Insurance verbal skill generation method and device, electronic equipment and storage medium

Abstract

The invention provides an insurance verbal skill generation method and device, electronic equipment and a storage medium, and relates to the technical field of artificial intelligence and the field of financial science and technology. The method comprises the steps of obtaining an initial attention insurance text of a target interested insured object; obtaining object information of a target entity object to obtain target entity object information; performing text semantic correction on the initial concerned insurance text through an insurance text correction model to obtain a target concerned insurance text; performing clause screening on an insurance clause database according to the target concerned insurance text to obtain target insurance clauses; performing verbal skill generation on the target concerned insurance text, the target entity object information and the target insurance clause through a large language model to obtain an insurance verbal skill text; and performing audio generation on the insurance verbal skill text through an audio generation model to obtain a target insurance verbal skill audio. According to the invention, customized insurance verbal skill audios can be provided for different interested insured objects.

Description

Technical Field

[0001] The present application relates to the fields of artificial intelligence technology and financial technology, and in particular to a method and device for generating insurance scripts, an electronic device, and a storage medium. Background Art

[0002] In the field of fintech, insurance companies rely on online channels (websites, apps, social media, etc.) for business promotion and customer service, such as placing insurance advertisements on websites. Audio promotional materials are an important way of communication, which can effectively attract the attention of potential customers and increase the popularity of insurance products.

[0003] Among the related technologies, insurance audio production technology mainly relies on manual script writing and dubbing, which is not only time-consuming and labor-intensive, but also difficult to meet personalized needs. Summary of the invention

[0004] The main purpose of the embodiments of the present application is to propose an insurance script generation method and device, electronic device, and storage medium, which can provide customized insurance script audio for insured persons with different interests.

[0005] To achieve the above-mentioned purpose, a first aspect of an embodiment of the present application proposes a method for generating insurance scripts, the method comprising:

[0006] Acquire the initial insurance text of the target interested insured object; wherein the target interested insured object tends to insure the target entity object;

[0007] Acquire the object information of the target entity object to obtain the target entity object information;

[0008] Performing text semantic correction on the initial insurance text of interest through a preset insurance text correction model to obtain a target insurance text of interest;

[0009] Screening a preset insurance clause database according to the target insurance text to obtain target insurance clauses;

[0010] Generate speech for the target insurance text, the target entity object information and the target insurance clause through a preset large language model to obtain an insurance speech text;

[0011] The insurance speech text is audio-generated using a preset audio generation model to obtain target insurance speech audio.

[0012] In some embodiments, the acquiring the object information of the target entity object to obtain the target entity object information includes:

[0013] Acquire the category of the target entity object to obtain the entity object category;

[0014] Searching a preset category and information item mapping library according to the entity object category to obtain a target information item;

[0015] Information is collected according to the target information item to obtain the target entity object information.

[0016] In some embodiments, collecting information according to the target information item to obtain the target entity object information includes:

[0017] Acquire a target entity object image uploaded by the target interested insured person;

[0018] Searching the preset information items and the model mapping library according to the target information items to obtain a model search result; wherein the model search result indicates whether the target information extraction model is found or not;

[0019] If the model search result indicates that the target information extraction model is found, the target entity object image is identified by using the found target information extraction model to obtain the target entity object information.

[0020] In some embodiments, after searching the preset information item and the model mapping library according to the target information item to obtain the model search result, the method further includes:

[0021] If the model search result indicates that the target information extraction model is not found, sending an information filling request to the target interested insured person;

[0022] Receiving information filled in by the target interested insured person based on the information filling request, and obtaining target filled-in information;

[0023] The target entity object information is updated according to the target filled information.

[0024] In some embodiments, the generating of the speech text for the target insurance text, the target entity object information and the target insurance clause by using a preset large language model to obtain the insurance speech text includes:

[0025] Parsing the target insurance clauses through the large language model to obtain insurance compensation information and insurance restriction information;

[0026] Performing insurance detection on the insurance restriction information and the target entity object information by using the large language model to obtain an insurance detection result;

[0027] If the insurance detection result indicates that the target entity object can be insured, the large language model is used to generate speech for the target insurance text, the target entity object information and the insurance claim information to obtain an insurance speech text;

[0028] The insurance concern speech text is determined as the insurance speech text.

[0029] In some embodiments, after performing insurance detection on the insurance restriction information and the target entity object information by using the large language model to obtain the insurance detection result, performing speech generation on the target insurance text, the target entity object information and the target insurance clause by using the preset large language model to obtain the insurance speech text further includes:

[0030] If the insurance detection result indicates that the target entity object cannot be insured, then generating information for the target insurance text and preset reference recommendation information through the large language model to obtain insurance recommendation information;

[0031] Generate speech for the insurance recommendation information and the target entity object information by using the large language model to obtain insurance recommendation speech text;

[0032] The insurance recommendation speech text is determined as the insurance speech text.

[0033] In some embodiments, obtaining the initial insurance text of interest to the target insured person includes:

[0034] Acquire the voice data input by the target interested insured person from a preset consultation channel to obtain the insurance consultation voice data of the target;

[0035] Performing text conversion on the insurance consultation voice data of the subject through a preset insurance voice-to-text model to obtain a first insurance consultation text of the subject;

[0036] Acquire the text data input by the target interested insured person from the consultation channel to obtain the second insurance consultation text of the target;

[0037] The first insurance consultation text of the object and the second insurance consultation text of the object are integrated to obtain the initial insurance text of interest.

