the error coding information and the context information. When the decoder completes all steps, it outputs the complete target recording information.

[0099] In this embodiment, the sequence-to-sequence model can be used to automatically process errors in intermediate recording information without the need for manual checking and correction of each item, thereby improving the efficiency of policy entry; and the sequence-to-sequence model can provide contextual information and error types to accurately identify and correct errors in intermediate recording information, reducing misjudgments caused by manual corrections and improving the accuracy of policy entry.

[0100] Continuing with the above-mentioned feasible example A, the intermediate order information in the form of structured data is corrected through the dictionary, random forest model, and insurance knowledge graph. When there is erroneous information, the erroneous statement is replaced through the sequence-to-sequence model to generate complete target order information. The target order information is then fed back to the salesperson for confirmation. Once the salesperson confirms that it is correct, the policy entry is completed.

[0101] In this embodiment, the insurance knowledge graph can integrate data from different sources and different structures, and perform error correction through the insurance knowledge graph, which can accurately identify errors in intermediate recording information, thereby improving the accuracy of error correction of the insurance knowledge graph and thus improving the accuracy of insurance recording; and through the combination of the insurance knowledge graph and the sequence-to-sequence model, automatic error correction of intermediate recording information can be achieved, thereby improving the efficiency of policy entry.

[0102] It should be emphasized that in order to further ensure the privacy and security of the above-mentioned order recording information, the above-mentioned order recording information can also be stored in a blockchain node.

[0103] The blockchain referred to in this application is a new application model for computer technologies such as distributed data storage, peer-to-peer transmission, consensus mechanisms, and encryption algorithms. Blockchain is essentially a decentralized database, a series of data blocks generated using cryptographic methods. Each data block contains information about a batch of network transactions, which is used to verify the validity of this information (to prevent counterfeiting) and generate the next block. Blockchain can include the underlying blockchain platform, the platform product service layer, and the application service layer.

[0104] The embodiments of the present application can acquire and process relevant data based on artificial intelligence technology. Artificial Intelligence (AI) is the theory, method, technology, and application system that uses digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use knowledge to achieve optimal results.

[0105] Fundamental AI technologies generally include sensors, dedicated AI chips, cloud computing, distributed storage, big data processing, operating / interaction systems, and mechatronics. AI software technologies primarily encompass computer vision, robotics, biometrics, speech processing, natural language processing, and machine learning / deep learning.

[0106] Those skilled in the art will appreciate that all or part of the processes in the above-described method embodiments can be implemented by instructing related hardware via computer-readable instructions. The computer-readable instructions can be stored in a computer-readable storage medium, and when the program is executed, it can include the processes in the above-described method embodiments. The aforementioned storage medium can be a non-volatile storage medium such as a magnetic disk, an optical disk, a read-only memory (ROM), or a random access memory (RAM).

[0107] It should be understood that although the steps in the flowcharts of the accompanying drawings are shown in sequence as indicated by the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some of the steps in the flowcharts of the accompanying drawings may include multiple sub-steps or multiple stages, and these sub-steps or stages are not necessarily executed at the same time, but can be executed at different times, and their execution order is not necessarily sequential, but can be executed in turn or alternately with other steps or at least a portion of the sub-steps or stages of other steps.

[0108] Further references Figure 3 , as a response to the above Figure 2 In order to realize the method shown in the figure, the present application provides an embodiment of a device for entering insurance policies. Figure 2 Corresponding to the method embodiment shown, the apparatus can be specifically applied to various computer devices.

[0109] like Figure 3As shown, the insurance policy entry device 300 of this embodiment includes: a graph construction module 301, an insurance policy information acquisition module 302, a model processing module 303, and an error information correction module 304. Among them:

[0110] A graph construction module 301 is used to access a preset insurance database and construct an insurance knowledge graph based on the preset insurance database;

[0111] In one embodiment, the graph construction module includes:

[0112] An entity recognition submodule is configured to obtain insurance data information from the preset insurance database, perform entity recognition on the insurance data information, and obtain at least one graph entity;

[0113] A relationship extraction submodule is used to analyze the semantic relationship of each of the graph entities, and extract the relationship of at least one of the graph entities according to the semantic relationship to obtain a graph relationship;

[0114] The graph construction submodule is used to construct the insurance knowledge graph using the graph entities as nodes and the graph relationships as edges.

