Insurance policy data annotation method and device, electronic equipment and storage medium

Abstract

The embodiment of the invention provides an insurance policy data annotation method and device, electronic equipment and a storage medium, and belongs to the technical field of artificial intelligence. The method comprises the steps that an insurance policy resource pool is acquired, the insurance policy resource pool comprises an original insurance policy, and the original insurance policy has an insurance policy deadline attribute and a contact object attribute; performing insurance policy screening on the original insurance policies based on the insurance policy deadline attribute to obtain an immediate insurance policy; determining an original contact object based on the contact object attribute of the on-time insurance policy, and obtaining initial reminding call data of the original contact object; performing voice recognition on the initial reminding call data to obtain an original call text; performing intention recognition on the original call text through a target renewal intention recognition model to obtain a renewal intention category; performing verbal skill generation based on the renewing intention category and the original conversation text to obtain a target verbal skill; and based on the renewing intention category and the target verbal skill, carrying out data labeling on the temporary insurance policy. According to the embodiment of the invention, the accuracy of policy data annotation can be improved.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence technology and is applicable to financial technology scenarios, and in particular to a method and device for annotating insurance policy data, an electronic device, and a storage medium. Background Art

[0002] Data annotation refers to the process of adding additional explanatory or tagging information to specific data items, datasets, or data features during data processing, analysis, or labeling. Data annotation can be applied to a variety of scenarios. For example, in the fintech insurance renewal management scenario, data annotation can be applied to policies after issuing renewal reminders for expiring policies.

[0003] Currently, insurance agents primarily provide renewal reminders to customers with expiring policies through proactive outbound calls. After the call, the agent manually annotates the customer's renewal intention based on the conversation content, providing a basis for subsequent renewal reminder strategy development. However, in actual application, different insurance agents may have subjective biases in their understanding of customer renewal intentions, affecting the accuracy of the annotated data.

[0004] Therefore, how to improve the accuracy of insurance policy data annotations has become a technical problem that needs to be solved urgently. Summary of the Invention

[0005] The main purpose of the embodiments of the present application is to propose a method and device for annotating insurance policy data, an electronic device, and a storage medium, aiming to improve the accuracy of insurance policy data annotation.

[0006] To achieve the above-mentioned purpose, a first aspect of an embodiment of the present application provides a method for annotating insurance policy data, the method comprising:

[0007] Obtaining a policy resource pool; wherein the policy resource pool includes original policies, and the original policies have policy term attributes and contact object attributes;

[0008] Screening the original policies based on the policy term attribute to obtain near-expiry policies;

[0009] Determining an original contact object based on the contact object attributes of the near-expiring insurance policy, and obtaining initial reminder call data of the original contact object;

[0010] Performing voice recognition on the initial reminder call data to obtain an original call text;

[0011] Performing intent recognition on the original call text using a preset target renewal intention recognition model to obtain a renewal intention category of the original contact object;

[0012] Generate a speech based on the renewal intention category and the original call text to obtain a target speech;

[0013] Data annotation is performed on the expiring insurance policy based on the renewal intention category and the target speech.

[0014] In some embodiments, generating a call script based on the renewal intention category and the original call text to obtain a target call script includes:

[0015] Performing entity extraction on the original call text to obtain call semantic features;

[0016] Obtain reference scripts based on the renewal intention category;

[0017] The call semantic features and the reference speech are integrated to obtain the target speech.

[0018] In some embodiments, extracting entities from the original call text to obtain call semantic features includes:

[0019] Performing policy entity recognition on the original call text to obtain policy entity features;

[0020] Performing relational entity recognition on the original call text to obtain policy relationship features;

[0021] Extracting emotional features from the original call text to obtain emotional features of the call;

[0022] Feature fusion is performed based on the policy entity feature, the policy relationship feature and the call emotion feature to obtain the call semantic feature.

[0023] In some embodiments, performing intent recognition on the original call text using a preset target renewal intention recognition model to obtain the renewal intention category of the original contact object includes:

[0024] Performing text preprocessing on the original call text to obtain a target call text;

[0025] Performing intent classification on the target call text using the target renewal intention recognition model to obtain an original intent category prediction value;

[0026] The renewal intention category of the original contact object is determined based on the original intention category prediction value.

[0027] In some embodiments, performing speech recognition on the initial reminder call data to obtain the original call text includes:

[0028] Performing speaker recognition on the initial reminder call data to obtain speaker voice annotation information;

[0029] Performing voice extraction on the initial reminder call data based on the speaker voice annotation information to obtain the call voice data of the original contact object;

[0030] The call voice data is converted into text to obtain the original call text.

[0031] In some embodiments, screening the original policies based on the policy term attribute to obtain expiring policies includes:

[0032] Calculate the duration of the policy based on the current time and the policy duration attribute to obtain the duration of the policy expiration;

[0033] Sorting the original policies based on the policy terms to obtain an initial policy sequence;

[0034] The initial policy sequence is screened based on a preset near-expiry duration threshold to obtain the near-expiry policy; wherein the near-expiry duration of the temporary policy is less than or equal to the near-expiry duration threshold.

[0035] In some embodiments, before performing intent recognition on the original call text using a preset target renewal intention recognition model to obtain the renewal intention category of the original contact object, the method further includes:

[0036] Acquire call sample data; wherein the call sample data includes sample call data and a sample renewal intention tag;

[0037] Performing speech recognition on the sample call data to obtain a sample call text;

[0038] Performing intent recognition on the sample call text using a preset original renewal intention recognition model to obtain a sample renewal intention category;

[0039] The parameters of the original renewal intention recognition model are adjusted based on the sample renewal intention label and the sample renewal intention category to obtain the target renewal intention recognition model.