[0038] To achieve the above-mentioned purpose, a second aspect of an embodiment of the present application provides an insurance speech generation device, the device comprising:

[0039] A text acquisition module is used to acquire the initial insurance text of the target interested insured object; wherein the target interested insured object tends to insure the target entity object;

[0040] An information acquisition module, used to acquire the object information of the target entity object, and obtain the target entity object information;

[0041] A text correction module, used for performing text semantic correction on the initial insurance text of interest through a preset insurance text correction model to obtain a target insurance text of interest;

[0042] A clause screening module, used to screen the preset insurance clause database according to the target insurance text to obtain the target insurance clause;

[0043] A text generation module, used to generate speech for the target insurance text, the target entity object information and the target insurance clause through a preset large language model to obtain an insurance speech text;

[0044] The audio generation module is used to generate audio for the insurance speech text through a preset audio generation model to obtain the target insurance speech audio.

[0045] To achieve the above-mentioned purpose, the third aspect of an embodiment of the present application proposes an electronic device, which includes a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the insurance script generation method described in the first aspect is implemented.

[0046] To achieve the above-mentioned purpose, the fourth aspect of an embodiment of the present application proposes a storage medium, which is a computer-readable storage medium, and the storage medium stores a computer program. When the computer program is executed by a processor, the insurance script generation method described in the first aspect is implemented.

[0047] The insurance speech generation method, insurance speech generation device, electronic device and computer-readable storage medium proposed in the present application can effectively and accurately find the target insurance terms by first using the insurance text correction model to perform text semantic correction on the initial insurance text after obtaining the initial insurance text of the target interest insured object, and then screening the insurance terms database according to the obtained target insurance text, so as to effectively and accurately find the target insurance terms. The target entity object information is also obtained, and then the target insurance text, target entity object information and target insurance terms are generated through the large language model to obtain the insurance speech text, and then the target insurance speech audio is obtained through the audio generation model. It can be seen that the present application generates speech through the above three dimensions of information (i.e., the target insurance text, target entity object information and target insurance terms), and can generate different target insurance speech audio based on the differences between the target interest insured object and the target entity object, so it can provide customized insurance speech audio for different interest insured objects.

[0048] Additional aspects and advantages of the present application will be given in part in the description below, and in part will become apparent from the description below, or will be learned through the practice of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Figure 1 It is a flow chart of the insurance speech generation method provided in the embodiment of the present application;

[0050] Figure 2 yes Figure 1 Flow chart of step 101 in;

[0051] Figure 3 yes Figure 1 Flowchart of step 102 in FIG.

[0052] Figure 4 yes Figure 3 Flow chart of step 303 in;

[0053] Figure 5 is a flow chart of a method for generating insurance scripts provided by another embodiment of the present application;

[0054] Figure 6 yes Figure 1 Flowchart of step 105 in;

[0055] Figure 7 yes Figure 1 Another flow chart of step 105 in;

[0056] Figure 8 This is a module structure diagram of the insurance speech generation device provided in an embodiment of the present application;

[0057] Fig. 9 It is a schematic diagram of the hardware structure of the electronic device provided in the embodiment of the present application. DETAILED DESCRIPTION

[0058] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0059] It should be noted that, although the functional modules are divided in the device schematic diagram and the logical order is shown in the flowchart, in some cases, the steps shown or described may be performed in a different order than the module division in the device or the order in the flowchart. The terms "first", "second", etc. in the specification, claims and the above drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence.

[0060] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art to which this application belongs. The terms used herein are only for the purpose of describing the embodiments of this application and are not intended to limit this application.

[0061] First, some nouns involved in this application are analyzed:

[0062] Artificial intelligence (AI) is a new technical science that studies and develops theories, methods, technologies and application systems for simulating, extending and expanding human intelligence. AI is a branch of computer science that attempts to understand the essence of intelligence and produce a new intelligent machine that can respond in a similar way to human intelligence. Research in this field includes robots, language recognition, image recognition, natural language processing and expert systems. AI can simulate the information process of human consciousness and thinking. AI is also a theory, method, technology and application system that uses digital computers or machines controlled by digital computers to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use knowledge to obtain the best results.

[0063] Natural language processing (NLP): NLP uses computers to process, understand and apply human languages ​​(such as Chinese and English). NLP is a branch of artificial intelligence and an interdisciplinary subject between computer science and linguistics. It is often referred to as computational linguistics. Natural language processing includes grammatical analysis, semantic analysis, and text understanding. Natural language processing is often used in technical fields such as machine translation, handwritten and printed character recognition, speech recognition and text-to-speech conversion, information image processing, information extraction and filtering, text classification and clustering, public opinion analysis and opinion mining. It involves data mining related to language processing, machine learning, knowledge acquisition, knowledge engineering, artificial intelligence research, and linguistic research related to language computing.

[0064] The embodiments of the present application propose an insurance speech generation method, an insurance speech generation device, an electronic device and a computer-readable storage medium, which can automatically generate corresponding speech text and then generate speech audio, thereby realizing fast, efficient and customized insurance speech audio production.