[0115] The order information acquisition module 302 is used to acquire initial order information and convert the initial order information into text graph structure information;

[0116] In one embodiment, the order information acquisition module includes:

[0117] A preprocessing submodule, configured to perform text preprocessing operations on the initial order record information, wherein the text preprocessing operations include word segmentation, stop word removal, and part-of-speech tagging;

[0118] An entity recognition submodule is configured to recognize the initial order information after the text preprocessing operation using a named entity recognition technology to obtain a plurality of text entities;

[0119] An entity relationship identification submodule, configured to perform relationship identification on the plurality of text entities using a preset natural language processing tool to obtain entity relationships among the plurality of text entities;

[0120] The graph structure information construction submodule is used to construct the text graph structure information by taking the multiple text entities as nodes of the text graph structure and the multiple entity relationships as edges.

[0121] The model processing module 303 processes the text graph structure information using a graph neural network to generate intermediate order information;

[0122] In one embodiment, the model processing module includes:

[0123] A node feature extraction submodule is used to identify the nodes of the text graph structure information, extract the feature vector of each node, and obtain the initial feature vector corresponding to each node;

[0124] A model input submodule, configured to input all the initial feature vectors into the graph neural network through the input layer of the graph neural network;

[0125] A graph convolution submodule is used to collect neighbor node information of each initial feature vector through the graph convolution layer of the graph neural network, and fuse the initial feature vector with the neighbor node information to obtain a target feature vector;

[0126] The feature conversion submodule is used to convert the target feature vector into structured data through the output layer of the graph neural network to obtain the intermediate order information.

[0127] In one embodiment, the apparatus further comprises:

[0128] A parsing module, configured to parse the intermediate order information to obtain the primary insurance type and insured person characteristics;

[0129] A first screening module is configured to screen out a first set of supplementary insurances related to the main insurance type based on the insurance knowledge graph and the main insurance type;

[0130] a second screening module, configured to screen the first set of supplementary insurance policies according to the characteristics of the insured, to obtain a second set of supplementary insurance policies;

[0131] The push module is used to sort the second supplementary insurance set according to the amount factor of each supplementary insurance in the second supplementary insurance set, select the supplementary insurance set with a preset ranking from the sorted second supplementary insurance set as the target supplementary insurance set, and push the target supplementary insurance set.

[0132] The error information correction module 304 is used to correct the intermediate order information according to the insurance knowledge graph to obtain error information, and convert the intermediate order information using a sequence-to-sequence model based on the error information to generate target order information.

[0133] In one embodiment, the apparatus further comprises:

[0134] A dictionary error correction module, configured to correct the intermediate order information using a dictionary to obtain the intermediate order information after dictionary error correction;

[0135] A model error correction module, configured to calculate the intermediate record information after the dictionary error correction using a random forest model to obtain the probability of an erroneous sentence;

[0136] The marking module is used to mark the erroneous statement when the probability of the erroneous statement exceeds a preset error threshold to obtain the error information.

[0137] In one embodiment, the error information correction module includes:

[0138] An information identification submodule, configured to identify the error type and the position of the error information in the intermediate order information according to the type mark of the error information;

[0139] an encoding submodule, configured to encode the error information through an encoder of the sequence-to-sequence model to obtain an error information code;

[0140] The decoding submodule is used to obtain context information according to the position of the error information in the intermediate recording information, and decode the error information through the sequence-to-sequence decoder according to the context information and the error type to generate the target recording information.

[0141] In this embodiment, the insurance knowledge graph constructed based on the preset insurance database can provide a basis for error correction of subsequent policy information, indirectly improving the accuracy of policy entry.