[0040] To achieve the above-mentioned purpose, a second aspect of an embodiment of the present application provides a device for annotating insurance policy data, the device comprising:

[0041] An insurance policy resource pool acquisition module is used to acquire an insurance policy resource pool; wherein the insurance policy resource pool includes original insurance policies, and the original insurance policies have an insurance policy term attribute and a contact object attribute;

[0042] a near-expiring policy screening module, configured to screen the original policies based on the policy term attribute to obtain near-expiring policies;

[0043] A contact object acquisition module, configured to determine an original contact object based on the contact object attributes of the expiring policy, and acquire initial reminder call data of the original contact object;

[0044] A speech recognition module, configured to perform speech recognition on the initial reminder call data to obtain an original call text;

[0045] An intention recognition module, configured to perform intention recognition on the original call text using a preset target renewal intention recognition model to obtain a renewal intention category of the original contact object;

[0046] A speech generation module, configured to generate speech based on the renewal intention category and the original call text to obtain a target speech;

[0047] The target data annotation module is used to annotate the data of the expiring insurance policy based on the renewal intention category and the target speech.

[0048] 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 the processor implements the method described in the first aspect when executing the computer program.

[0049] To achieve the above-mentioned purpose, the fourth aspect of the embodiments of the present application proposes a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, it implements the method described in the first aspect.

[0050] The policy data annotation method and device, electronic device, and storage medium proposed in this application screen all original policies in a policy resource pool based on their policy term attributes to identify near-expiring policies. This helps focus on customer groups with expiring policies and improves renewal reminder efficiency. Next, the original contact person is determined based on the contact person attributes of the near-expiring policy. The initial reminder call data for the initial renewal reminder with the original contact person is obtained. Voice recognition is performed on the initial reminder call data to obtain the original call text, facilitating subsequent intent recognition. Furthermore, intent recognition is performed on the original call text using a preset target renewal intent recognition model, accurately identifying the renewal intent category of the original contact person. Targeted language is generated based on the renewal intent category and the original call text, enabling tailored language tailored to the original contact person's intent and communication content. Finally, data for near-expiring policies is annotated based on the renewal intent category and target language, improving the accuracy of policy data annotation. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1This is a flow chart of the insurance policy data annotation method provided in an embodiment of the present application;

[0052] Figure 2 yes Figure 1 Flowchart of step S102 in FIG.

[0053] Figure 3 yes Figure 1 Flowchart of step S104 in FIG.

[0054] Figure 4 This is a flow chart of a method for annotating insurance policy data provided by another embodiment of the present application;

[0055] Figure 5 yes Figure 1 Flowchart of step S105 in FIG.

[0056] Figure 6 yes Figure 1 Flowchart of step S106 in FIG.

[0057] Figure 7 yes Figure 6 Flowchart of step S601 in FIG.

[0058] Figure 8 This is a schematic diagram of the structure of the insurance policy data annotation device provided in an embodiment of the present application;

[0059] Figure 9 This is a schematic diagram of the hardware structure of the electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0060] In order to make the purpose, technical solutions and advantages of this application more clear, the following further describes this application in detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain this application and are not intended to limit this application.

[0061] It should be noted that although the device schematics illustrate functional module divisions and the flowcharts illustrate logical sequences, in certain circumstances, the steps shown or described may be performed in a sequence that differs from the module divisions in the device or the sequence in the flowcharts. The terms "first," "second," and so on, in the specification, claims, and drawings, are used to distinguish similar items and are not necessarily used to describe a specific sequence or precedence.

[0062] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the art to which this application pertains. The terms used herein are for the purpose of describing the embodiments of this application only and are not intended to limit this application.

[0063] First, let’s analyze some of the terms used in this application:

[0064] Artificial intelligence (AI) is a new technical discipline that studies and develops theories, methods, technologies, and application systems for simulating, extending, and expanding human intelligence. A branch of computer science, AI seeks to understand the essence of intelligence and produce new intelligent machines that can respond in a manner similar to human intelligence. Research in this field includes robotics, speech recognition, image recognition, natural language processing, and expert systems. AI can simulate the information processes of human consciousness and thinking. It also encompasses the theories, methods, technologies, and application systems that use digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, to perceive the environment, acquire knowledge, and use that knowledge to achieve optimal results.

[0065] Natural language processing (NLP): NLP uses computers to process, understand, and apply human languages ​​(such as Chinese and English). A branch of artificial intelligence, NLP is an interdisciplinary field between computer science and linguistics, often referred to as computational linguistics. Natural language processing encompasses grammatical analysis, semantic analysis, and discourse comprehension. Natural language processing is commonly used in technical fields such as machine translation, handwritten and printed character recognition, speech recognition and text-to-speech conversion, information intent recognition, information extraction and filtering, text classification and clustering, public opinion analysis, and opinion mining. It encompasses data mining, machine learning, knowledge acquisition, knowledge engineering, artificial intelligence research related to language processing, and linguistics research related to language computing.

[0066] Information Extraction: A text processing technology that extracts specified types of entity, relationship, event, and other factual information from natural language text and forms structured data output. Information extraction is a technology that extracts specific information from text data. Text data is composed of some specific units, such as sentences, paragraphs, and chapters. Text information is composed of some small specific units, such as characters, words, phrases, sentences, paragraphs, or a combination of these specific units. Extracting noun phrases, names, place names, etc. from text data is all text information extraction. Of course, the information extracted by text information extraction technology can be of various types.

