Insurance transaction method and device based on large model and artificial intelligence

Abstract

The invention discloses an insurance transaction method and device based on a large model and artificial intelligence, and the method comprises the steps: carrying out the similarity matching of consultation text data inputted by a user and a standard text vector, obtaining similar text data, sequentially carrying out the prompt word generation and question reasoning of the similar text data, obtaining corresponding answer text data, and carrying out the calculation of the corresponding answer text data. Generating a user contour according to the consultation text data; and in response to a demand transaction signal triggered by the user, demand response information is generated according to the user contour, the demand response information is fed back to the user, and insurance transaction information is generated according to the supplementary data input by the user. The specific consultation content and consultation direction of the user are determined through similar text data obtained through similarity matching of consultation text data, and the user is helped to rapidly understand insurance and rapidly select intentional insurance; the full-automatic insurance transaction process not only saves the human resource cost of insurance service personnel, but also ensures the security and confidentiality of the generated insurance transaction information.

Description

Technical Field

[0001] The present invention relates to the field of Internet insurance technology, and in particular to an insurance transaction method and device based on a large model and artificial intelligence. Background Art

[0002] With the popularization of the Internet and the continuous development of big data technology, the insurance industry is facing tremendous changes, and the insurance industry is also constantly exploring more intelligent and efficient transaction models. The insurance business process generally consists of three parts: insurance inquiry, insurance application, and maintenance. Among them, insurance inquiry is the process of users understanding and comparing insurance. Through insurance inquiry, users can select the insurance that suits them from a large number of insurance brands, types, and models; insurance application is the process in which users themselves or with the assistance of insurance agents fill in the insurance information on the insurance application page and submit it, and the insurance company reviews and pays to form an insurance policy to protect users; maintenance refers to the continuous interaction and service between users and insurance companies after the insurance contract takes effect; including policy changes, renewals, claims and other services.

[0003] Traditional insurance transaction methods have limitations in risk assessment, pricing and services, lack of pertinence and intelligence, and cannot meet the needs of modern consumers. In addition, traditional insurance consultation services require customers to screen related insurance businesses by themselves, which is not only time-consuming and labor-intensive, but also cannot guarantee that the insurance services screened are the best match for users. The same problem exists in insurance services and maintenance services. Summary of the invention

[0004] The present invention provides an insurance transaction method and device based on a big model and artificial intelligence to achieve the technical effect of improving the efficiency of insurance business processing and reducing the manpower cost and operating cost of the insurance industry.

[0005] In order to solve the above technical problems, an embodiment of the present invention provides an insurance transaction method based on a big model and artificial intelligence, comprising the following steps:

[0006] Acquire and parse the consultation text data input by the user, and perform similarity matching between the consultation text data and a preset standard text vector to obtain similar text data;

[0007] Generate prompt words and perform question reasoning on the similar text data in sequence, obtain and display answer text data corresponding to the consultation text data, and generate a corresponding user profile according to the consultation text data;

[0008] In response to a demand transaction signal triggered by a user, corresponding demand response information is generated according to the user profile, the demand response information is fed back to the user, and insurance transaction information is generated according to supplementary data input by the user.

[0009] The insurance transaction method provided by the present invention first receives consultation text data input by the user, that is, accepts question text data about insurance content input by the user, and performs similarity matching between the received consultation text data and the standard text vector preset in the database within the system to obtain similar text data, thereby determining the user's specific consultation content and consultation direction, that is, determining the user's interest direction in insurance business, so as to facilitate the subsequent generation of corresponding answer text data, and at the same time, the user profile of the user can be constructed by analyzing the consultation text data input by the user, so as to facilitate the subsequent recommendation of corresponding insurance services to the user, and also help the user to quickly understand insurance and quickly choose intended insurance.

[0010] After determining the similar text data, the system can generate prompt words and perform question reasoning on the determined similar text data to obtain the answer corresponding to the question input by the user, that is, the answer text data, and display the answer text data to the user after obtaining it so that the user can subsequently process the insurance service. At the same time, the system will also generate a corresponding user profile based on the consulting text data input by the user, that is, the user's image will be modeled based on the input consulting text data, so as to facilitate more insurance services for the user.

[0011] After determining the user profile, the system will also respond to the demand transaction signal triggered by the user, that is, the insurance service need signal, and generate corresponding demand response information based on the generated user profile and feedback it to the user, so as to facilitate the user to screen and supplement insurance services, while also saving the human resource costs of insurance service personnel. After the user supplements and confirms the demand response information fed back by the system, the system can generate insurance transaction information based on the determined demand response information. The above fully automatic insurance transaction process not only further saves the human resource costs of insurance service personnel, but also ensures the security and confidentiality of the generated insurance transaction information because the system is responsible for the entire process of insurance transactions.

[0012] As a preferred example, the similarity matching between the consultation text data and the preset standard vector text to obtain similar text data includes:

[0013] Vectorizing the consultation text data into a consultation text vector, and performing similarity matching between the consultation text vector and the standard text vector in a preset vector database to obtain a plurality of matching similarities;

[0014] The corresponding standard text vectors are sorted in descending order of the matching similarity to generate a similarity vector table, and a corresponding number of similar text data are obtained from the similarity vector table according to a preset number of matches.

[0015] Before performing similarity matching on the consulting text data input by the user, the system must first quantify the consulting text data so that the data format of the consulting text data can be similarly matched with the standard text vectors pre-stored in the system's vector database, and sort the matched standard text vectors according to the calculated matching similarity, and then select a preset number of text data as similar text data for output in the sorted order, so as to facilitate the subsequent system to perform problem reasoning on the consulting text data input by the user.

[0016] As a preferred example, the step of sequentially generating prompt words and performing question reasoning on the similar text data to obtain answer text data corresponding to the consultation text data includes:

[0017] Generate prompt words for the similar text data and the consulting text data using a prompt word template to obtain inference prompt words;

[0018] The inference prompt words are inferred through the large model, and the parameters of the large model in inference are adjusted according to the preset large model temperature, and then the inference result output by the adjusted large model is used as the answer text data.

[0019] In order to obtain the answer text data corresponding to the consultation text data, the system will first generate prompt words for the similar text data and consultation text data obtained by matching through the prompt word template, obtain the prompt words of the above two types of data, and input the prompt words into the big model. The big model will perform question reasoning on it to obtain the answer text data corresponding to the consultation text data, that is, obtain the answer to the question entered by the user.

[0020] As a preferred example, the step of generating a corresponding user profile according to the consultation text data includes:

[0021] The consultation text data is used to understand the intention, and the understood user intention is stored in a corresponding user database so that the user database is updated, and then the user profile is constructed based on the data in the updated user database.

[0022] After outputting the generated answer text data, the system will also understand the intent of the consultation text data input by the user, and store the understood user intent in the user database, so that the data in the user database is updated accordingly, and the corresponding user profile is updated thereby, further improving the integrity and practicality of the user profile.

[0023] As a preferred example, the demand transaction signal triggered in response to the user includes: an insurance transaction signal, a renewal transaction signal and a claim transaction signal.

[0024] As a preferred example, when the demand transaction signal is the insurance transaction signal, generating corresponding demand response information according to the user profile, feeding back the demand response information to the user, and generating insurance transaction information according to the supplementary data input by the user, includes:

[0025] In response to the insurance transaction signal, generating corresponding insurance information according to the user profile and feeding it back to the user, so that the user can supplement the insurance information and generate a corresponding insurance request;

[0026] Calling the insurance interface of the corresponding insurance company according to the insurance request, and sending the insurance request to the insurance interface, so that the insurance company conducts an insurance review on the user according to the insurance request, and feeds back the generated insurance status information to the user after completing the insurance review;

[0027] The insurance status information is received, and the user profile is updated according to the insurance status information.