[0142] By converting the initial order information into text-graph structure information, it facilitates subsequent graph neural network recognition and improves the accuracy of insurance order recording; and through structured text-graph structure information, it facilitates rapid computer processing and improves the efficiency of the insurance order information process.

[0143] By using graph neural networks to generate intermediate order information, text graph structure information can be automatically processed, reducing manual intervention and error rates, and automatically extracting key information from the graph structure, thereby realizing automatic entry of insurance policies, improving the efficiency of policy entry, and increasing customer satisfaction. Graph neural networks can directly process graph structure data, and through information propagation and parameter updates between nodes, the accuracy of policy entry is improved.

[0144] The insurance knowledge graph can integrate data from different sources and structures. By performing error correction through the insurance knowledge graph, it can accurately identify errors in intermediate recording information, thereby improving the accuracy of error correction in the insurance knowledge graph and, in turn, improving the accuracy of insurance recording. Moreover, by combining the insurance knowledge graph with the sequence-to-sequence model, it is possible to achieve automated error correction of intermediate recording information, thereby improving the efficiency of policy entry.

[0145] In order to solve the above technical problems, the embodiment of the present application also provides a device (computer device). Figure 4 , Figure 4 This is a basic structural block diagram of the computer device in this embodiment.

[0146] The computer device 4 includes a memory 41, a processor 42, and a network interface 43 that are interconnected through a system bus. It should be noted that the figure only shows a computer device 4 with a memory 41, a processor 42, and a network interface 43, but it should be understood that it is not required to implement all the components shown, and more or fewer components can be implemented instead. Among them, those skilled in the art can understand that the computer device here is a device that can automatically perform numerical calculations and / or information processing according to pre-set or stored instructions, and its hardware includes but is not limited to a microprocessor, an application specific integrated circuit (ASIC), a field-programmable gate array (FPGA), a digital signal processor (DSP), an embedded device, etc.

[0147] The computer device may be a desktop computer, notebook computer, PDA, cloud server, etc. The computer device may interact with the user via a keyboard, mouse, remote control, touchpad, or voice control device.

[0148] The memory 41 includes at least one type of readable storage medium, including flash memory, a hard disk, a multimedia card, a card-type memory (e.g., SD or DX memory), random access memory (RAM), static random access memory (SRAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), programmable read-only memory (PROM), magnetic storage, a magnetic disk, an optical disk, etc. In some embodiments, the memory 41 may be an internal storage unit of the computer device 4, such as the hard disk or memory of the computer device 4. In other embodiments, the memory 41 may also be an external storage device of the computer device 4, such as a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash memory card, etc. Of course, the memory 41 may also include both the internal storage unit of the computer device 4 and its external storage device. In this embodiment, the memory 41 is generally used to store the operating system and various application software installed on the computer device 4, such as computer-readable instructions for the insurance policy entry method. In addition, the memory 41 can also be used to temporarily store various types of data that have been output or are to be output.

[0149] In some embodiments, the processor 42 may be a central processing unit (CPU), a controller, a microcontroller, a microprocessor, or other data processing chip. The processor 42 is generally used to control the overall operation of the computer device 4. In this embodiment, the processor 42 is used to execute computer-readable instructions stored in the memory 41 or process data, such as computer-readable instructions for executing the insurance policy entry method.

[0150] The network interface 43 may include a wireless network interface or a wired network interface. The network interface 43 is generally used to establish a communication connection between the computer device 4 and other electronic devices.

[0151] During implementation, the electronic device of this application can provide a basis for error correction of subsequent policy entry information by constructing an insurance knowledge graph based on a preset insurance database, thereby indirectly improving the accuracy of policy entry.

[0152] By converting the initial order information into text-graph structure information, it facilitates subsequent graph neural network recognition and improves the accuracy of insurance order recording; and through structured text-graph structure information, it facilitates rapid computer processing and improves the efficiency of the insurance order information process.