[0065] The insurance speech generation method provided in the embodiment of the present application can be applied to the terminal and the server side, and can also be software running on the server side. The server side can be configured as an independent physical server, or a server cluster or distributed system composed of multiple physical servers, or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, CDN, and big data and artificial intelligence platforms; the software can be an application that implements the insurance speech generation method, but is not limited to the above forms.

[0066] The present application can be used in many general or special computer system environments or configurations. For example: server computers, multiprocessor systems, microprocessor-based systems, set-top boxes, programmable consumer electronics, network PCs, distributed computing environments including any of the above systems or devices, etc. The present application can be described in the general context of computer executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform specific tasks or implement specific abstract data types. The present application can also be practiced in distributed computing environments, in which tasks are performed by remote processing devices connected through a communication network. In a distributed computing environment, program modules can be located in local and remote computer storage media including storage devices.

[0067] The embodiments of the present application provide an insurance script generation method and an insurance script generation device, an electronic device, and a storage medium, which are specifically explained through the following embodiments. First, the insurance script generation method in the embodiments of the present application is described.

[0068] It should be noted that in each specific implementation method of the present application, when it comes to processing data related to the user's identity or characteristics such as the user's behavioral data, the user's permission or consent will be obtained first, and the collection, use and processing of such data will comply with relevant laws, regulations and standards.

[0069] Reference Figure 1 , Figure 1 It is an optional flowchart of the insurance script generation method provided in the embodiment of the present application, which may include but is not limited to steps 101 to 106.

[0070] Step 101, obtaining the initial insurance text of the target interested insured object; wherein the target interested insured object tends to insure the target entity object;

[0071] Step 102, obtaining object information of a target entity object, and obtaining the target entity object information;

[0072] Step 103, performing text semantic correction on the initial insurance text of interest by using a preset insurance text correction model to obtain a target insurance text of interest;

[0073] Step 104, screening the preset insurance clause database according to the target insurance text to obtain the target insurance clause;

[0074] Step 105, generating speech texts for the target insurance text, target entity object information and target insurance clauses through a preset large language model to obtain insurance speech texts;

[0075] Step 106, generate audio for the insurance talk text through a preset audio generation model to obtain the target insurance talk audio.

[0076] In the steps 101 to 106 shown in the embodiment of the present application, after obtaining the initial insurance text of interest of the insured object, the insurance text correction model is first used to perform text semantic correction on the initial insurance text of interest, and then the insurance clause database is screened according to the obtained target insurance text of interest, so that the target insurance clause can be found effectively and accurately. The target entity object information is also obtained, and then the target insurance text of interest, the target entity object information and the target insurance clause are generated through the preset large language model to obtain the insurance text of interest, and then the target insurance audio is obtained through the audio generation model. It can be seen that the present application generates speech through three dimensions of information (i.e., the target insurance text of interest, the target entity object information and the target insurance clause), and can generate different target insurance audio based on the differences between the target interest insured object and the target entity object, so it can provide customized insurance audio for different interest insured objects.

[0077] In step 101 of some embodiments, the initial insurance text of the target interest insurance object is obtained. The target interest insurance object is an object (user, customer) that is interested in insurance. The target entity object is an insured entity, such as a person, an object, or a pet. The target interest insurance object tends to insure the target entity object, which can also be said to be the target interest insurance object is interested in insuring the target entity object.

[0078] The target interested insured persons may be people who have consulted or conducted insurance business on insurance websites or related offline outlets. In one example, insurance business records of insurance websites or offline outlets systems may be obtained, and the target interested insured persons may be determined based on the insurance business records.

[0079] After determining the target interested insured person, obtain the initial insurance text. The initial insurance text is used to indicate the insurance information that the target interested insured person cares about. It should be noted that there are many types of existing insurance products. When generating insurance scripts, first determine which type of insurance the target person is more interested in, so as to improve the accuracy of the generated insurance scripts.

[0080] In one embodiment, referring to Figure 2 , step 101 may include:

[0081] Step 201, obtaining voice data input by the target interested insured person from a preset consultation channel to obtain the insurance consultation voice data of the target;

[0082] Step 202, converting the insurance consultation voice data of the subject into text using a preset insurance voice-to-text model to obtain a first insurance consultation text of the subject;

[0083] Step 203, obtaining text data input by the target interested insured person from the consultation channel, and obtaining the second insurance consultation text of the target;

[0084] Step 204, integrating the object's first consultation insurance text and the object's second consultation insurance text to obtain an initial focus insurance text.

[0085] Specifically, the consulting channel refers to a channel that provides insurance consulting services to the target interested insured person (user / customer), such as a website, APP, or social media. Considering the wide variety of insurance products, the target interested insured person may not be able to accurately express the information about their concern for a certain type of insurance. Therefore, this embodiment will receive voice data and text data at the same time, and then integrate them to obtain the initial concerned insurance text.

[0086] The insurance speech text model is a deep learning network model that can convert the input insurance consultation speech into insurance consultation text. In one embodiment, in the model training phase, the insurance speech text model is trained using a sample insurance consultation speech data set; in the model use phase, the object insurance consultation speech data is input into the insurance speech text model for text conversion to obtain the object's first insurance consultation text.