[0067] Data annotation is the process of adding additional explanatory or tagging information to specific data items, datasets, or features during data processing, analysis, or labeling. Data annotation can be applied to a variety of scenarios. For example, in finance, data annotation can be applied to insurance policies after issuing renewal reminders for expiring policies.

[0068] Currently, insurance renewal management primarily involves proactive outbound calls from insurance agents to remind customers with expiring policies to renew their policies. After the call, the agent manually annotates the customer's renewal intention based on the conversation, providing a basis for subsequent renewal reminder strategy development. However, in practice, subjective interpretations of customer renewal intentions can vary among different insurance agents, impacting the accuracy of the annotated data.

[0069] Based on this, the embodiments of the present application provide a method and device for annotating insurance policy data, an electronic device, and a storage medium, aiming to improve the accuracy of insurance policy data annotation.

[0070] The insurance policy data annotation method and device, electronic device and storage medium provided in the embodiments of the present application are specifically explained through the following embodiments. First, the insurance policy data annotation method in the embodiments of the present application is described.

[0071] The embodiments of the present application can acquire and process relevant data based on artificial intelligence technology. Artificial Intelligence (AI) is the theory, method, technology, and application system that uses digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use knowledge to achieve optimal results.

[0072] Fundamental AI technologies generally include sensors, dedicated AI chips, cloud computing, distributed storage, big data processing, operating / interaction systems, and mechatronics. AI software technologies primarily encompass computer vision, robotics, biometrics, speech processing, natural language processing, and machine learning / deep learning.

[0073] The insurance policy data annotation method provided in the embodiment of the present application relates to the field of artificial intelligence technology. The insurance policy data annotation method provided in the embodiment of the present application can be applied to the terminal, can also be applied to the server side, and can also be software running in the terminal or the server side. In some embodiments, the terminal can be a smart phone, a tablet computer, a laptop computer, a desktop computer, etc.; the server side can be configured as an independent physical server, or can be configured as a server cluster or distributed system composed of multiple physical servers, and can also be configured as 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 policy data annotation method, etc., but is not limited to the above forms.

[0074] The present application can be used in many general or special computer system environments or configurations. For example: personal computers, server computers, handheld or portable devices, tablet devices, multiprocessor systems, microprocessor-based systems, set-top boxes, programmable consumer electronics, network PCs, minicomputers, mainframe computers, distributed computing environments including any of the above systems or devices, and the like. 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, and the like 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 via a communication network. In a distributed computing environment, program modules can be located in local and remote computer storage media, including storage devices.

[0075] It should be noted that in each specific embodiment of the present application, when it comes to the need to perform relevant processing based on data related to the user's identity or characteristics, such as user information, user behavior data, user historical data, and user location information, 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. In addition, when the embodiment of the present application needs to obtain the user's sensitive personal information, the user's separate permission or consent will be obtained through a pop-up window or by jumping to a confirmation page. After clearly obtaining the user's separate permission or consent, the necessary user-related data for the normal operation of the embodiment of the present application will be obtained.

[0076] Figure 1 This is an optional flow chart of the insurance policy data annotation method provided in the embodiment of the present application. Figure 1 The method may include but is not limited to steps S101 to S107.

[0077] Step S101: Obtain a policy resource pool; wherein the policy resource pool includes original policies, and the original policies have policy term attributes and contact object attributes;

[0078] Step S102: Screening the original policies based on the policy term attribute to obtain expiring policies;

[0079] Step S103: determining the original contact object based on the contact object attributes of the expiring policy, and obtaining the initial reminder call data of the original contact object;

[0080] Step S104, performing voice recognition on the initial reminder call data to obtain the original call text;

[0081] Step S105: performing intent recognition on the original call text using a preset target renewal intention recognition model to obtain a renewal intention category of the original contact object;

[0082] Step S106: Generate a target call based on the renewal intention category and the original call text.

[0083] Step S107: Data labeling of expiring policies based on renewal intention categories and target sales pitches.

[0084] In the steps S101 to S107 shown in the embodiment of the present application, all original policies in the policy resource pool are screened based on the policy term attributes of the original policies to obtain expiring policies, which facilitates focusing on the customer group that is about to expire and improves the efficiency of renewal reminders. Next, the original contact object is determined based on the contact object attributes of the expiring policy, the initial reminder call data for the first renewal reminder with the original contact object is obtained, and the initial reminder call data is voice recognized to obtain the original call text, which facilitates subsequent intent recognition. Furthermore, the original call text is subjected to intent recognition through a preset target renewal intention recognition model, which can accurately identify the renewal intention category of the original contact object. Based on the renewal intention category and the original call text, a speech is generated to obtain the target speech, which can formulate targeted speech according to the intention and communication content of the original contact object. Finally, the data of the expiring policy is annotated based on the renewal intention category and the target speech, which improves the accuracy of the policy data annotation.

[0085] In step S101 of some embodiments, the policy resource pool is a pre-built collection of multiple original policies. The policy resource pool can be stored in a database, data warehouse or other system to provide data support for different businesses in insurance application scenarios. For example, policies that are about to expire can be screened out from the policy resource pool, and high-risk policies can be screened out from the policy resource pool.

[0086] Specifically, an original policy is an insured policy that has been purchased and is currently insured. Each original policy has at least a policy term attribute and a contact person attribute. The policy term attribute indicates the policy protection period of the original policy, including the effective date and expiration date. The contact person attribute indicates the policyholder or designated contact person of the original policy, who can be contacted to communicate about policy-related matters.

[0087] It should be noted that the original insurance policy may also have information related to the insurance policy, such as the insurance policy type attribute and the insurance policy amount attribute. The specific selection needs to be made based on the actual application scenario, but is not limited to this.