[0153] By using graph neural networks to generate intermediate order information, text graph structure information can be automatically processed, reducing manual intervention and error rates, and automatically extracting key information from the graph structure, thereby realizing automatic entry of insurance policies, improving the efficiency of policy entry, and increasing customer satisfaction. Graph neural networks can directly process graph structure data, and through information propagation and parameter updates between nodes, the accuracy of policy entry is improved.

[0154] The insurance knowledge graph can integrate data from different sources and structures. By performing error correction through the insurance knowledge graph, it can accurately identify errors in intermediate recording information, thereby improving the accuracy of error correction in the insurance knowledge graph and, in turn, improving the accuracy of insurance recording. Moreover, by combining the insurance knowledge graph with the sequence-to-sequence model, it is possible to achieve automated error correction of intermediate recording information, thereby improving the efficiency of policy entry.

[0155] The present application also provides another embodiment, namely, providing a storage medium (computer-readable storage medium), wherein the computer-readable storage medium stores computer-readable instructions, and the computer-readable instructions can be executed by at least one processor to enable the at least one processor to perform the steps of the insurance policy entry method as described above.

[0156] During implementation, the computer-readable storage medium of the present application can provide a basis for error correction of subsequent policy entry information by constructing an insurance knowledge graph based on a preset insurance database, thereby indirectly improving the accuracy of policy entry.

[0157] By converting the initial order information into text-graph structure information, it facilitates subsequent graph neural network recognition and improves the accuracy of insurance order recording; and through structured text-graph structure information, it facilitates rapid computer processing and improves the efficiency of the insurance order information process.

[0158] By using graph neural networks to generate intermediate order information, text graph structure information can be automatically processed, reducing manual intervention and error rates, and automatically extracting key information from the graph structure, thereby realizing automatic entry of insurance policies, improving the efficiency of policy entry, and increasing customer satisfaction. Graph neural networks can directly process graph structure data, and through information propagation and parameter updates between nodes, the accuracy of policy entry is improved.

[0159] The insurance knowledge graph can integrate data from different sources and structures. By performing error correction through the insurance knowledge graph, it can accurately identify errors in intermediate recording information, thereby improving the accuracy of error correction in the insurance knowledge graph and, in turn, improving the accuracy of insurance recording. Moreover, by combining the insurance knowledge graph with the sequence-to-sequence model, it is possible to achieve automated error correction of intermediate recording information, thereby improving the efficiency of policy entry.

[0160] The non-Company software tools or components appearing in the embodiments of this application are merely examples and do not represent actual use.

[0161] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus the necessary general hardware platform, and of course can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk), and includes a number of instructions for enabling a terminal device (which can be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the methods described in each embodiment of the present application.

[0162] Obviously, the embodiments described above are only some of the embodiments of the present application, rather than all of the embodiments. The preferred embodiments of the present application are given in the accompanying drawings, but they do not limit the patent scope of the present application. The present application can be implemented in many different forms. On the contrary, the purpose of providing these embodiments is to make the understanding of the disclosure of the present application more thorough and comprehensive. Although the present application has been described in detail with reference to the aforementioned embodiments, for those skilled in the art, it is still possible to modify the technical solutions described in the aforementioned specific embodiments, or to make equivalent replacements for some of the technical features therein. Any equivalent structure made using the contents of the present application specification and the accompanying drawings, directly or indirectly used in other related technical fields, is also within the scope of patent protection of the present application.

Claims

1. A method for entering an insurance policy, characterized in that: The method comprises: Accessing a preset insurance database and constructing an insurance knowledge graph based on the preset insurance database; Obtaining initial order information, and converting the initial order information into text graph structure information; Using a graph neural network to process the text graph structure information to generate intermediate order information; The intermediate order information is corrected according to the insurance knowledge graph to obtain error information. Based on the error information, the intermediate order information is converted using a sequence-to-sequence model to generate target order information.