[0028] After determining the user profile, the system will also respond to the demand transaction signal triggered by the user, that is, the insurance transaction signal, and provide insurance services to the user. When the demand transaction signal triggered by the user is an insurance transaction signal, the system will generate corresponding insurance information based on the insurance transaction signal and the user profile and feedback it to the user. The user will supplement the insurance information and generate a corresponding insurance request, thereby triggering the system to retrieve the insurance interface of the corresponding insurance company according to the insurance request, and send the insurance request to the corresponding insurance interface. The insurance company will conduct an insurance review of the user based on the received insurance request, that is, determine whether the user is eligible to apply for insurance, that is, whether it meets the underwriting requirements, and conduct a corresponding assessment of the user's insurance risk.

[0029] After completing the corresponding underwriting review, the system will generate corresponding insurance status information based on the underwriting results, that is, underwriting success or underwriting failure, and while feeding back to the user, it will also update the parameters in the system user database based on the underwriting results, thereby updating the user profile, making the updated user profile more timely and practical, and improving the authenticity and reliability of the subsequent demand response information generated based on the updated user profile.

[0030] As a preferred example, when the demand transaction signal is the renewal transaction signal, generating corresponding demand response information according to the user profile, feeding back the demand response information to the user, and generating insurance transaction information according to the supplementary data input by the user, includes:

[0031] In response to the renewal transaction signal, generating corresponding renewal information according to the user profile and feeding it back to the user, so that the user confirms the renewal information and generates a corresponding renewal request;

[0032] Calling a renewal interface of a corresponding insurance company according to the renewal request, and sending the renewal request to the renewal interface, so that the insurance company generates and feeds back a corresponding renewal transaction request according to the renewal request, so that the user adjusts the renewal status information according to the renewal transaction request;

[0033] The renewal status information fed back by the user is received, and the user profile is updated according to the renewal status information.

[0034] If the signal responded by the system is a renewal transaction signal, the system will generate renewal information based on the content related to insurance renewal in the user profile and feedback it to the user for confirmation. After the user confirms and returns the corresponding renewal request, the renewal interface of the corresponding insurance company is called according to the renewal request. The insurance company receives the renewal request through the renewal interface and generates a corresponding renewal transaction request and feeds it back to the user. The user further confirms the transaction request and then adjusts the insurance status information between the insurance company and the user through the confirmed renewal transaction request, that is, the renewal status information. After the system receives the renewal status information fed back by the user, the user database can be updated through the received renewal status information, that is, the user profile corresponding to the user is updated, thereby ensuring the timeliness of the user profile update.

[0035] As a preferred example, when the demand transaction signal is the claim transaction signal, generating corresponding demand response information according to the user profile, feeding back the demand response information to the user, and generating insurance transaction information according to the supplementary data input by the user, includes:

[0036] In response to the claim transaction signal, generating corresponding claim information according to the user profile and feeding it back to the user, so that the user can supplement the claim information and generate a corresponding claim request;

[0037] Calling a claim interface of a corresponding insurance company according to the claim request, and sending the claim request to the claim interface, so that the insurance company conducts a claim review of the user according to the claim request, and feeds back the generated claim status information to the user after completing the claim review;

[0038] The claim status information is received, and the user profile is updated according to the claim status information.

[0039] If the signal responded by the system is a claim transaction signal, the system will generate claim information based on the content related to insurance claims in the user profile and feedback it to the user for confirmation. After the user confirms and returns the corresponding claim request, the system will call the claim interface of the corresponding insurance company according to the claim request. The insurance company will receive the claim request through the claim interface and conduct a claim review on the claim request sent by the user to confirm whether the claim request applied by the user meets the claim requirements, so as to avoid abnormal claims problems that may cause losses to the insurance company.

[0040] After confirming that the claim request can be approved, the corresponding claim status information will be generated and fed back to the user, who will confirm the claim status information. At the same time, the system will also update the user database based on the claim status information, that is, update the user profile corresponding to the user, so as to ensure the timeliness of the user profile update.

[0041] As a preferred example, before obtaining and parsing the consultation text data input by the user, the method further includes:

[0042] The knowledge base file to be processed is obtained, and the knowledge base file to be processed is vectorized, the vectorization result is stored in a vector database, and the parameters of the vector database are updated according to the directory information of the knowledge base file to obtain the updated vector database.

[0043] In order to conduct a more in-depth analysis of the consulting text data input by the user and make the output answer text data more in line with the user's real needs, the system will build a vector database for analyzing the consulting text data input by the user before analyzing the consulting text data. The system will vectorize the acquired knowledge base file and store the vectorized result in the vector database. The directory information of the vector database will be updated according to the directory information of the knowledge base file, thereby realizing the update of the vector database. At the same time, the updated directory information can also be used as an index to retrieve the vector data required by the user, so that the user can adjust the data content in the vector database.

[0044] Accordingly, an embodiment of the present invention provides an insurance transaction device based on a large model and artificial intelligence, the insurance transaction device comprising a vector database construction module, a consultation text matching module, an answer text reasoning module and a transaction demand response module;

[0045] The consulting text matching module is used to obtain and analyze the consulting text data input by the user, and perform similarity matching between the consulting text data and a preset standard text vector to obtain similar text data;

[0046] The answer text reasoning module is used to perform prompt word generation and question reasoning on the similar text data in sequence, obtain answer text data corresponding to the consultation text data, and generate a corresponding user profile according to the consultation text data;

[0047] The transaction demand response module is used to respond to the demand transaction signal triggered by the user, generate corresponding demand response information according to the user profile, feed back the demand response information to the user, and generate insurance transaction information according to the supplementary data input by the user;

[0048] The vector database construction module is used to obtain the knowledge base file to be processed, and vectorize the knowledge base file to be processed, store the vectorization result in the vector database, and update the parameters of the vector database according to the directory information of the knowledge base file to obtain the updated vector database. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Figure 1 :A flow chart of an embodiment of an insurance transaction method based on a big model and artificial intelligence provided by the present invention;

[0050] Figure 2 : A structural diagram of an embodiment of an insurance transaction device based on a large model and artificial intelligence provided by the present invention;

[0051] Figure 3 : A structural diagram of an embodiment of the insurance intelligent assistant system provided by the present invention;

[0052] Figure 4 : A flowchart of insurance inquiry based on the insurance intelligent assistant system provided by the present invention;

[0053] Figure 5 : A flowchart of insurance application based on the insurance intelligent assistant system provided by the present invention;

[0054] Figure 6 : A renewal flow chart based on the insurance intelligent assistant system provided by the present invention;

[0055] Figure 7 : A claims settlement flow chart based on the insurance intelligent assistant system provided by the present invention;

[0056] Figure 8 : Add a flow chart to the industry knowledge base based on the insurance intelligent assistant system provided by the present invention;

[0057] Fig. 9 : A flowchart for modifying the industry knowledge base based on the insurance intelligent assistant system provided by the present invention;

[0058] Fig.10: A flowchart of deleting the industry knowledge base based on the insurance intelligent assistant system provided by the present invention;

[0059] Fig.11 : A flowchart of an industry knowledge base query based on the insurance intelligent assistant system provided by the present invention;

[0060] Fig.12 : A schematic diagram of the system deployment of the insurance intelligent assistant system provided by the present invention. DETAILED DESCRIPTION

[0061] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0062] Embodiment 1

[0063] Please refer to Figure 1 , which is a flow chart of an embodiment of the insurance transaction method based on a big model and artificial intelligence provided by the present invention, including steps 101 to 103, each of which is specifically as follows:

[0064] Step 101: Acquire and parse consultation text data input by a user, and perform similarity matching between the consultation text data and a preset standard text vector to obtain similar text data.

[0065] The insurance transaction method provided by the embodiment of the present invention first receives consultation text data input by the user, that is, accepts question text data about insurance content input by the user, and performs similarity matching on the received consultation text data with the standard text vector preset in the database within the system to obtain similar text data, thereby determining the user's specific consultation content and consultation direction, that is, determining the user's interest direction in insurance business, so as to facilitate the subsequent generation of corresponding answer text data, and at the same time, the user profile of the user can be constructed by analyzing the consultation text data input by the user, so as to facilitate the subsequent recommendation of corresponding insurance services to the user, and also help the user to quickly understand insurance and quickly choose intended insurance.