[0088] See also Figure 2In some embodiments, step S102 may include but is not limited to steps S201 to S203:

[0089] Step S201, calculate the duration of the policy based on the current time and the policy duration attribute to obtain the duration of the policy expiration;

[0090] Step S202: sorting the original policies based on the policy term length to obtain an initial policy sequence;

[0091] Step S203, screening the initial policy sequence based on a preset near-expiry time threshold to obtain near-expiry policies; wherein the near-expiry time of the temporary policy is less than or equal to the near-expiry time threshold.

[0092] Steps S201 to S203 shown in the embodiment of the present application calculate the policy expiration time based on the current moment and the policy term attribute, and obtain the policy expiration time, which can accurately quantify the remaining time of each original policy until expiration. Then, the original policies are sorted based on the policy term to obtain an initial policy sequence, so that the original policies are clearly arranged in order of expiration, which is convenient for subsequent screening of the original policies. Finally, the initial policy sequence is screened based on the preset expiration time threshold, and the policies whose expiration time is less than or equal to the expiration time threshold are screened out to obtain the expiring policies, thereby facilitating targeted renewal reminders for the policies that are about to expire.

[0093] In step S201 of some embodiments, for each original policy, the policy expiration time of the original policy is calculated by calculating the expiration time in the policy term attribute of the original policy with the current time. The policy expiration time is generally measured in days. For example, the policy expiration time of original policy A is 52 days.

[0094] In step S202 of some embodiments, the multiple original policies are sorted in ascending order based on the policy term lengths to obtain an initial policy sequence. Alternatively, the multiple original policies may be sorted in descending order based on the policy term lengths, and the specific selection depends on the actual application scenario and is not limited thereto.

[0095] It should be noted that the preset threshold of the duration of the imminent expiration is set according to the actual application scenario. For example, the threshold of the duration of the imminent expiration is set to 30 days, 60 days or 90 days, but is not limited thereto.

[0096] In step S203 of some embodiments, the policies in the initial policy sequence whose expiring time is less than or equal to the expiring time threshold are screened out to obtain the expiring policies.

[0097] It is understandable that since the initial policy sequence is a sorted sequence, it is only necessary to find the corresponding policy as the boundary based on the temporary duration threshold. If the policy is arranged in order from short to long, the policies at and above the boundary will be filtered out to obtain the near-expiring policies; if the policy is arranged in order from long to short, the policies at and below the boundary will be filtered out to obtain the near-expiring policies.

[0098] It should be noted that the expiring insurance policies are screened from the original insurance policies, and the expiring insurance policies also have the attribute of the contact object.

[0099] In step S103 of some embodiments, the original contact object is determined based on the contact object attributes of the expiring insurance policy, and the initial reminder call data of the first renewal reminder between the manual customer service or intelligent customer service and the original contact object is obtained.

[0100] It is understandable that the renewal intention of the original contact person can only be known after the initial renewal reminder, so that targeted sales talk can be formulated based on the renewal intention of the original contact person, and the sales talk and renewal intention can be annotated together, so that when the policy renewal reminder is made again in the future, the efficiency of the renewal reminder and the success rate of the policy renewal can be improved.

[0101] See also Figure 3 In some embodiments, step S104 may include but is not limited to steps S301 to S303:

[0102] Step S301, performing speaker recognition on the initial reminder call data to obtain speaker voice annotation information;

[0103] Step S302: performing voice extraction on the initial reminder call data based on the speaker's voice annotation information to obtain the call voice data of the original contact object;

[0104] Step S303: convert the call voice data into text to obtain the original call text.

[0105] In steps S301 to S303 shown in the embodiment of the present application, by performing speaker recognition on the initial reminder call data, the speaker voice annotation information is obtained, which can accurately distinguish the voice segments of different speakers in the call. Then, based on the speaker voice annotation information, the initial reminder call data is subjected to voice extraction, and the call voice data of the original contact object can be accurately obtained, eliminating the interference of other speakers, such as manual customer service or intelligent customer service, to ensure that the extracted voice data is accurate and focused. Finally, the extracted call voice data is converted into text to obtain the original call text, so that the call content is presented in text form, which helps to better understand the call content and facilitates subsequent intent recognition and speech generation to improve the accuracy of insurance policy data annotation.

[0106] In step S301 of some embodiments, a pre-trained speaker recognition model may be used to perform speaker recognition on the initial reminder call data to obtain speaker voice annotation information. The speaker voice annotation information is used to indicate speech segments of different speakers. For example, the speech segment corresponding to the timeline 00:01:20-00:01:45 is the speech segment of the original contact, and the speech segment corresponding to the timeline 00:01:47-00:02:15 is the speech segment of the human customer service representative.

[0107] Specifically, the speaker recognition model can be an I-Vector model, an X-Vector model, a convolutional neural network model, a Transformer model, etc., and the specific selection needs to be based on the actual application scenario, but is not limited to these.

[0108] In step S302 of some embodiments, voice data corresponding to the original contact object is extracted from the initial reminder call data according to the speaker's voice annotation information to obtain multiple voice segments to constitute the call voice data.

[0109] In step S303 of some embodiments, a pre-trained text conversion model can be used to convert the call voice data into text to obtain the original call text, where the original call text is an unedited transcription result that retains all information in the voice data, including interjections, pauses, incorrectly recognized content, etc.

[0110] Specifically, the text conversion model can be a Wav2Vec model, a convolutional neural network model, a long short-term memory network model, a Transformer-based audio feature extraction model, etc. The specific selection needs to be based on the actual application scenario, but is not limited to this.