[0087] The above-mentioned sample consultation insurance voice data set contains multiple sample consultation insurance voice data. The sample consultation insurance voice data is voice data input from the consultation channel by multiple different interested insured objects. The sample consultation insurance voice data has a sample label, and the sample label indicates the reference conversion text corresponding to the sample consultation insurance voice data. The sample label can be obtained by manually annotating the sample consultation insurance voice data. In the model training stage, the sample consultation insurance voice data is input into the initial voice-text model to obtain the predicted conversion text; the loss of the predicted conversion text and the reference conversion text is calculated through the loss function to obtain the text loss data; the parameters of the initial voice-text model are adjusted according to the text loss data to obtain the insurance voice-text model.

[0088] The benefit of the above-mentioned embodiment of step 201 to step 204 is that, on the basis of being able to obtain the initial insurance text of interest, the information richness of the initial insurance text of interest is improved.

[0089] In step 102 of some embodiments, object information of a target entity object is acquired to obtain the target entity object information.

[0090] The target entity object can be determined in the following manner: (1) After obtaining the initial insurance text of the target interest insured object, entity extraction is performed on the initial insurance text to obtain the target entity object. (2) After obtaining the initial insurance text of the target interest insured object, an entity object confirmation request is sent to the target interest insured object, and the target entity object sent by the target interest insured object based on the entity object confirmation request is received.

[0091] The target entity object information can be regarded as a description of the target entity object. For example, the target entity object information includes age, gender, region, health status, etc.

[0092] In one embodiment, referring to Figure 3 , step 102 may include:

[0093] Step 301, obtaining the category of the target entity object to obtain the entity object category;

[0094] Step 302, searching a preset category and information item mapping library according to the entity object category to obtain a target information item;

[0095] Step 303: collect information according to the target information item to obtain the target entity object information.

[0096] In step 301, the entity object category may indicate the category of the target entity object. For example, if the target entity object is a cat or a dog, then the entity object category is a cat or a dog. In one example, an image of the target entity object may be captured to obtain a target entity object image; the target entity object image is input into an image classification model to obtain the entity object category. In another example, a category filling request may be sent to the target interest insured object, and then the category filled in by the target interest insured object based on the category filling request may be received to obtain the entity object category.

[0097] In step 302, the category and information item mapping library stores the mapping relationship between categories and information items. For example, if the category is a cat, the target information items include age, weight, mood, health status, etc. For another example, if the category is a dog, the target information items include age, weight, mood, fur color, body shape, health status, etc.

[0098] In step 303, after the target information item is determined, information is collected according to the target information item to finally obtain the target entity object information.

[0099] The benefit of the above-mentioned embodiment of step 301 to step 303 is that by distinguishing different categories of target entity objects, information collection can be carried out in a targeted manner, which can reduce the collection of unnecessary information, thereby improving information collection efficiency and saving processing resources.

[0100] In one embodiment, referring to Figure 4 , step 303 may include:

[0101] Step 401, obtaining a target entity object image uploaded by a target interested insured person;

[0102] Step 402, searching the preset information item and the model mapping library according to the target information item to obtain the model search result;

[0103] Step 403: If the model search result indicates that the target information extraction model is found, the target entity object image is recognized by using the found target information extraction model to obtain the target entity object information.

[0104] In step 401, in combination with the above, the target entity object image can be obtained by collecting images of the target entity object. The target entity object image contains appearance information of the target entity object, such as category, fur color, body shape, expression, etc.

[0105] In step 402, the information item and model mapping library stores the mapping relationship between the information item and the model. For example, if the information item is age, the target information extraction model is the age extraction model. For another example, if the information item is emotion, the target information extraction model is the emotion extraction model.

[0106] Specifically, the information item and the model mapping library are searched according to the target information item. If the target information extraction model can be found, the model search result indicates that the target information extraction model is found. If the target information extraction model cannot be found, the model search result indicates that the target information extraction model is not found. For example, for age and emotion, the age extraction model and the emotion extraction model can be found, so the model search result indicates that the target information extraction model is found. For another example, for health status, there is no corresponding health status extraction model, so the model search result indicates that the target information extraction model is not found.

[0107] In step 403, if the model search result indicates that the target information extraction model is found, for example, the target information extraction model is an age extraction model and an emotion extraction model, then the age of the target entity object image is recognized according to the age extraction model to obtain the age of the target entity object; the emotion of the target entity object image is recognized according to the emotion extraction model to obtain the emotion of the target entity object; the target entity object age and the target entity object emotion are integrated to obtain the target entity object information.

[0108] The benefit of the above-mentioned embodiment of step 401 to step 403 is that the target entity object information is obtained by identifying the target entity object image, thereby improving the efficiency and flexibility of information acquisition.

[0109] In one embodiment, referring to Figure 5 After step 402, the insurance script generation method further includes:

[0110] Step 501: If the model search result indicates that the target information extraction model is not found, an information filling request is sent to the target interested insured person;

[0111] Step 502, receiving information filled in by the target interested insured person based on the information filling request, and obtaining target filled-in information;

[0112] Step 503: Update the target entity object information according to the target filled information.

[0113] In this embodiment, considering that some information cannot be extracted from the target entity object image, that is, some target information items do not have corresponding target information extraction models in the information item and model mapping library, an information filling request can be generated based on the target information item for which the target information extraction model has not been found, and sent to the target interest insured object. After the target interest insured object receives the information filling request, the information filled in by the target interest insured object based on the information filling request is received to obtain the target filling information, and then the target entity object information is updated according to the target filling information. In this way, the information richness of the target entity object information can be further improved.