[0066] Specifically, the method of performing similarity matching between the consultation text data and the preset standard vector text to obtain similar text data in this embodiment includes:

[0067] Vectorizing the consultation text data into a consultation text vector, and performing similarity matching between the consultation text vector and the standard text vector in a preset vector database to obtain a plurality of matching similarities;

[0068] The corresponding standard text vectors are sorted in descending order of the matching similarity to generate a similarity vector table, and a corresponding number of similar text data are obtained from the similarity vector table according to a preset number of matches.

[0069] Before performing similarity matching on the consulting text data input by the user, the system must first quantify the consulting text data so that the data format of the consulting text data can be similarly matched with the standard text vectors pre-stored in the system's vector database, and sort the matched standard text vectors according to the calculated matching similarity, and then select a preset number of text data as similar text data for output in the sorted order, so as to facilitate the subsequent system to perform problem reasoning on the consulting text data input by the user.

[0070] In this embodiment, before obtaining and parsing the consultation text data input by the user, the method further includes:

[0071] The knowledge base file to be processed is obtained, and the knowledge base file to be processed is vectorized, the vectorization result is stored in a vector database, and the parameters of the vector database are updated according to the directory information of the knowledge base file to obtain the updated vector database.

[0072] In order to conduct a more in-depth analysis of the consulting text data input by the user and make the output answer text data more in line with the user's real needs, the system will build a vector database for analyzing the consulting text data input by the user before analyzing the consulting text data. The system will vectorize the acquired knowledge base file and store the vectorized result in the vector database. The directory information of the vector database will be updated according to the directory information of the knowledge base file, thereby realizing the update of the vector database. At the same time, the updated directory information can also be used as an index to retrieve the vector data required by the user, so that the user can adjust the data content in the vector database.

[0073] Step 102: Prompt word generation and question reasoning are performed on the similar text data in sequence, answer text data corresponding to the consultation text data is obtained and displayed, and a corresponding user profile is generated according to the consultation text data.

[0074] After determining the similar text data, the system can generate prompt words and perform question reasoning on the determined similar text data to obtain the answer corresponding to the question input by the user, that is, the answer text data, and display the answer text data to the user after obtaining it so that the user can subsequently process the insurance service. At the same time, the system will also generate a corresponding user profile based on the consulting text data input by the user, that is, the user's image will be modeled based on the input consulting text data, so as to facilitate more insurance services for the user.

[0075] Specifically, the present embodiment sequentially generates prompt words and performs question reasoning on the similar text data to obtain answer text data corresponding to the consultation text data, including:

[0076] Generate prompt words for the similar text data and the consulting text data using a prompt word template to obtain inference prompt words;

[0077] The inference prompt words are inferred through the large model, and the parameters of the large model in inference are adjusted according to the preset large model temperature, and then the inference result output by the adjusted large model is used as the answer text data.

[0078] In order to obtain the answer text data corresponding to the consultation text data, the system will first generate prompt words for the similar text data and consultation text data obtained by matching through the prompt word template, obtain the prompt words of the above two types of data, and input the prompt words into the big model. The big model will perform question reasoning on it to obtain the answer text data corresponding to the consultation text data, that is, obtain the answer to the question entered by the user.

[0079] Furthermore, the step of generating a corresponding user profile according to the consultation text data in this embodiment includes:

[0080] The consultation text data is used to understand the intention, and the understood user intention is stored in a corresponding user database so that the user database is updated, and then the user profile is constructed based on the data in the updated user database.

[0081] After outputting the generated answer text data, the system will also understand the intent of the consultation text data input by the user, and store the understood user intent in the user database, so that the data in the user database is updated accordingly, and the corresponding user profile is updated thereby, further improving the integrity and practicality of the user profile.

[0082] Step 103: In response to the demand transaction signal triggered by the user, corresponding demand response information is generated according to the user profile, the demand response information is fed back to the user, and insurance transaction information is generated according to the supplementary data input by the user.

[0083] After determining the user profile, the system will also respond to the demand transaction signal triggered by the user, that is, the insurance service need signal, and generate corresponding demand response information based on the generated user profile and feedback it to the user, so as to facilitate the user to screen and supplement insurance services, while also saving the human resource costs of insurance service personnel. After the user supplements and confirms the demand response information fed back by the system, the system can generate insurance transaction information based on the determined demand response information. The above fully automatic insurance transaction process not only further saves the human resource costs of insurance service personnel, but also ensures the security and confidentiality of the generated insurance transaction information because the system is responsible for the entire process of insurance transactions.

[0084] Among them, the demand transaction signal in response to the user trigger in this embodiment includes: an insurance transaction signal, a renewal transaction signal and a claim transaction signal.

[0085] Specifically, when the demand transaction signal in this embodiment is the insurance transaction signal, generating corresponding demand response information according to the user profile, feeding back the demand response information to the user, and generating insurance transaction information according to the supplementary data input by the user include:

[0086] In response to the insurance transaction signal, generating corresponding insurance information according to the user profile and feeding it back to the user, so that the user can supplement the insurance information and generate a corresponding insurance request;

[0087] Calling the insurance interface of the corresponding insurance company according to the insurance request, and sending the insurance request to the insurance interface, so that the insurance company conducts an insurance review on the user according to the insurance request, and feeds back the generated insurance status information to the user after completing the insurance review;

[0088] The insurance status information is received, and the user profile is updated according to the insurance status information.

[0089] After determining the user profile, the system will also respond to the demand transaction signal triggered by the user, that is, the insurance transaction signal, and provide insurance services to the user. When the demand transaction signal triggered by the user is an insurance transaction signal, the system will generate corresponding insurance information based on the insurance transaction signal and the user profile and feedback it to the user. The user will supplement the insurance information and generate a corresponding insurance request, thereby triggering the system to retrieve the insurance interface of the corresponding insurance company according to the insurance request, and send the insurance request to the corresponding insurance interface. The insurance company will conduct an insurance review of the user based on the received insurance request, that is, determine whether the user is eligible to apply for insurance, that is, whether it meets the underwriting requirements, and conduct a corresponding assessment of the user's insurance risk.

[0090] After completing the corresponding underwriting review, the system will generate corresponding insurance status information based on the underwriting results, that is, underwriting success or underwriting failure, and while feeding back to the user, it will also update the parameters in the system user database based on the underwriting results, thereby updating the user profile, making the updated user profile more timely and practical, and improving the authenticity and reliability of the subsequent demand response information generated based on the updated user profile.

[0091] Further, when the demand transaction signal in this embodiment is the renewal transaction signal, generating corresponding demand response information according to the user profile, feeding back the demand response information to the user, and generating insurance transaction information according to the supplementary data input by the user include:

[0092] In response to the renewal transaction signal, generating corresponding renewal information according to the user profile and feeding it back to the user, so that the user confirms the renewal information and generates a corresponding renewal request;

[0093] Calling a renewal interface of a corresponding insurance company according to the renewal request, and sending the renewal request to the renewal interface, so that the insurance company generates and feeds back a corresponding renewal transaction request according to the renewal request, so that the user adjusts the renewal status information according to the renewal transaction request;

[0094] The renewal status information fed back by the user is received, and the user profile is updated according to the renewal status information.

[0095] If the signal responded by the system is a renewal transaction signal, the system will generate renewal information based on the content related to insurance renewal in the user profile and feedback it to the user for confirmation. After the user confirms and returns the corresponding renewal request, the renewal interface of the corresponding insurance company is called according to the renewal request. The insurance company receives the renewal request through the renewal interface and generates a corresponding renewal transaction request and feeds it back to the user. The user further confirms the transaction request and then adjusts the insurance status information between the insurance company and the user through the confirmed renewal transaction request, that is, the renewal status information. After the system receives the renewal status information fed back by the user, the user database can be updated through the received renewal status information, that is, the user profile corresponding to the user is updated, thereby ensuring the timeliness of the user profile update.