[0111] See also Figure 4 In some embodiments, before step S105, the policy data annotation method may include but is not limited to steps S401 to S404:

[0112] Step S401: Acquire call sample data; wherein the call sample data includes sample call data and a sample renewal intention tag;

[0113] Step S402: performing speech recognition on the sample call data to obtain a sample call text;

[0114] Step S403: performing intent recognition on the sample call text using a preset original renewal intention recognition model to obtain a sample renewal intention category;

[0115] Step S404 : adjusting parameters of the original renewal intention recognition model based on the sample renewal intention label and the sample renewal intention category to obtain a target renewal intention recognition model.

[0116] Steps S401 to S404 shown in the embodiment of the present application provide a reliable data basis for model training by obtaining call sample data containing sample call data and sample renewal intention labels, ensuring that the training data is closely related to the target task. Next, the sample call data is converted into sample call text through voice recognition, and the unstructured voice information is converted into structured text information to facilitate subsequent model processing. Then, the original renewal intention recognition model is used to perform intent recognition on the sample call text to obtain the sample renewal intention category, and the intent recognition function is preliminarily realized. Finally, the model parameters are adjusted based on the sample renewal intention label and the identified sample renewal intention category. Through this supervised learning method, the model continuously learns the difference between the real intention and the predicted intention, and gradually optimizes the model parameters, thereby obtaining a more accurate target renewal intention recognition model, significantly improving the accuracy of renewal intention recognition, and providing strong support for the company's precision marketing, customer service and other businesses.

[0117] In step S401 of some embodiments, the call sample data is data collected in advance in the policy renewal management scenario, including sample call data of the initial renewal reminder between manual customer service or intelligent customer service and the sample contact object, and sample renewal intention labels used to characterize the renewal intention category of the sample contact object; wherein, the sample renewal intention labels are pre-labeled by insurance business personnel, and include three types, namely: willing to renew, unwilling to renew, and unclear renewal intention.

[0118] In some embodiments, when collecting call sample data, cross-labeling verification is also performed by different insurance business personnel to improve the accuracy of the sample renewal intention label, thereby improving the accuracy of the target renewal intention recognition model in intent recognition.

[0119] In some embodiments, step S402 may include but is not limited to the following steps:

[0120] By performing speaker recognition on the sample call data, sample speech annotation information is obtained;

[0121] Perform voice extraction on the sample call data based on the sample voice annotation information to obtain sample voice data of the sample contact object;

[0122] The sample voice data is converted into text to obtain a sample call text.

[0123] Specifically, the implementation of the above step S402 is basically the same as the specific implementation of the above steps S301 to S303, and will not be repeated here.

[0124] In step S403 of some embodiments, the original renewal intention recognition model is used to perform intent recognition on the sample call text to obtain a sample renewal intention category; wherein the sample renewal intention category is obtained through model prediction and includes three categories, namely, willingness to renew, unwillingness to renew, and unclear renewal intention.

[0125] Among them, the original renewal intention recognition model can adopt a large language model, a BERT model, a random forest model, etc. The specific selection needs to be based on the actual application scenario, but is not limited to this.

[0126] In some embodiments, step S404 may include but is not limited to the following steps:

[0127] Calculate the recognition loss based on the sample renewal intention label and sample renewal intention category to obtain the intent recognition loss function;

[0128] Based on the intention recognition loss function, the parameters of the original renewal intention recognition model are adjusted to obtain the target renewal intention recognition model.

[0129] Specifically, a corresponding loss function is selected according to the original renewal intention recognition model, and the loss is calculated for the sample renewal intention label and the sample renewal intention category to obtain the intention recognition loss function; among them, the loss function can select the mean square error loss function, the cross entropy loss function, etc., which is not limited in the embodiment of the present application.

[0130] Next, the backpropagation method, gradient descent method, momentum update method and other methods can be used to adjust the parameters of the original renewal intention recognition model based on the intention recognition loss function, so as to optimize the learning ability and generalization ability of the model and obtain the target renewal intention recognition model.

[0131] In some embodiments, if the original renewal intention recognition model is a large language model, Lora technology can be used to fine-tune the parameters of the original renewal intention recognition model, such as adjusting the learning rate, batch size, etc.

[0132] The target renewal intention recognition model is then deployed on the local device of the policy renewal management scenario for subsequent intent recognition, which can ensure the security and privacy of the data.

[0133] See also Figure 5 In some embodiments, step S105 may also include but is not limited to steps S501 to S503:

[0134] Step S501, performing text preprocessing on the original call text to obtain the target call text;

[0135] Step S502: classify the target call text using the target renewal intention recognition model to obtain a predicted value of the original intention category;

[0136] Step S503: determining the renewal intention category of the original contact object based on the original intention category prediction value.

[0137] In steps S501 to S503 of the embodiment of the present application, by performing text preprocessing on the original call text to obtain the target call text, the original call text can be standardized, thereby improving the accuracy of call information extraction and facilitating subsequent intent classification. Next, the target call text is classified by intent using the target renewal intention recognition model to obtain the original intent category prediction value, and then category generation is performed based on the original intent category prediction value, accurately identifying the renewal intent category of the original contact object, which helps to determine the subsequent reminder strategy based on the intent category, improves the accuracy of the insurance policy data annotation, and facilitates the efficiency of renewal reminders when the insurance policy renewal reminder is issued again in the future.

[0138] In step S501 of some embodiments, text preprocessing includes but is not limited to: removing punctuation marks, special symbols, stop words, word segmentation, correcting common typos, unifying synonyms, etc., thereby normalizing the original call text and improving the accuracy of call information extraction.