[0114] In one example, the target information item is a health status. If the information item and the model mapping library do not store a health status extraction model corresponding to the health status, the model search result indicates that the target information extraction model is not found. A health status filling request can be generated based on the health status, and the health status filling request can be sent to the target interest insured object, and the target interest insured object fills in the health status. The health status filled in by the target interest insured object is received, the health status filling information is obtained, and then the health status filling information is merged into the target entity object information. Among them, the health status filling information may include vaccination records, disease history, allergy information, etc.

[0115] The benefit of the above-mentioned embodiment of steps 501 to 503 is that it can improve the information richness of the target entity information, help improve the fit between the insurance script and the target entity object, and have a high degree of customization.

[0116] Before step 103, in one embodiment, the initial insurance text is subjected to special symbol removal. Removing special symbols is one of the most basic steps in preprocessing. There may be various special symbols in the text, such as punctuation marks, numbers, English letters, etc. These symbols often have no practical meaning in speech synthesis, but will interfere with the model's understanding of semantics. Therefore, these special symbols need to be deleted from the text or replaced with spaces. For example, "$50" can be replaced with "50".

[0117] In step 103 of some embodiments, the text semantics of the initial insurance text of interest is corrected using a preset insurance text correction model to obtain a target insurance text of interest.

[0118] The insurance text correction model is a model trained with an insurance text dataset. The sample insurance texts in the insurance text dataset contain various types of insurance products. The insurance text correction model can predict the output text containing at least one insurance product based on the input text. For example, the initial insurance text of interest is "I want to know about pet products", and the insurance text correction model can obtain the target insurance text of interest as "I want to know about pet insurance". For another example, the initial insurance text of interest is "I have a kind of cat and want to know about related insurance", and the insurance text correction model can obtain the target insurance text of interest as "I want to know about cat-related pet insurance".

[0119] In step 104 of some embodiments, a preset insurance clause database is screened according to the target insurance text to obtain the target insurance clause. The insurance clause database stores a plurality of candidate insurance clauses. The candidate insurance clauses can be screened according to the distance between the target insurance text and the candidate insurance clauses to obtain the target insurance clause. The distance formula can be a cosine similarity formula, a Euclidean distance formula, etc.

[0120] In one embodiment, step 104 includes:

[0121] Perform semantic recognition on the target insurance text to obtain the insurance semantic vector;

[0122] Perform semantic recognition on candidate insurance clauses in the insurance clause database to obtain clause semantic vectors;

[0123] Calculate similarity based on the insurance semantic vector and the clause semantic vector to obtain semantic similarity;

[0124] The candidate insurance clauses are screened according to semantic similarity to obtain the target insurance clauses.

[0125] In this embodiment, the target insurance text and the candidate insurance clauses can be input into the double tower model, and the insurance semantic vector and the clause semantic model are first obtained respectively, and then similarity is calculated based on the insurance semantic vector and the clause semantic vector to obtain semantic similarity. The formula for similarity calculation can be the cosine similarity formula.

[0126] In one example, the candidate insurance clauses may be sorted from high to low according to semantic similarity to obtain a sorting result; and the first candidate insurance clause is selected from the sorting result as the target insurance clause.

[0127] In one example, the target insurance terms include product name, insurance period, minimum quote, maximum insured amount, insured age and insurance instructions. Insurance instructions may include: 1. This product is limited to one pet, and only insured pet cats or dogs are insured. Please fill in the true situation of your pet carefully when applying for insurance. If the insurer finds obvious differences with the insurance information (including age, pet photos, etc.) when applying for claims, and the insurer cannot identify the same pet, you will not be able to obtain compensation. 2. The age range of pets insured for the first time is 3 months to 10 years old, and renewal is not affected by the age range. If this product has been discontinued at the end of the insurance period, the renewal application will no longer be accepted, but reasonable suggestions and assistance will be provided to you for other insurance products. 3. The waiting period for the first insurance of this product is 30 days, and there is no waiting period for renewal. This insurance does not bear compensation liability for diseases or symptoms that existed before the insurance contract takes effect and during the waiting period after the insurance contract takes effect. 4. This product has a single deductible of 200 yuan for pet medical expenses, a compensation ratio of 60%, a single compensation limit of 500 yuan, and an annual cumulative compensation limit of 1,000 yuan.

[0128] The benefit of the above embodiment is that the accuracy of determining the target insurance clauses is improved by using semantic similarity to screen the clauses. Here, the semantic similarity of this embodiment corresponds to the text semantic correction in the above step, further improving the relevance of the target insurance clauses and the initial insurance text of the target interested insured person.

[0129] In step 105 of some embodiments, a preset large language model is used to generate speech for the target insurance text, target entity object information and target insurance clauses to obtain the insurance speech text.

[0130] A large language model is an artificial intelligence model trained with a large amount of text data. Its main purpose is to understand and generate natural language. Qwen72B can be used as a large language model.

[0131] In one embodiment, referring to Figure 6 , step 105 may include:

[0132] Step 601, parsing the target insurance clauses through a large language model to obtain insurance payment information and insurance restriction information;

[0133] Step 602, performing insurance detection on the insurance restriction information and the target entity object information through a large language model to obtain an insurance detection result;

[0134] Step 603: If the insurance detection result indicates that the target entity object can be insured, a large language model is used to generate a speech for the target insurance text, the target entity object information, and the insurance claim information to obtain an insurance speech text;

[0135] Step 604, determining the insurance concern speech text as the insurance speech text.