[0096] Furthermore, when the demand transaction signal in this embodiment is the claim transaction signal, generating corresponding demand response information according to the user profile, feeding back the demand response information to the user, and generating insurance transaction information according to the supplementary data input by the user include:

[0097] In response to the claim transaction signal, generating corresponding claim information according to the user profile and feeding it back to the user, so that the user can supplement the claim information and generate a corresponding claim request;

[0098] Calling a claim interface of a corresponding insurance company according to the claim request, and sending the claim request to the claim interface, so that the insurance company conducts a claim review of the user according to the claim request, and feeds back the generated claim status information to the user after completing the claim review;

[0099] The claim status information is received, and the user profile is updated according to the claim status information.

[0100] If the signal responded by the system is a claim transaction signal, the system will generate claim information based on the content related to insurance claims in the user profile and feedback it to the user for confirmation. After the user confirms and returns the corresponding claim request, the system will call the claim interface of the corresponding insurance company according to the claim request. The insurance company will receive the claim request through the claim interface and conduct a claim review on the claim request sent by the user to confirm whether the claim request applied by the user meets the claim requirements, so as to avoid abnormal claims problems that may cause losses to the insurance company.

[0101] After confirming that the claim request can be approved, the corresponding claim status information will be generated and fed back to the user, who will confirm the claim status information. At the same time, the system will also update the user database based on the claim status information, that is, update the user profile corresponding to the user, so as to ensure the timeliness of the user profile update.

[0102] See also Figure 2 , Figure 2 This is a structural diagram of an embodiment of an insurance transaction device based on a large model and artificial intelligence provided by the present invention. Figure 2 As shown, accordingly, an embodiment of the present invention also provides an insurance transaction device based on a big model and artificial intelligence, and the insurance transaction device includes a vector database construction module 201, a consultation text matching module 202, an answer text reasoning module 203 and a transaction demand response module 204.

[0103] Among them, the vector database construction module 201 is used to obtain the knowledge base file to be processed, and vectorize the knowledge base file to be processed, store the vectorization result in the vector database, and update the parameters of the vector database according to the directory information of the knowledge base file to obtain the updated vector database.

[0104] The consulting text matching module 202 is used to obtain and analyze the consulting text data input by the user, and perform similarity matching between the consulting text data and a preset standard text vector to obtain similar text data.

[0105] Furthermore, the consulting text matching module 202 performs similarity matching on the consulting text data and a preset standard vector text to obtain similar text data, including:

[0106] The consulting text data is vectorized into a consulting text vector, and the consulting text vector is matched with the standard text vector in a preset vector database for similarity to obtain a number of matching similarities; the corresponding standard text vectors are sorted in order from high to low according to the matching similarities to generate a similarity vector table, and a corresponding number of similar text data are obtained from the similarity vector table according to a preset number of matches.

[0107] The answer text reasoning module 203 is used to perform prompt word generation and question reasoning on the similar text data in sequence, obtain answer text data corresponding to the consultation text data, and generate a corresponding user profile based on the consultation text data.

[0108] Furthermore, the answer text reasoning module 203 sequentially generates prompt words and performs question reasoning on the similar text data to obtain answer text data corresponding to the consultation text data, including:

[0109] Prompt words are generated for the similar text data and the consultation text data through a prompt word template to obtain inference prompt words; the inference prompt words are inferred through a large model, and the parameters of the large model in reasoning are adjusted according to a preset large model temperature, and then the inference result output by the adjusted large model is used as the answer text data.

[0110] Furthermore, the answer text reasoning module 203 generates a corresponding user profile according to the consultation text data, including:

[0111] The consultation text data is used to understand the intention, and the understood user intention is stored in a corresponding user database so that the user database is updated, and then the user profile is constructed based on the data in the updated user database.

[0112] The transaction demand response module 204 is used to respond to the demand transaction signal triggered by the user, generate corresponding demand response information according to the user profile, feed back the demand response information to the user, and generate insurance transaction information according to the supplementary data input by the user. The transaction demand response module 204 responds to the demand transaction signal triggered by the user, including: insurance transaction signal, renewal transaction signal and claim transaction signal.

[0113] Furthermore, when the demand transaction signal is the insurance transaction signal, the transaction demand response module 204 generates corresponding demand response information according to the user profile, feeds back the demand response information to the user, and generates insurance transaction information according to the supplementary data input by the user, including:

[0114] In response to the insurance transaction signal, generating corresponding insurance information according to the user profile and feeding it back to the user, so that the user can supplement the insurance information and generate a corresponding insurance request;

[0115] The insurance interface of the corresponding insurance company is called according to the insurance request, and the insurance request is sent to the insurance interface, so that the insurance company conducts an insurance underwriting review on the user according to the insurance request, and feeds back the generated insurance status information to the user after completing the insurance underwriting review; the insurance status information is received, and the user profile is updated according to the insurance status information.

[0116] Furthermore, when the demand transaction signal is the renewal transaction signal, the transaction demand response module 204 generates corresponding demand response information according to the user profile, feeds back the demand response information to the user, and generates insurance transaction information according to the supplementary data input by the user, including:

[0117] In response to the renewal transaction signal, generating corresponding renewal information according to the user profile and feeding it back to the user, so that the user confirms the renewal information and generates a corresponding renewal request;

[0118] The renewal interface of the corresponding insurance company is called according to the renewal request, and the renewal request is sent to the renewal interface, so that the insurance company generates and feeds back a corresponding renewal transaction request according to the renewal request, so that the user adjusts the renewal status information according to the renewal transaction request; the renewal status information fed back by the user is received, and the user profile is updated according to the renewal status information.

[0119] Furthermore, when the demand transaction signal is the claim transaction signal, the transaction demand response module 204 generates corresponding demand response information according to the user profile, feeds back the demand response information to the user, and generates insurance transaction information according to the supplementary data input by the user, including:

[0120] In response to the claim transaction signal, generating corresponding claim information according to the user profile and feeding it back to the user, so that the user can supplement the claim information and generate a corresponding claim request;

[0121] The claim interface of the corresponding insurance company is called according to the claim request, and the claim request is sent to the claim interface, so that the insurance company conducts a claim review on the user according to the claim request, and feeds back the generated claim status information to the user after completing the claim review; the claim status information is received, and the user profile is updated according to the claim status information.

[0122] Embodiment 2

[0123] See also Figure 3 , Figure 3 This is a structural diagram of an embodiment of the insurance intelligent assistant system provided by the present invention. Figure 3 As shown, the insurance intelligent assistant system, i.e., the insurance transaction device described in Example 1, is composed of the industry knowledge base / prompt word / large model management subsystem, i.e., the vector database construction module 201 in Example 1, the insurance inquiry subsystem, i.e., the consultation text matching module 202 and the answer text reasoning module 203 in Example 1, the insurance subsystem, the maintenance subsystem, and the user subsystem. Among them, the transaction demand response module described in Example 1 includes the insurance subsystem, the maintenance subsystem, and the user subsystem described in this embodiment.

[0124] like Figure 3 As shown, the industry knowledge base / prompt word / large model management subsystem is composed of modules such as industry knowledge base management M11, prompt word management M12, large model management M13, industry knowledge base M14, prompt word library M15 and large model library M16.

[0125] in, Figure 3 The industry knowledge base management module M11 is responsible for the management of the industry knowledge base, including functions such as adding an industry knowledge base, modifying an industry knowledge base, deleting an industry knowledge base, and querying an industry knowledge base; when adding a new industry knowledge base, Word2Vec will be used to vectorize the text to ensure that semantically similar words are close to each other in the vector space, and the vectorization results will be written into the vector database, which can be either Milvus or Faiss.

[0126] The industry knowledge base of the large model refers to the industry data set that the artificial intelligence model relies on, and in this embodiment, the industry knowledge base refers to the industry text data used for large model reasoning. The Word2Vec is an algorithm for generating word embeddings, which can convert words into vectors so that semantically similar words are close to each other in the vector space.