[0139] Next, the target call text is classified by the trained target renewal intention recognition model to obtain the original intent category prediction value, which is used to indicate the prediction value corresponding to different categories.

[0140] For example, if there are three renewal intention categories, the original intention category prediction value can be an array [x,x,x] containing three elements, where each element corresponds to a category. The first element corresponds to the category "willing to renew," the second element corresponds to the category "unwilling to renew," and the third element corresponds to the category "unclear renewal intention." A value of 1 indicates that the intention belongs to that category, and a value of 0 indicates that the intention does not belong to that category.

[0141] Therefore, the renewal intention category of the original contact object can be determined based on the original intention category prediction value. For example, the original intention category prediction value of original contact object A is [1,0,0], indicating that the renewal intention category of original type object A is willing to renew; the original intention category prediction value of original contact object B is [0,1,0], indicating that the renewal intention category of original type object B is not willing to renew; and the original intention category prediction value of original contact object C is [0,0,1], indicating that the renewal intention category of original type object C is unclear.

[0142] It should be noted that, among the elements in the original intention category prediction value, only one element has a value of 1. Furthermore, the renewal intention category includes three types: willing to renew, unwilling to renew, and unclear renewal intention.

[0143] See also Figure 6 In some embodiments, step S106 includes but is not limited to steps S601 to S603:

[0144] Step S601: extracting entities from the original call text to obtain call semantic features;

[0145] Step S602: Obtain reference scripts based on the renewal intention category;

[0146] Step S603: integrating the call semantic features and the reference speech to obtain the target speech.

[0147] In the steps S601 to S603 shown in the embodiment of the present application, the unstructured original call text is converted into structured call semantic features by performing entity extraction on the original call text. Then, the predefined reference speech template is accurately matched according to the renewal intention category to ensure that the service strategy matches the intention category, thereby avoiding the mechanical response of the traditional fixed speech. Finally, the speech is integrated with the call semantic features and the reference speech to obtain the target speech, which not only retains the standardization of professional speech, but also enables the customization of speech, which is more in line with the needs of the original contact object, thereby helping to improve the accuracy of the policy data annotation, and facilitating the subsequent renewal reminder of the policy, which can improve the efficiency of the renewal reminder and the success rate of the policy renewal.

[0148] See also Figure 7 In some embodiments, step S601 may include but is not limited to steps S701 to S704:

[0149] Step S701: Perform policy entity recognition on the original call text to obtain policy entity features;

[0150] Step S702: performing relationship entity recognition on the original call text to obtain insurance policy relationship features;

[0151] Step S703: extracting emotional features from the original call text to obtain call emotional features;

[0152] Step S704: performing feature fusion based on the policy entity features, policy relationship features, and call sentiment features to obtain call semantic features.

[0153] In the steps S701 to S704 shown in the embodiment of the present application, by performing policy entity recognition on the original call text, key information in the insurance business is extracted from the original call text to obtain policy entity features; by performing relationship entity recognition on the original call text, policy relationship features between policy entity features are obtained; by performing sentiment feature extraction on the original call text, call sentiment features are obtained, which can accurately capture the implicit attitude of the original contact object towards policy renewal and provide a psychological decision-making basis for the renewal reminder strategy. Finally, based on the policy entity features, policy relationship features and call sentiment features, feature fusion is performed to obtain call semantic features, which can more accurately guide the generation of speech, improve the matching degree between the policy renewal strategy and the needs of the original contact object, thereby helping to improve the accuracy of policy data annotations, and facilitate subsequent policy renewal reminders, which can improve the efficiency of renewal reminders and the success rate of policy renewals.

[0154] In step S701 of some embodiments, the original call text may contain unstructured information such as the policy number, product name, policyholder, beneficiary, and expiration date. A named entity recognition model can be used to annotate and extract policy-related entities from the original call text, extracting the unstructured information and obtaining policy entity features, which facilitates rapid location of policy information related to the original contact. The named entity recognition model can be a BiLSTM-CRF model, a BERT-NER model, or other such model, and the specific selection should be based on the actual application scenario, but is not limited thereto.

[0155] In step S702 of some embodiments, the relationship between entities, such as the relationship between the policyholder and the beneficiary, and the relationship between the policy number and the expiration date, can be identified through dependency syntax analysis or relationship extraction models, so as to accurately understand the original contact object's demands for the policy and improve the accuracy of information understanding.

[0156] In step S703 of some embodiments, a sentiment analysis model is used to extract sentiment features from the original call text, thereby accurately understanding the original contact's emotional attitude towards the insurance policy and obtaining sentiment features. Communication strategies can be adjusted based on sentiment, for example, avoiding sales pitches for an "angry" contact.

[0157] In step S704 of some embodiments, the policy entity features, policy relationship features, and call emotion features are integrated into multi-dimensional call semantic features, which can support intelligent decision-making in policy renewal management scenarios.

[0158] It should be noted that different renewal intention categories are provided with corresponding reference scripts, which are stored in a preset script database.

[0159] Understandably, when the renewal intention category is unwilling to renew or unclear, sales agents often face communication difficulties and struggle to effectively guide the original contact to complete policy renewal. Therefore, it is even more necessary to refer to sales pitches for guidance to improve the success rate of policy renewal.

[0160] In step S602 of some embodiments, corresponding reference speech is obtained from a speech database based on the renewal intention category.

[0161] Furthermore, the preset speech generation model can be used to integrate the call semantic features and reference speech to obtain the target speech, which can better meet the needs of the original contact object, thereby helping to improve the efficiency of policy renewal reminders and increase the success rate of policy renewal.