[0136] In step 601, the target insurance clause is input into the large language model for information analysis, and the insurance compensation information and insurance restriction information can be obtained. For example, the insurance compensation information includes: the single deductible of pet medical expenses of this product is 200 yuan, the compensation ratio is 60%, the single compensation limit is 500 yuan, and the annual cumulative compensation limit is 1,000 yuan. The insurance restriction information includes: the age range of pets insured for the first time is 3 months to 10 years old, the waiting period for the first insurance of this product is 30 days, and the insurance does not bear compensation liability for existing diseases or symptoms before and after the insurance contract takes effect.

[0137] In step 602, the insurance restriction information and the target entity object information are input into the large language model for insurance detection, and the insurance detection result can be obtained. The insurance detection result indicates whether the target entity object can be insured or not. For example, if the target entity object information includes an age of 11 years old, which cannot meet the pet age range in the insurance restriction information, the target entity object cannot be insured. For another example, if the target entity object information includes a health condition of suffering from XX disease, which cannot make the insurance company bear compensation liability, the target entity object cannot be insured.

[0138] Then, execute step 603 to obtain the insurance attention script text, and then execute step 604 to finally obtain the insurance script text. For example, the insurance script text may be "XX pet insurance can protect pets from all diseases and accidents. Different versions have different coverage amounts. When taking out insurance, you need to fill in the actual situation of the pet to avoid differences in claims. The insurance period is 1 year. For pet medical expenses, there are different deductibles and compensation ratios."

[0139] It should be noted that the purpose of the insurance script text in this embodiment is to guide the target interested insured object to insure the target entity object. Therefore, in this embodiment, if the insurance detection result has indicated that the target entity object can be insured, there is no need to generate scripts based on the entire target insurance clause, but the insurance compensation information, the target insurance text and the target entity object information contained in the target insurance clause are used to generate scripts to obtain the insurance script text. In this way, the script generation efficiency can be improved, and the insurance success rate can also be improved.

[0140] In another embodiment, referring to Figure 7 After step 602, step 105 may further include:

[0141] Step 701: if the insurance detection result indicates that the target entity object cannot be insured, then information generation is performed on the target insurance text and preset reference recommendation information through a large language model to obtain insurance recommendation information;

[0142] Step 702, generating speech text for the insurance recommendation information and the target entity object information through a large language model to obtain the insurance recommendation speech text;

[0143] Step 703, determining the insurance recommendation script text as the insurance script text.

[0144] In step 701, the reference recommendation information is information set in advance and stored in a database. The reference recommendation information is used to indicate information about recommended services provided to the target interested insured person. For example, the reference recommendation information includes food, health check-ups, training courses, etc. for a specific age group. For health reminders, regular health check-up reminders, vaccination reminders, etc. can be provided according to the age and health status of the pet, so that the health of the pet can be better managed.

[0145] In step 702, the insurance recommendation information and the target entity object information are input into the large language model for speech generation, and the insurance recommendation speech text can be obtained. Then in step 703, the insurance recommendation speech text is determined as the insurance speech text. For example, the insurance speech text can be "You can choose to subscribe to the health check reminder service for your pet. We will provide you with regular health check reminders, vaccination reminders, etc., so that you can better manage your pet's health."

[0146] It should be noted that the purpose of the insurance script text of this embodiment is to guide the target interest insured object to subscribe to the recommended service for the target entity object. Therefore, in this embodiment, if the insurance detection result has indicated that the target entity object cannot be insured, there is no need to generate scripts based on the insurance claim information, but to generate scripts based on the reference recommendation information and the target entity object information to obtain the insurance recommendation script text. In this way, when the target entity object cannot be insured, the target interest insured object can be guided to subscribe to the recommended service for the target entity object.

[0147] In step 106 of some embodiments, the insurance speech text is audio-generated by a preset audio generation model to obtain the target insurance speech audio. The audio generation model is a deep learning network model that generates audio based on text. The audio generation model can be AudioLDM2.

[0148] After step 106, in one embodiment, the target insurance speech audio may be optimized, including adjusting parameters such as speech speed, volume, and intonation, so that the target insurance speech audio is more in line with human hearing habits. The optimized target insurance speech audio may also be saved in a common audio format, such as MP3, WAV, etc.; and then the generated target insurance speech audio may be applied to online channels such as websites, apps, and social media for delivery.

[0149] In summary, the present application can at least achieve the following effects: 1. The artificial intelligence technology is used to realize the automatic conversion of text to audio, which greatly improves the efficiency of insurance script production. 2. Customized insurance script audio can be quickly generated according to the needs of insured persons with different target interests to meet diverse insurance needs. 3. The generated target insurance script audio content is of high quality and has high auditory appeal, which is conducive to improving the success rate of insurance or the success rate of recommended service subscription.