[0127] Figure 3 The prompt word management module M12 is shown: used for managing prompt word templates, including functions such as adding, modifying, deleting and querying prompt word templates.

[0128] Among them, the prompt words of the large model refer to specific keywords or phrases that users can use to activate or customize certain functions or behaviors of the assistant when communicating with the artificial intelligence assistant. For example, in some artificial intelligence systems, users can switch languages, change the conversation style, and request specific types of information or services by entering specific commands or prompt words.

[0129] The prompt word template of the large model is a text sequence designed to guide the model to respond in a specific way. The prompt word template described in this embodiment can be customized for different tasks and goals.

[0130] Figure 3 The large model management module M13 shown is responsible for the management of large models, including functions such as adding, modifying, deleting and querying large models. Among them, "large model" refers to an artificial intelligence model with a large number of parameters and complex structure, which can process and understand a large amount of data and perform complex tasks such as language understanding, image recognition and data analysis. For example, in the field of natural language processing, "large model" can be used to refer to models such as BERT (Bidirectional Encoder Representations from Transformers) or GPT (Generated Pre-trai ned Transformer), which have billions or even hundreds of billions of parameters and can generate coherent text, translate language, answer questions, and even create stories.

[0131] Assume that the large model is GPT-3, a large-scale language model developed by OpenAI that contains 175 billion parameters and can generate high-quality text and understand complex instructions. Users can ask questions or give instructions to GPT-3, and the corresponding model can generate corresponding answers or text to display to users.

[0132] Figure 3 The industry knowledge base M14 shown is a library managed by M11. It is used to store the industry knowledge base directory in the database and store the knowledge base text files on the disk. The stored file formats include txt, word, pdf, excel, md and other formats.

[0133] Figure 3 The prompt word library M15 shown is a prompt word template library managed by M12. Since the prompt word is generally no longer than 10KB, it can be directly stored in the database.

[0134] Figure 3The large model library M16 shown is a large model managed by M13. It stores the large model directory in the database and stores the large model files in the disk. The larger the model size, the larger the model file will generally be.

[0135] The insurance consultation subsystem provides users with insurance consultation functions, and is composed of modules such as question and answer M21, similarity matching M22, forming prompt words M23, large model reasoning M24, counter-question M25 and intention understanding M26.

[0136] in, Figure 3 The question-and-answer module M21 shown is used to receive questions from users and send the questions to the similarity matching module M22 for similarity matching. After large model reasoning, the user's answers are obtained and answered to the user.

[0137] Figure 3 The similarity matching module M22 shown is used to vectorize the user's question and match it with all knowledge base vectors in the industry knowledge base based on the vector similarity matching algorithm, and then obtain a corresponding number of knowledge base vectors according to the similarity from high to low according to the pre-set maximum matching number. Among them, the vector similarity matching algorithm is a method for determining the similarity of two or more vectors in a multidimensional space, specifically including: cosine similarity algorithm, Euclidean distance algorithm and dot product algorithm.

[0138] Figure 3 The prompt word formation module M23 is shown: it is used to form the prompt words for large model reasoning together with the vector text matched by M22 from the knowledge base, along with the question text, based on the prompt word template.

[0139] Figure 3 The large model reasoning module M24 shown is used to preload the large model into the GPU or CPU, and perform question reasoning based on the question of the prompt word, and output the answer to the question based on the preset large model temperature, and then send the answer to the question to the question and answer module M21. Among them, "large model temperature" refers to the factor that controls the randomness in the large model generation process. In the large model, lower temperature values ​​will cause the model to generate more deterministic and predictable results; while higher temperature values ​​will increase the diversity and randomness of the results. For example: in text generation, a lower temperature may cause the model to generate more grammatical and logical text, while a higher temperature may produce more diverse, and sometimes even creative or unconventional sentences.

[0140] Figure 3 The reverse question module M25 shown is used to obtain the user's true intention. When the system does not obtain relevant information, the module will form reverse questions to the user according to a specific template and ask the user reverse questions.

[0141] Figure 3The schematic understanding module M26 is used to obtain the user's true intention during the question-and-answer and counter-question process with the user, such as the type of insurance the user intends to purchase, the intended insurance product, the budget amount range, and the purchase date.

[0142] The insurance subsystem provides users with a one-click insurance function, which consists of modules such as one-click insurance M31 and insurance call distribution M32.

[0143] Figure 3 The one-click insurance module M31 shown is used to receive the user's insurance requirements, intentions and user information, and automatically generate an insurance call API request, and send it to the insurance call distribution M32; the above process will eliminate the process of users filling in insurance information line by line on the insurance pages of different insurance companies, greatly improving insurance efficiency.

[0144] Figure 3 The insurance call distribution M32 shown is used to route and distribute the API request to the information system of the corresponding insurance company to complete the insurance of the corresponding insurance.

[0145] The maintenance subsystem provides users with a one-click maintenance function, which consists of modules such as one-click renewal M41, intelligent claims settlement M42 and maintenance call distribution M43.

[0146] Figure 3 The one-click renewal module M41 shown is used to enable the user to renew the insurance with one click by clicking on the module according to the reminder link of the maintenance interface M55 in the user subsystem. If the insurance requires payment, the renewal can be completed by completing the payment.

[0147] Figure 3 The intelligent claims module M42 shown is used for when a user has an insurance claim. The system can automatically generate a claims API request based on the user's matter description and related requirements, and send it to the maintenance call distribution module M43; at the same time, this module can also follow up on the claims progress by querying the claims API interface and remind the user of the corresponding progress.

[0148] Figure 3 The maintenance call distribution module M43 shown is used to route and distribute API requests to the information system of the corresponding insurance company to complete the maintenance process of the corresponding insurance, including renewal, claims, etc.

[0149] The user subsystem provides an access interface for users and stores user-related data in the user database. It consists of modules such as user registration / login M51, user settings M52, insurance inquiry interface M53, insurance purchase interface M54, maintenance interface M55 and user database M56.

[0150] Figure 3The user registration / login module M51 shown is used for user registration or login. The user registration / login module M51 described in this embodiment supports multiple interaction modes such as h5, app, web, etc.

[0151] Figure 3 The user setting module M52 shown is used for the user to set relevant information, including user personal information (name, ID number, etc.), insurance information (information of the policyholder, the insured, the beneficiary, etc.), etc.

[0152] Figure 3 The insurance inquiry interface M53 shown is used to provide users with an interface for insurance inquiry interaction. Users can ask insurance-related questions on the interface, and the insurance inquiry subsystem will receive the questions and respond to them. At the same time, in order to understand the user's true intentions, the insurance inquiry subsystem may form counter-questions and ask users to reply to the questions. In addition, the insurance inquiry interface M53 also supports multiple interaction methods such as h5, app, web, etc., and the interactive content supports text, voice and pictures.

[0153] Figure 3 The insurance interface M54 shown is an interface for providing users with insurance interaction. Users can purchase insurance with one click on the interface. The insurance subsystem can form an insurance link according to the user's intention. When the user clicks the link, some insurance information will be automatically generated. After the missing information is supplemented by the user, the user can confirm the insurance with one click. At the same time, the insurance interface M54 also supports multiple interaction methods such as h5, app, web, etc., and the interactive content supports text, voice and pictures.

[0154] Figure 3 The maintenance interface M55 shown is an interface for providing users with maintenance interaction. The maintenance subsystem will automatically form a maintenance reminder link within the renewal period based on the insurance that the user has purchased, so that the user can renew the insurance with one click. At the same time, when the user has an insurance claim, an interactive interface for insurance claim application is provided, so that the user can describe the claim and make relevant requirements, and send them to the maintenance subsystem for intelligent claim processing. In addition, when the claim progress is updated, the system can also remind the user of the corresponding progress through this interface. The maintenance interface M55 also supports multiple interaction methods such as h5, app, web, etc., and the interactive content supports text, voice and pictures.