[0162] In some embodiments, the speech generation model can adopt a large language model, a BERT model, a RAG model, etc., and the specific selection needs to be made according to the actual application scenario, but is not limited to this.

[0163] In step S107 of some embodiments, by combining the renewal intention category and the target speech, target annotation data is obtained, and data annotation is performed on the expiring insurance policy based on the target annotation data. Specifically,

[0164] It should be noted that the target annotation data can be stored in the policy resource pool, so that when the expiring policies are screened out from the policy resource pool next time for policy renewal reminders, the content of the target annotation data can be combined to achieve targeted policy renewal reminders, thereby improving the efficiency of renewal reminders and the success rate of policy renewal.

[0165] In some embodiments, after step S107, when a policy renewal reminder is next needed, expiring policies are screened from the resource pool, and target annotation data corresponding to the expiring policies is obtained, thereby determining the target sales pitch. After determining the target sales pitch, human or intelligent customer service representatives can send a policy renewal reminder to the original contact based on the target sales pitch.

[0166] It is understandable that the target script is customized based on the original contact's intention at the time of the initial renewal reminder and the content of the call, which can help improve the efficiency of policy renewal reminders and increase the success rate of policy renewal.

[0167] In some embodiments, the policy data annotation method provided in the embodiments of the present application can be applied to the policy renewal management scenario in the field of financial technology. It can automatically annotate the initial reminder call data of the first renewal reminder to the original contact object of the expiring policy, obtain the renewal intention type corresponding to the original contact object, and then formulate corresponding words according to the renewal intention type and the call content, thereby improving the accuracy of the policy data annotation and facilitating the subsequent renewal reminder of the policy, which can improve the efficiency of the renewal reminder.

[0168] In addition, the ideas adopted in the insurance policy data annotation method provided in this application can also be extended to segmented scenarios such as renewal of financial products and financial services, transaction risk assessment, and intelligent customer service outbound calls. The model needs to be trained specifically for different segmented scenarios.

[0169] See also Figure 8 The present application also provides a device for annotating insurance policy data, which can implement the above-mentioned method for annotating insurance policy data. The device includes:

[0170] The policy resource pool acquisition module 801 is used to acquire the policy resource pool; wherein the policy resource pool includes the original policy, and the original policy has the policy term attribute and the contact object attribute;

[0171] The near-expiring policy screening module 802 is used to screen the original policies based on the policy term attribute to obtain the near-expiring policies;

[0172] The contact object acquisition module 803 is used to determine the original contact object based on the contact object attributes of the expiring insurance policy and obtain the initial reminder call data of the original contact object;

[0173] The speech recognition module 804 is used to perform speech recognition on the initial reminder call data to obtain the original call text;

[0174] Intent recognition module 805, configured to perform intent recognition on the original call text using a preset target renewal intention recognition model to obtain a renewal intention category of the original contact object;

[0175] A speech generation module 806 is configured to generate a speech based on the renewal intention category and the original call text to obtain a target speech;

[0176] The target data annotation module 807 is used to annotate data of expiring insurance policies based on renewal intention categories and target words.

[0177] The specific implementation of the insurance policy data annotation device is basically the same as the specific embodiment of the above-mentioned insurance policy data annotation method, and will not be repeated here.

[0178] The present application also provides an electronic device comprising a memory and a processor, wherein the memory stores a computer program, and the processor implements the above-described insurance policy data annotation method when executing the computer program. The electronic device can be any smart terminal, including a tablet computer and an in-vehicle computer.

[0179] See also Figure 9 , Figure 9 The hardware structure of an electronic device according to another embodiment is shown. The electronic device includes:

[0180] The processor 901 can be implemented as 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;

[0181] 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 solutions provided in the embodiments of this specification are implemented through software or firmware, the relevant program code is stored in the memory 902 and is called by the processor 901 to execute the policy data annotation method of the embodiments of this application.

[0182] Input / output interface 903, used to implement information input and output;

[0183] Communication interface 904, used to implement communication interaction between this device and other devices, which can be achieved through wired means (such as USB, network cable, etc.) or wireless means (such as mobile network, WiFi, Bluetooth, etc.);

[0184] Bus 905 , which transmits information between various components of the device (e.g., processor 901 , memory 902 , input / output interface 903 , and communication interface 904 );

[0185] 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 .

[0186] An embodiment of the present application also provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, it implements the above-mentioned insurance policy data annotation method.

[0187] 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 arranged 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.

[0188] The policy data annotation method and device, electronic device, and storage medium provided in the embodiments of the present application screen all original policies in a policy resource pool based on the policy term attributes of the original policies to obtain near-expiring policies, thereby focusing on customer groups with expiring policies and improving the efficiency of renewal reminders. Next, the original contact person is determined based on the contact person attributes of the near-expiring policy, and the initial reminder call data for the initial renewal reminder with the original contact person is obtained. Voice recognition is performed on the initial reminder call data to obtain the original call text, facilitating subsequent intent recognition. Furthermore, intent recognition is performed on the original call text using a preset target renewal intent recognition model, enabling accurate identification of the renewal intent category of the original contact person. Scripts are generated based on the renewal intent category and the original call text to obtain target scripts, enabling tailored scripts to be tailored to the original contact person's intent and communication content. Finally, data annotation is performed on the near-expiring policies based on the renewal intent category and target scripts, improving the accuracy of policy data annotation.

[0189] The embodiments described in the embodiments of this application are intended to more clearly illustrate the technical solutions of the embodiments of this application and do not constitute a limitation on the technical solutions provided by the embodiments of this 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 this application are also applicable to similar technical problems.

[0190] Those skilled in the art will understand 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.