[0150] See also Figure 8 The embodiment of the present application also provides an insurance speech generation device, which can implement the above-mentioned insurance speech generation method. Figure 8The module structure block diagram of the insurance speech generation device provided in the embodiment of the present application includes: a text acquisition module 801, an information acquisition module 802, a text correction module 803, a clause screening module 804, a text generation module 805 and an audio generation module 806. Among them, the text acquisition module 801 is used to obtain the initial insurance text of the target interest insurance object, and the target interest insurance object tends to insure the target entity object; the information acquisition module 802 is used to obtain the object information of the target entity object and obtain the target entity object information; the text correction module 803 is used to perform text semantic correction on the initial insurance text of interest through a preset insurance text correction model to obtain the target insurance text of interest; the clause screening module 804 is used to perform clause screening on the preset insurance clause database according to the target insurance text of interest to obtain the target insurance clause; the text generation module 805 is used to generate speech for the target insurance text of interest, the target entity object information and the target insurance clause through a preset large language model to obtain the insurance speech text; the audio generation module 806 is used to generate audio for the insurance speech text through a preset audio generation model to obtain the target insurance speech audio.

[0151] In one embodiment, the insurance script generation device also includes an information update module, which is used to send an information filling request to the target interest insured object if the model search result indicates that the target information extraction model is not found, receive the information filled in by the target interest insured object based on the information filling request, obtain the target filled information, and update the target entity object information according to the target filled information.

[0152] In one embodiment, the text generation module is specifically used to:

[0153] The target insurance clause is parsed through a large language model to obtain insurance compensation information and insurance restriction information;

[0154] Performing insurance detection on the insurance restriction information and the target entity object information by using a large language model to obtain an insurance detection result;

[0155] If the insurance detection result indicates that the target entity object can be insured, the target insurance text, the target entity object information and the insurance claim information are used to generate a speech text through a large language model to obtain an insurance speech text, and the insurance speech text is determined as an insurance speech text;

[0156] If the insurance detection result indicates that the target entity object cannot be insured, the target insurance text and the preset reference recommendation information are generated through a large language model to obtain insurance recommendation information, and the insurance recommendation information and the target entity object information are generated through a large language model to obtain insurance recommendation speech text, and the insurance recommendation speech text is determined as the insurance speech text.

[0157] It should be noted that the specific implementation of the insurance speech generation device is basically the same as the specific implementation example of the above-mentioned insurance speech generation method, and will not be repeated here.

[0158] The embodiment of the present application also provides an electronic device, which includes: a memory, a processor, a program stored in the memory and executable on the processor, and a data bus for realizing connection and communication between the processor and the memory, and the program is executed by the processor to realize the above-mentioned insurance speech generation method. The electronic device can be any intelligent terminal including a tablet computer, a car computer, etc.

[0159] See also Fig. 9 , Fig. 9 The hardware structure of an electronic device of another embodiment is illustrated, and the electronic device includes:

[0160] The processor 901 may be implemented by a general-purpose CPU (Central Processing Unit), a microprocessor, an application-specific integrated circuit (ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of the present application.

[0161] The memory 902 can be implemented in the form of a read-only memory (ROM), a static storage device, a dynamic storage device, or a random access memory (RAM). The memory 902 can store an operating system and other application programs. When the technical solution provided in the embodiment of this specification is implemented by software or firmware, the relevant program code is stored in the memory 902, and the processor 901 calls and executes the insurance speech generation method of the embodiment of this application;

[0162] Input / output interface 903, used to implement information input and output;

[0163] Communication interface 904, used to realize communication interaction between the device and other devices, which can be realized through wired mode (such as USB, network cable, etc.) or wireless mode (such as mobile network, WI FI, Bluetooth, etc.);

[0164] A bus 905 that transmits information between various components of the device (e.g., the processor 901, the memory 902, the input / output interface 903, and the communication interface 904);

[0165] The processor 901 , the memory 902 , the input / output interface 903 and the communication interface 904 are connected to each other in communication within the device via a bus 905 .

[0166] The embodiment of the present application also provides a storage medium, which is a computer-readable storage medium used for computer-readable storage. The storage medium stores one or more programs, and the one or more programs can be executed by one or more processors to implement the above-mentioned insurance script generation method.

[0167] The memory, as a non-transient computer-readable storage medium, can be used to store non-transient software programs and non-transient computer executable programs. In addition, the memory may include a high-speed random access memory, and may also include a non-transient memory, such as at least one disk storage device, a flash memory device, or other non-transient solid-state storage device. In some embodiments, the memory may optionally include a memory remotely disposed relative to the processor, and these remote memories may be connected to the processor via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0168] The embodiments described in the embodiments of the present application are intended to more clearly illustrate the technical solutions of the embodiments of the present application and do not constitute a limitation on the technical solutions provided in the embodiments of the present application. Those skilled in the art will appreciate that with the evolution of technology and the emergence of new application scenarios, the technical solutions provided in the embodiments of the present application are also applicable to similar technical problems.

[0169] Those skilled in the art will appreciate that the technical solutions shown in the figures do not constitute a limitation on the embodiments of the present application, and may include more or fewer steps than shown in the figures, or a combination of certain steps, or different steps.

[0170] The device embodiments described above are merely illustrative, and the units described as separate components may or may not be physically separated, that is, they may be located in one place or distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0171] Those skilled in the art will appreciate that all or some of the steps in the methods disclosed above, and the functional modules / units in the systems and devices may be implemented as software, firmware, hardware, or a suitable combination thereof.