[0155] Figure 3 The user database M56 shown is used to store user registration and login information (including user name, nickname, etc.) and user-set information (including user personal information and insurance information). At the same time, it stores contextual data such as text, voice, and pictures that the user has interacted with the system during the process of insurance inquiry, insurance purchase, and maintenance, to form a personalized user profile, and establish a data foundation for the user's personalized "insurance smart assistant" service.

[0156] In this embodiment, see Figure 4 , Figure 4 This is a flowchart of insurance inquiry based on the insurance intelligent assistant system provided by the present invention. Figure 4 As shown, the insurance inquiry process provided in this embodiment specifically includes 9 steps, namely, S401 user logs in to the system, S402 user enters the insurance inquiry interface, S403 user inquires about insurance-related issues, S404 the system performs similarity matching on the issues in the industry knowledge base, S405 forms prompt words based on prompt word templates, S406 infers answers based on the large model, S407 understands the intention, S408 forms counter-questions, and S409 sends messages to the user. The specific implementation of each step is as follows:

[0157] S401 User logs into the system: The user logs into the system through the user registration / login module M51 of the user subsystem.

[0158] S402 User enters the insurance inquiry interface: User enters the insurance inquiry interface M53.

[0159] S403 User inquires about insurance-related issues: The user inquires about insurance-related issues on the interface and sends the questions to the question-answering module M21 of the insurance inquiry subsystem.

[0160] S404 The system performs similarity matching on the question in the industry knowledge base: the question-answering module M21 receives the user question, sends it to the similarity matching module M22, performs similarity matching, and obtains the top N highly similar vector texts, that is, the N vectors with higher similarity are obtained from high to low according to the similarity ranking (N can be set by the system, and the value range can be any positive integer greater than 0. The larger the value of N, the greater the performance loss, which will result in a long time for answering the question; therefore, in order to ensure low latency in the response, this embodiment preferably takes the value N=3).

[0161] S405 forms prompt words based on the prompt word template: the user's question and the N vector texts matched are added to the prompt word template by the prompt word forming module M23 to form prompt words.

[0162] S406 Reasoning the answer based on the large model: the prompt word module M23 sends the prompt word to the large model reasoning module M24 to perform question reasoning and obtain the answer to the question.

[0163] S407 Intent understanding: According to whether the system setting requires intent understanding, the intent understanding module M26 understands the intent of the question and answer; at the same time, the intent can also be written into the user database to form a more complete user profile.

[0164] S408 forms rhetorical questions: according to the system setting whether it is necessary to ask the user questions to obtain more user needs, the rhetorical question module M25 can form rhetorical questions; at the same time, a method of forming rhetorical questions provided in this embodiment is based on large model reasoning, that is, the questions that need to be understood are formed into rhetorical question sentences by using prompt word templates and large model reasoning.

[0165] S409 sends a message to the user: the question-answering module M21 replies the answer to the question to the question-answering interface M53; if it is a counter-question, the counter-questioning module M25 replies the counter-question to the question-answering interface.

[0166] See also Figure 5 , Figure 5 The insurance process flow chart based on the insurance intelligent assistant system provided by the present invention is as follows: Figure 5 As shown, the insurance process provided in this embodiment includes 10 steps: S501, the system forms an insurance link on the insurance interface M54 according to the user's intention, S502 the user clicks the link, S503 the system automatically generates insurance information, S504 the user supplements the missing information, S505 the user confirms the insurance, S506 the system calls the insurance interface of the corresponding insurance company, S507 the insurance company underwrites, S508 the insurance company forms an insurance policy, S509 the user pays the premium, and S510 the system sends a message to the user. The specific implementation method of each step is as follows:

[0167] S501 Based on the user's intention, the system forms an insurance link on the insurance interface M54: Based on the user's intention, the insurance subsystem forms an insurance link on the insurance interface M54 for the user to click on to purchase insurance.

[0168] S502 User clicks link: The user clicks the insurance link to purchase insurance.

[0169] S503 The system automatically generates insurance information: The insurance subsystem automatically generates insurance information based on the user intention obtained by the insurance inquiry subsystem and the user profile in the user database, including the policyholder, the insured, the beneficiary, the insurance product, the premium, and other information, and presents it to the user on the insurance interface M54; optionally, the insurance information that is not automatically generated will be highlighted in red for the user to supplement.

[0170] S504 User supplements missing information: The user supplements the missing insurance information in the insurance interface M54 to form an insurance application.

[0171] S505 User confirms insurance: The user clicks the confirmation button to purchase insurance.

[0172] S506 The system calls the insurance interface of the corresponding insurance company: the insurance subsystem receives the user's insurance confirmation request, and the insurance call distribution module M32 calls the insurance interface of the corresponding insurance company according to the insurance product.

[0173] S507 Insurance company underwriting: The insurance company assesses the customer's risk and decides whether to underwrite the insurance based on the insurance application. Risks include reviewing personal information, health status, financial status, historical records, etc. to ensure that the risks are controllable and in compliance with regulations.

[0174] S508 The insurance company forms a policy: If the underwriting is passed, the insurance company system will form a policy for the user to pay.

[0175] S509 User pays premium: On the insurance application interface, the user pays the premium for the policy that has passed the underwriting and is awaiting payment.

[0176] S510 system sends information to the user: when underwriting fails or succeeds or payment is successful, the insurance subsystem will send insurance status information to the user and present it on the insurance interface M54; at the same time, the insurance subsystem will also write the insurance / underwriting information into the user database to complete the user profile.

[0177] See also Figure 6 , Figure 6 The renewal flow chart based on the insurance intelligent assistant system provided by the present invention is as follows: Figure 6 As shown, the renewal process provided in this embodiment includes 7 steps: S601 the system forms a renewal link on the maintenance interface M55, S602 the user clicks the link, S603 the system automatically generates renewal information, S604 the user confirms the renewal, S605 the system calls the renewal interface of the corresponding insurance company, S606 the user pays the premium, and S607 the system sends a message to the user. The specific implementation method of each step is as follows:

[0178] The S601 system forms a renewal link on the maintenance interface M55: The maintenance subsystem reminds the user to renew the insurance within the renewal period on the maintenance interface M55 based on the insurance purchased by the user, and forms a renewal link for the user to click to renew the insurance.

[0179] S602 User clicks link: The user clicks the renewal link to renew the policy.

[0180] S603 The system automatically generates renewal information: The renewal subsystem automatically generates renewal information based on user information and insurance information in the user database, including policy number, insurance type, policyholder, insured, beneficiary, insurance product, premium, and other information, and presents it to the user on the maintenance interface M55.

[0181] S604 User confirms renewal: The user clicks the confirmation button to renew.

[0182] S605 The system calls the renewal interface of the corresponding insurance company: The maintenance subsystem receives the user's renewal confirmation request, and the maintenance call distribution module M43 calls the renewal interface of the corresponding insurance company according to the insurance product.

[0183] S606 User pays premium: The user pays the premium for the renewal policy to be paid on the maintenance interface.

[0184] S607 The system sends information to the user: When the payment is successful, the maintenance subsystem will send the renewal status information to the user and present it on the maintenance interface M55; at the same time, the maintenance subsystem will also write the renewal information into the user database to complete the user profile.

[0185] See also Figure 7 , Figure 7 The claims process flow chart based on the insurance intelligent assistant system provided by the present invention is as follows: Figure 7 As shown, the claim process provided in this embodiment includes: S701 the user describes the insurance claim matters and related requirements on the maintenance interface M55, S702 the system forms a claim link on the maintenance interface M55, S703 the user clicks the link, S704 the system automatically generates an insurance claim application form, S705 the user supplements the missing information, S706 the user confirms the claim, S707 the system calls the claim interface of the corresponding insurance company, S708 the insurance company reviews the claim, S709 the insurance company makes the claim, the user receives the claim amount, S710 the system sends a message to the user, etc. 10 steps, and the specific implementation method of each step is as follows:

[0186] S701 The user describes the insurance claim matters and related requirements on the maintenance interface M55: When the insurance purchased by the user meets the claim requirements, the user can initiate a claim through this system and describe the claim matters and related requirements corresponding to the insurance on the maintenance interface M55.