[0191] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, i.e., they may be located in one place or distributed across multiple network units. Some or all of the modules may be selected based on actual needs to achieve the objectives of this embodiment.

[0192] Those skilled in the art will appreciate that all or some of the steps in the methods, systems, and functional modules / units in the devices disclosed above may be implemented as software, firmware, hardware, or appropriate combinations thereof.

[0193] 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 sequential order. 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 variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes 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.

[0194] It should be understood that in this 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 previous and next associated objects are in an "or" relationship. "At least one of the following items" or similar expressions refers to any combination of these items, including any combination of single items 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.

[0195] In the several embodiments provided in this 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 merely schematic. For example, the division of the above-mentioned units is only a logical function division. In actual implementation, there may be other division methods, 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.

[0196] The units described above as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of these units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0197] In addition, the functional units in the various embodiments of the present application may be integrated into a single processing unit, or each unit may exist physically separately, or two or more units may be integrated into a single unit. The aforementioned integrated units may be implemented in the form of hardware or software functional units.

[0198] 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, 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, which is stored in a storage medium and includes multiple instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of various embodiments of the present application. The aforementioned storage medium includes: various media that can store programs, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0199] The preferred embodiments of the present invention are described above with reference to the accompanying drawings, but are not intended to limit the scope of the present invention. Any modifications, equivalent substitutions, and improvements made by those skilled in the art without departing from the scope and essence of the present invention should be within the scope of the present invention.

Claims

1. A method for annotating insurance policy data, characterized in that: The method comprises: Obtaining a policy resource pool; wherein the policy resource pool includes original policies, and the original policies have policy term attributes and contact object attributes; Screening the original policies based on the policy term attribute to obtain near-expiry policies; Determining an original contact object based on the contact object attributes of the near-expiring insurance policy, and obtaining initial reminder call data of the original contact object; Performing voice recognition on the initial reminder call data to obtain an original call text; Performing intent recognition on the original call text using a preset target renewal intention recognition model to obtain a renewal intention category of the original contact object; Generate a speech based on the renewal intention category and the original call text to obtain a target speech; Data annotation is performed on the expiring insurance policy based on the renewal intention category and the target speech.

2. The method according to claim 1, characterized in that The generating of a speech based on the renewal intention category and the original call text to obtain a target speech includes: Performing entity extraction on the original call text to obtain call semantic features; Obtain reference scripts based on the renewal intention category; The call semantic features and the reference speech are integrated to obtain the target speech.

3. The method according to claim 2, characterized in that The entity extraction of the original call text to obtain the call semantic features includes: Performing policy entity recognition on the original call text to obtain policy entity features; Performing relational entity recognition on the original call text to obtain policy relationship features; Extracting emotional features from the original call text to obtain emotional features of the call; Feature fusion is performed based on the policy entity feature, the policy relationship feature and the call emotion feature to obtain the call semantic feature.

4. The method according to claim 1, wherein The performing intent recognition on the original call text by using a preset target renewal intention recognition model to obtain the renewal intention category of the original contact object includes: Performing text preprocessing on the original call text to obtain a target call text; Performing intent classification on the target call text using the target renewal intention recognition model to obtain an original intent category prediction value; The renewal intention category of the original contact object is determined based on the original intention category prediction value.

5. The method according to claim 1, wherein The performing voice recognition on the initial reminder call data to obtain the original call text includes: Performing speaker recognition on the initial reminder call data to obtain speaker voice annotation information; Performing voice extraction on the initial reminder call data based on the speaker voice annotation information to obtain the call voice data of the original contact object; The call voice data is converted into text to obtain the original call text.

6. The method according to any one of claims 1 to 5, characterized in that The screening of the original policies based on the policy term attribute to obtain near-expiring policies includes: Calculate the duration of the policy based on the current time and the policy duration attribute to obtain the duration of the policy expiration; Sorting the original policies based on the policy terms to obtain an initial policy sequence; The initial policy sequence is screened based on a preset near-expiry duration threshold to obtain the near-expiry policy; wherein the near-expiry duration of the temporary policy is less than or equal to the near-expiry duration threshold.

7. The method according to any one of claims 1 to 5, characterized in that Before performing intent recognition on the original call text using a preset target renewal intention recognition model to obtain the renewal intention category of the original contact object, the method further includes: Acquire call sample data; wherein the call sample data includes sample call data and a sample renewal intention tag; Performing speech recognition on the sample call data to obtain a sample call text; Performing intent recognition on the sample call text using a preset original renewal intention recognition model to obtain a sample renewal intention category; The parameters of the original renewal intention recognition model are adjusted based on the sample renewal intention label and the sample renewal intention category to obtain the target renewal intention recognition model.

8. A device for annotating insurance policy data, characterized in that: The device comprises: An insurance policy resource pool acquisition module is used to acquire an insurance policy resource pool; wherein the insurance policy resource pool includes original insurance policies, and the original insurance policies have an insurance policy term attribute and a contact object attribute; a near-expiring policy screening module, configured to screen the original policies based on the policy term attribute to obtain near-expiring policies; A contact object acquisition module, configured to determine an original contact object based on the contact object attributes of the expiring policy, and acquire initial reminder call data of the original contact object; A speech recognition module, configured to perform speech recognition on the initial reminder call data to obtain an original call text; An intention recognition module, configured to perform intention recognition on the original call text using a preset target renewal intention recognition model to obtain a renewal intention category of the original contact object; A speech generation module, configured to generate speech based on the renewal intention category and the original call text to obtain a target speech; The target data annotation module is used to annotate the data of the expiring insurance policy based on the renewal intention category and the target speech.

9. An electronic device, characterized in that: The electronic device includes 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