[0172] The terms "first", "second", "third", "fourth", etc. (if any) in the specification of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0173] It should be understood that in the present application, "at least one (item)" means one or more, and "plurality" means two or more. "And / or" is used to describe the association relationship of associated objects, indicating that three relationships may exist. For example, "A and / or B" can mean: only A exists, only B exists, and A and B exist at the same time, where A and B can be singular or plural. The character " / " generally indicates that the objects associated before and after are in an "or" relationship. "At least one of the following" or similar expressions refers to any combination of these items, including any combination of single or plural items. For example, at least one of a, b or c can mean: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, c can be single or multiple.

[0174] In the several embodiments provided in the present application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0175] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0176] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0177] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including multiple instructions to enable an electronic device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (Read-Only Memory, referred to as ROM), random access memory (Random Access Memory, referred to as RAM), disk or optical disk and other media that can store programs.

[0178] The preferred embodiments of the present invention are described above with reference to the accompanying drawings, but the scope of the rights of the present invention is not limited thereto. Any modification, equivalent substitution and improvement made by a person skilled in the art without departing from the scope and essence of the present invention should be within the scope of the rights of the present invention.

Claims

1. A method for generating insurance scripts, characterized in that: The method comprises: Acquire the initial insurance text of the target interested insured object; wherein the target interested insured object tends to insure the target entity object; Acquire the object information of the target entity object to obtain the target entity object information; Performing text semantic correction on the initial insurance text of interest through a preset insurance text correction model to obtain a target insurance text of interest; Screening a preset insurance clause database according to the target insurance text to obtain target insurance clauses; Generate speech for the target insurance text, the target entity object information and the target insurance clause through a preset large language model to obtain an insurance speech text; The insurance speech text is audio-generated using a preset audio generation model to obtain target insurance speech audio.

2. The method according to claim 1, characterized in that The acquiring the object information of the target entity object to obtain the target entity object information includes: Acquire the category of the target entity object to obtain the entity object category; Searching a preset category and information item mapping library according to the entity object category to obtain a target information item; Information is collected according to the target information item to obtain the target entity object information.

3. The method according to claim 2, characterized in that The step of collecting information according to the target information item to obtain the target entity object information includes: Acquire a target entity object image uploaded by the target interested insured person; Searching the preset information items and the model mapping library according to the target information items to obtain a model search result; wherein the model search result indicates whether the target information extraction model is found or not; If the model search result indicates that the target information extraction model is found, the target entity object image is identified by using the found target information extraction model to obtain the target entity object information.

4. The method according to claim 3, characterized in that After searching the preset information item and the model mapping library according to the target information item to obtain the model search result, the method further includes: If the model search result indicates that the target information extraction model is not found, sending an information filling request to the target interested insured person; Receiving information filled in by the target interested insured person based on the information filling request, and obtaining target filled-in information; The target entity object information is updated according to the target filled information.

5. The method according to any one of claims 1 to 4, characterized in that: The method of generating speech texts for the target insurance text, the target entity object information and the target insurance clauses by using a preset large language model to obtain insurance speech texts includes: Parsing the target insurance clauses through the large language model to obtain insurance compensation information and insurance restriction information; Performing insurance detection on the insurance restriction information and the target entity object information by using the large language model to obtain an insurance detection result; If the insurance detection result indicates that the target entity object can be insured, the large language model is used to generate speech for the target insurance text, the target entity object information and the insurance claim information to obtain an insurance speech text; The insurance concern speech text is determined as the insurance speech text.

6. The method according to claim 5, characterized in that After performing insurance detection on the insurance restriction information and the target entity object information by using the large language model to obtain the insurance detection result, generating speech texts for the target insurance text, the target entity object information and the target insurance clause by using the preset large language model to obtain the insurance speech text further includes: If the insurance detection result indicates that the target entity object cannot be insured, then generating information for the target insurance text and preset reference recommendation information through the large language model to obtain insurance recommendation information; Generate speech for the insurance recommendation information and the target entity object information by using the large language model to obtain insurance recommendation speech text; The insurance recommendation speech text is determined as the insurance speech text.

7. The method according to any one of claims 1 to 4, characterized in that: The step of obtaining the initial insurance text of the target interested insured person includes: Acquire the voice data input by the target interested insured person from a preset consultation channel to obtain the insurance consultation voice data of the target; Performing text conversion on the insurance consultation voice data of the subject through a preset insurance voice-to-text model to obtain a first insurance consultation text of the subject; Acquire the text data input by the target interested insured person from the consultation channel to obtain the second insurance consultation text of the target; The first insurance consultation text of the object and the second insurance consultation text of the object are integrated to obtain the initial insurance text of interest.

8. An insurance script generation device, characterized in that: The device comprises: A text acquisition module is used to acquire the initial insurance text of the target interested insured object; wherein the target interested insured object tends to insure the target entity object; An information acquisition module, used to acquire the object information of the target entity object, and obtain the target entity object information; A text correction module, used for performing text semantic correction on the initial insurance text of interest through a preset insurance text correction model to obtain a target insurance text of interest; A clause screening module, used to screen the preset insurance clause database according to the target insurance text to obtain the target insurance clause; A text generation module, used to generate speech for the target insurance text, the target entity object information and the target insurance clause through a preset large language model to obtain an insurance speech text; The audio generation module is used to generate audio for the insurance speech text through a preset audio generation model to obtain the target insurance speech audio.

9. An electronic device, characterized in that: The electronic device comprises a memory and a processor, the memory stores a computer program, and the processor implements the method according to any one of claims 1 to 7 when executing the computer program.

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

Images