[0187] S702 The system forms a claim link on the maintenance interface M55: Based on the insurance purchased by the user, the claim items entered by the user and related requirements, the system forms a claim link on the maintenance interface M55 for the user to click to make a claim.

[0188] S703 User clicks link: The user clicks the claim link to make a claim.

[0189] S704 The system automatically generates an insurance claim application form: The maintenance subsystem automatically generates a claim application form based on the insurance purchased by the user, the claim items and related requirements entered, and the user profile of the user database. The form includes claim information such as the insured person information, insurance contract details, description of the claim reason, loss estimate, relevant supporting document attachments, claim amount, etc., and presents it to the user on the maintenance interface M55; optionally, claim information that is not automatically generated will be highlighted in red for the user to supplement.

[0190] S705 User supplements missing information: The user supplements the missing claim information in the maintenance interface M55 to form a complete claim application form.

[0191] S706 User confirms claim: The user clicks the confirmation button to make a claim.

[0192] S707 The system calls the claim interface of the corresponding insurance company: The maintenance subsystem receives the user's claim confirmation request, and the maintenance call distribution module M43 calls the claim interface of the corresponding insurance company according to the insurance product.

[0193] S708 Insurance company review and claim settlement (referred to as claim settlement): The insurance company will review relevant documents, verify accidents, assess losses, conduct risk analysis, fraud detection, and conduct internal approval based on the claim application form and the insurance policy purchased by the user, and then decide whether to pay the claim and the amount of the claim.

[0194] S709 The insurance company makes a claim and the user receives the claim payment: If the claim is approved, the insurance company makes a claim and the user receives the claim payment.

[0195] The S710 system sends information to the user: when the claim verification fails or succeeds or the claim payment is received, the maintenance subsystem will send the claim status information to the user and present it on the maintenance interface M55; at the same time, the maintenance subsystem will also write the claim / claim verification information into the user database to complete the user profile.

[0196] See also Figure 8 , Figure 8 A flow chart is added to the industry knowledge base based on the insurance intelligent assistant system provided by the present invention, such as Figure 8 As shown, the process of adding an industry knowledge base provided in this embodiment includes four steps, namely, S801 naming and describing, S802 importing and uploading files, S803 vectorizing data and writing it into a vector database, and S804 writing the directory into a database table. The specific implementation of each step is as follows:

[0197] S801 Naming and Description: Give the new industry knowledge base an understandable name and add a description of the knowledge base;

[0198] S802 Import and upload files: Import knowledge base files. The files will be uploaded to the industry knowledge base server and stored locally on the server.

[0199] S803 Data vectorization and writing into vector database: The text is vectorized by Word2Vec algorithm, and the vectorization result is written into the vector database, which can be Milvus or Faiss. When a single file is vectorized, it can be vectorized into multiple vectors according to the block size; for example, if the size of a file is 10KB (1KB=1024 bytes), and the block size is set to 700B (i.e. 700 bytes), the vectorization result is 15 vectors.

[0200] S804 writes the directory into the database table: writes the industry knowledge base directory information into the database, the directory information includes: knowledge base name, description, knowledge base file name, file type, file length, industry knowledge base server IP address, file path, vector quantity and other field contents.

[0201] See also Fig. 9 , Fig. 9 The industry knowledge base modification flow chart based on the insurance intelligent assistant system provided by the present invention is as follows: Fig. 9 As shown, the industry knowledge base modification process provided in this embodiment includes four steps: S901 selects the industry knowledge base to be modified, S902 modifies the name or description, S903 modifies the file and vectorizes it, and S904 refreshes the directory of the database table. The specific implementation method of each step is as follows:

[0202] S901 Select the industry knowledge base to be modified: select a knowledge base to be modified from multiple industry knowledge bases;

[0203] S902 Modify name or description: This step is optional. If you choose to modify, the name or description of the knowledge base will be modified.

[0204] S903 Modify the file and vectorize: This step can be omitted. If you choose to modify the file, the original file will be deleted and modified to the latest file. At the same time, the file will be vectorized and the original vector in the original vector database will be updated.

[0205] S904 Refresh the directory of the database table: refresh the industry knowledge base directory information of the database, the directory information includes: updated knowledge base name, updated description, updated knowledge base file name, updated file type, updated file length, industry knowledge base server IP address, file path, updated vector quantity and other field contents.

[0206] See also Fig.10 , Fig.10 The industry knowledge base deletion flow chart based on the insurance intelligent assistant system provided by the present invention is as follows: Fig.10 As shown, the industry knowledge base deletion process provided in this embodiment includes three steps: S1001 selecting the industry knowledge base to be deleted, S1002 deleting files and vectorization results, and S1003 deleting the directory from the database table. The implementation method of each step is as follows:

[0207] S1001 Select the industry knowledge base to be deleted: select a knowledge base to be deleted from multiple industry knowledge bases;

[0208] S1002: Deleting files and vectorization results: deleting the original files on the industry knowledge base server according to the directory information stored in the database, and at the same time, deleting the vectorization results corresponding to the files in the vector database;

[0209] S1003 Delete the directory from the database table: delete the directory information record of the industry knowledge base from the database.

[0210] join Fig.11 , Fig.11 The industry knowledge base query flow chart based on the insurance intelligent assistant system provided by the present invention is as follows: Fig.11 As shown, the industry knowledge base query process provided in this embodiment includes three steps: S1101 opens the knowledge base list, S1102 inputs the search keyword of the knowledge base name or description, and S1103 selects a certain knowledge base and views it. The implementation method of each step is as follows:

[0211] S1101 Open the knowledge base list: open the knowledge base list, and a management interface for querying the industry knowledge base will be presented;

[0212] S1102 Enter the search keyword of the knowledge base name or description: Enter the search keyword of the knowledge base name or description on the interface, and the industry knowledge base management module will match the keyword of the name or description and return a list of matching knowledge bases;

[0213] S1103 Select a knowledge base and view: Select a knowledge base on the list and view its detailed information. The information mainly includes the directory information of the knowledge base (including knowledge base name, description, knowledge base file name, file type, file length, industry knowledge base server IP address, file path, number of vectors and other field contents). You can also view the original text information of the vectorization results.

[0214] In addition, this embodiment also provides a system deployment method, see Fig.12 , Fig.12 This is a schematic diagram of the system deployment of the insurance intelligent assistant system provided by the present invention. Fig.12 As shown, the insurance intelligent assistant system is composed of a user-side mobile phone or computer, a user database server, a management server, a data storage server, an insurance inquiry server, an insurance server and a maintenance server. Among them:

[0215] 1) User-side mobile phone or computer: used to deploy all modules of a single user subsystem except the user database, provide an access interface for users, and store user-related data in the user database; interact with the user database server, insurance inquiry server, insurance server, and maintenance server; each user-side mobile phone or computer participating in the system is deployed in a homogeneous manner to ensure that users can obtain insurance smart assistant services in an easily scalable manner.

[0216] 2) User database server: used to deploy the user database module M56, which is used to store user registration and login information, user setting information, and user profile data formed during the insurance inquiry / insurance / maintenance process.

[0217] 3) Management server: used to deploy modules such as industry knowledge base management M11, prompt word management M12, and large model management M13, and manage the industry knowledge base, prompt word library, and large model.

[0218] 4) Data storage server: used to deploy modules such as industry knowledge base M14, prompt word base M15, and large model base M16, and used to store data such as industry knowledge base, prompt word base, and large models.

[0219] 5) Insurance inquiry server: used to deploy all modules of the insurance inquiry subsystem; accept requests from the user subsystem, interact with the user database server and data storage server, and provide insurance consulting functions for users.

[0220] 6) Insurance server: used to deploy all modules of the insurance subsystem; accept requests from the user subsystem, interact with the user database server and data storage server, and provide users with one-click insurance function.