2. The insurance policy entry method according to claim 1, wherein: The constructing of the insurance knowledge graph according to the preset insurance database includes: Obtaining insurance data information from the preset insurance database, performing entity recognition on the insurance data information, and obtaining at least one graph entity; Analyzing the semantic relationship of each of the graph entities, and performing relationship extraction on at least one of the graph entities according to the semantic relationship to obtain a graph relationship; The insurance knowledge graph is constructed by taking the graph entities as nodes and the graph relationships as edges.

3. The insurance policy entry method according to claim 1, wherein: The converting the initial order information into text graph structure information includes: Performing text preprocessing on the initial order information, wherein the text preprocessing includes word segmentation, stop word removal, and part-of-speech tagging; Based on the initial order information after the text preprocessing operation, the initial order information is identified using a named entity recognition technology to obtain multiple text entities; Relationship identification is performed on the plurality of text entities using a preset natural language processing tool to obtain entity relationships among the plurality of text entities; The text graph structure information is constructed by using the multiple text entities as nodes of the text graph structure and the multiple entity relationships as edges.

4. The insurance policy entry method according to claim 1, wherein: The processing of the text graph structure information by using a graph neural network to generate intermediate order information includes: Identify the nodes of the text graph structure information, extract the feature vector of each node, and obtain an initial feature vector corresponding to each node; Inputting all the initial feature vectors into the graph neural network through the input layer of the graph neural network; Based on each of the initial feature vectors, collecting neighbor node information of the initial feature vector through the graph convolution layer of the graph neural network, and fusing the initial feature vector with the neighbor node information to obtain a target feature vector; The target feature vector is converted into structured data through the output layer of the graph neural network to obtain the intermediate order information.

5. The insurance policy entry method according to claim 1, wherein: After processing the text graph structure information using the graph neural network to generate intermediate order information, the method further includes: Parsing the intermediate order information to obtain the main insurance type and insured person characteristics; Based on the insurance knowledge graph, screening out a first set of supplementary insurances related to the main insurance type according to the main insurance type; Filtering the first set of supplementary insurance policies according to the insured's characteristics to obtain a second set of supplementary insurance policies; According to the amount factor of each supplementary insurance in the second supplementary insurance set, the second supplementary insurance set is sorted, and a supplementary insurance set with a preset ranking is selected from the sorted second supplementary insurance set as the target supplementary insurance set, and the target supplementary insurance set is pushed.

6. The insurance policy entry method according to claim 1, wherein: Before correcting the intermediate order information according to the insurance knowledge graph, the method further includes: Correcting the intermediate record information using a dictionary to obtain the intermediate record information after dictionary correction; The intermediate record information after the dictionary error correction is calculated using a random forest model to obtain the probability of an erroneous sentence; When the probability of the erroneous statement exceeds a preset error threshold, the erroneous statement is marked to obtain the error information.

7. The insurance policy entry method according to claim 1, wherein: The converting the intermediate order information using a sequence-to-sequence model according to the error information to generate target order information includes: Identify the error type and the position of the error information in the intermediate order information according to the type mark of the error information; Encoding the error information through the encoder of the sequence-to-sequence model to obtain an error information code; According to the position of the error information in the intermediate recording information, context information is obtained, and according to the context information and the error type, the error information is decoded by the sequence-to-sequence decoder to generate the target recording information.

8. An insurance policy entry device, characterized in that: The device comprises: A graph construction module, used to access a preset insurance database and construct an insurance knowledge graph based on the preset insurance database; The order information acquisition module is used to obtain initial order information and convert the initial order information into text graph structure information; The model processing module uses a graph neural network to process the text graph structure information and generate intermediate order information; An error information correction module is used to correct the intermediate order information according to the insurance knowledge graph to obtain error information. According to the error information, a sequence-to-sequence model is used to convert the intermediate order information to generate target order information.

9. A computer device, characterized in that: The computer device comprises: at least one processor; and, a memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the insurance policy entry method according to any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the insurance policy entry method according to any one of claims 1 to 7 is implemented.

Images