[0221] 7) Maintenance server: used to deploy all modules of the maintenance subsystem; accept requests from user subsystems, interact with user database servers and data storage servers, and provide users with one-click maintenance functions.

[0222] In summary, the system provided in this embodiment empowers the insurance industry in multiple business links such as insurance inquiry, insurance purchase, and maintenance, helping users to quickly understand insurance and quickly choose the insurance of their choice. The insurance purchase subsystem and the maintenance subsystem automatically connect to the insurance purchase and maintenance interfaces of multiple insurance companies to achieve one-click insurance purchase, one-click renewal, and intelligent claims settlement, greatly improving the efficiency of insurance business. At the same time, the intelligent assistant introduced by the system will replace part of the work of insurance sales, intermediaries, and agents in the links of insurance inquiry, insurance purchase, and maintenance, thereby reducing the industry's labor costs and operating costs.

[0223] In addition, the system provided in this embodiment enables users to process insurance services through the application portal of the insurance intelligent assistant, which assists users in all aspects, and also improves users' peace of mind, understanding of insurance, intelligence of insurance, convenience of renewal, and peace of mind of claims.

[0224] In order to better illustrate the working principle and step flow of the insurance transaction method and device based on a big model and artificial intelligence of the present invention, you can but are not limited to refer to the relevant records above.

[0225] In summary, the embodiment of the present invention provides an insurance transaction method and device based on a large model and artificial intelligence, which obtains similar text data by performing similarity matching between the consultation text data input by the user and the standard text vector, generates prompt words and performs question reasoning on the similar text data in turn, obtains corresponding answer text data, and generates a user profile based on the consultation text data; in response to the demand transaction signal triggered by the user, generates demand response information based on the user profile, feeds it back to the user, and generates insurance transaction information based on the supplementary data input by the user. By matching the similar text data obtained by the similarity matching of the consultation text data, the specific consultation content and consultation direction of the user are determined, helping the user to quickly understand insurance and quickly select the intended insurance; the fully automatic insurance transaction process not only saves the human resource costs of insurance service personnel, but also ensures the security and confidentiality of the generated insurance transaction information.

[0226] The specific embodiments described above further illustrate the purpose, technical solutions and beneficial effects of the present invention. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. It is particularly pointed out that for those skilled in the art, any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention should be included in the scope of protection of the present invention.

Claims

1. An insurance transaction method based on a big model and artificial intelligence, characterized in that: The following steps are involved: Acquire and parse the consultation text data input by the user, and perform similarity matching between the consultation text data and a preset standard text vector to obtain similar text data; Generate prompt words and perform question reasoning on the similar text data in sequence, obtain and display answer text data corresponding to the consultation text data, and generate a corresponding user profile according to the consultation text data; In response to a demand transaction signal triggered by a user, corresponding demand response information is generated according to the user profile, the demand response information is fed back to the user, and insurance transaction information is generated according to supplementary data input by the user.

2. The insurance transaction method based on a big model and artificial intelligence as claimed in claim 1, characterized in that: The step of performing similarity matching between the consultation text data and a preset standard vector text to obtain similar text data includes: Vectorizing the consultation text data into a consultation text vector, and performing similarity matching between the consultation text vector and the standard text vector in a preset vector database to obtain a plurality of matching similarities; The corresponding standard text vectors are sorted in descending order of the matching similarity to generate a similarity vector table, and a corresponding number of similar text data are obtained from the similarity vector table according to a preset number of matches.

3. The insurance transaction method based on a big model and artificial intelligence as claimed in claim 1, characterized in that: The step of sequentially generating prompt words and performing question reasoning on the similar text data to obtain answer text data corresponding to the consultation text data includes: Generate prompt words for the similar text data and the consulting text data using a prompt word template to obtain inference prompt words; The inference prompt words are inferred through the large model, and the parameters of the large model in inference are adjusted according to the preset large model temperature, and then the inference result output by the adjusted large model is used as the answer text data.

4. The insurance transaction method based on a large model and artificial intelligence as claimed in claim 1, characterized in that: The generating a corresponding user profile according to the consultation text data comprises: The consultation text data is used to understand the intention, and the understood user intention is stored in a corresponding user database so that the user database is updated, and then the user profile is constructed based on the data in the updated user database.

5. The insurance transaction method based on big model and artificial intelligence as claimed in claim 1, characterized in that: The demand transaction signal triggered in response to the user includes: an insurance transaction signal, a renewal transaction signal and a claim transaction signal.

6. The insurance transaction method based on a large model and artificial intelligence as claimed in claim 5, characterized in that: When the demand transaction signal is the insurance transaction signal, generating corresponding demand response information according to the user profile, feeding back the demand response information to the user, and generating insurance transaction information according to the supplementary data input by the user, includes: In response to the insurance transaction signal, generating corresponding insurance information according to the user profile and feeding it back to the user, so that the user can supplement the insurance information and generate a corresponding insurance request; Calling the insurance interface of the corresponding insurance company according to the insurance request, and sending the insurance request to the insurance interface, so that the insurance company conducts an insurance review on the user according to the insurance request, and feeds back the generated insurance status information to the user after completing the insurance review; The insurance status information is received, and the user profile is updated according to the insurance status information.

7. The insurance transaction method based on a large model and artificial intelligence as claimed in claim 5, characterized in that: When the demand transaction signal is the renewal transaction signal, generating corresponding demand response information according to the user profile, feeding back the demand response information to the user, and generating insurance transaction information according to the supplementary data input by the user, includes: In response to the renewal transaction signal, generating corresponding renewal information according to the user profile and feeding it back to the user, so that the user confirms the renewal information and generates a corresponding renewal request; Calling a renewal interface of a corresponding insurance company according to the renewal request, and sending the renewal request to the renewal interface, so that the insurance company generates and feeds back a corresponding renewal transaction request according to the renewal request, so that the user adjusts the renewal status information according to the renewal transaction request; The renewal status information fed back by the user is received, and the user profile is updated according to the renewal status information.

8. The insurance transaction method based on a large model and artificial intelligence as claimed in claim 5, characterized in that: When the demand transaction signal is the claim transaction signal, generating corresponding demand response information according to the user profile, feeding back the demand response information to the user, and generating insurance transaction information according to the supplementary data input by the user, includes: In response to the claim transaction signal, generating corresponding claim information according to the user profile and feeding it back to the user, so that the user can supplement the claim information and generate a corresponding claim request; Calling a claim interface of a corresponding insurance company according to the claim request, and sending the claim request to the claim interface, so that the insurance company conducts a claim review of the user according to the claim request, and feeds back the generated claim status information to the user after completing the claim review; The claim status information is received, and the user profile is updated according to the claim status information.

9. The insurance transaction method based on a large model and artificial intelligence as claimed in claim 1, characterized in that: Before obtaining and parsing the consultation text data input by the user, the method further includes: The knowledge base file to be processed is obtained, and the knowledge base file to be processed is vectorized, the vectorization result is stored in a vector database, and the parameters of the vector database are updated according to the directory information of the knowledge base file to obtain the updated vector database.

10. An insurance transaction device based on a large model and artificial intelligence, characterized in that: The insurance transaction device includes a vector database construction module, a consultation text matching module, an answer text reasoning module and a transaction demand response module; The consulting text matching module is used to obtain and analyze the consulting text data input by the user, and perform similarity matching between the consulting text data and a preset standard text vector to obtain similar text data; The answer text reasoning module is used to perform prompt word generation and question reasoning on the similar text data in sequence, obtain answer text data corresponding to the consultation text data, and generate a corresponding user profile according to the consultation text data; The transaction demand response module is used to respond to the demand transaction signal triggered by the user, generate corresponding demand response information according to the user profile, feed back the demand response information to the user, and generate insurance transaction information according to the supplementary data input by the user; The vector database construction module is used to obtain the knowledge base file to be processed, and vectorize the knowledge base file to be processed, store the vectorization result in the vector database, and update the parameters of the vector database according to the directory information of the knowledge base file to obtain the updated vector database.

Images