Online mediation scheme generation method and system

Abstract

The invention relates to the technical field of artificial intelligence and natural language processing, in particular to an online mediation scheme generation method and system, which are used for communicating with a customer by constructing a guide text of a plurality of dialogue rounds and guiding the customer to input identity information and state information. According to the method, the complex dialogue process is coped with by constructing the background knowledge base and calling the large language model. And if any useful information is not extracted, directly outputting the universal mediation scheme to the customer after the dialogue is finished. And if the identity information is extracted, determining a corresponding mediation scheme by using the identity information and a pre-constructed prediction model, and outputting the mediation scheme to the customer. And if the identity information and the state information are extracted, outputting a more accurate personalized mediation scheme to the customer by using the identity information, the state information and a pre-constructed prediction model. According to the method, the online mediation mode is adopted, the efficiency is high, meanwhile, the corresponding mediation scheme is formulated based on the characteristics of the customer, and the precision is high.

Description

Technical Field

[0001] The present invention relates to the technical field of artificial intelligence and natural language processing, and in particular to an online mediation solution generation method and system. Background Art

[0002] Before filing a lawsuit for some overdue loans or borrowings, banks often entrust relevant units to mediate between the creditor and the debtor.

[0003] The existing mediation generally adopts the offline mediation method, which is very dependent on the experience of the mediator and is easy to intensify the conflict between the two parties. In addition, the way of generating mediation solutions for offline mediation takes up a lot of time for both parties, and the mediation efficiency is low. Summary of the invention

[0004] In view of this, an object of the present invention is to provide a method and system for generating an online mediation solution to solve the problem of low efficiency and poor accuracy of offline mediation in the prior art.

[0005] In order to achieve the above object, the present invention adopts the following technical solutions:

[0006] An online mediation solution generation method of the present invention comprises the steps of:

[0007] When receiving a conversation request from a customer, it uses the pre-built background knowledge base and the language model to perform multiple rounds of conversations with the customer.

[0008] When the conversation text of the multi-round conversation does not contain any target information, a pre-built universal mediation solution is sent to the customer, wherein the target information is the customer's identity information or the customer's status information;

[0009] When identity information exists in the conversation text of the multi-round conversation but state information does not exist, a mediation scheme with the highest acceptance probability is determined based on the identity information and a pre-built prediction model, wherein the prediction model includes acceptance probabilities of multiple mediation factors corresponding to multiple identity parameters under multiple states;

[0010] When identity information and state information exist in the dialogue text of multiple rounds of dialogue, a mediation solution with the highest acceptance probability is determined based on the identity information, the state information and a pre-built prediction model.

[0011] In one embodiment of the present application, based on a pre-built background knowledge base, a large language model is called to perform multiple rounds of dialogue with a customer, including:

[0012] S1, determining the current conversation turn, and sending a pre-built guide text to the customer based on the current conversation turn;

[0013] S2, when receiving a reply text from the customer, extracting information from the reply text; when no reply text is received from the customer after the target time, taking the next conversation round as the current conversation round, and returning to step S1 until all conversation rounds are completed;

[0014] S3, when the target information of the current dialogue round exists in the reply text, the next dialogue round is used as the current dialogue round, and the process returns to step S1 until all dialogue rounds are completed, wherein the target information is the customer identity or status feature;

[0015] S4, when the target information of the current dialogue round does not exist in the reply text and the question entity exists in the reply text, extract the keywords in the reply text, match the keywords with a pre-built background knowledge base, and when structured data matching the keywords exists in the background knowledge base, return the text in the structured data to the customer; when structured data matching the keywords does not exist in the background knowledge base, forward the reply text to the called language model for processing, and return to step S2.

[0016] In one embodiment of the present application, when there is no question entity in the reply text, the pre-constructed guide text is sent to the customer again based on the current dialogue turn, and the process returns to step S2.

[0017] In one embodiment of the present application, information extraction from the reply text includes:

[0018] Segmenting the reply text to obtain multiple words;

[0019] The recognition model is called to perform entity recognition on the multiple words. When the multiple words include an entity matching the current dialogue round, it is determined that the response text contains target information of the current dialogue round; otherwise, it is determined that the response text does not contain target information of the current dialogue round, wherein the recognition model is obtained by marking identity words or status words as training data and training an artificial neural network.

[0020] In one embodiment of the present application, the method for constructing the background knowledge base includes:

[0021] Obtaining background knowledge materials, wherein the background knowledge materials include the bank's policy documents, loan product information, and relevant legal provisions;

[0022] Segmenting the background knowledge material to obtain a plurality of paragraphs;

[0023] Calling the language big model to extract keywords and summarize the multiple paragraphs, and vectorizing the keywords, paragraph summaries, texts and corresponding titles of the multiple paragraphs to obtain structured data;

[0024] A background knowledge base is constructed based on the structured processing.

[0025] In one embodiment of the present application, the method for constructing the prediction model includes:

[0026] Obtaining questionnaire data of the target population, wherein the questionnaire data includes the current status and identity information of the respondents and the selection of multiple mediation factors, wherein the mediation factors include the repayment method, repayment period, and number of repayment installments;

[0027] Dividing the questionnaire data multiple times based on the current state and identity information to obtain multiple data units, wherein each data unit includes a selection of multiple mediation factors corresponding to a value or a value range of a single identity parameter in a current state;

[0028] Calculate the probability of selecting each moderator in each data unit Wherein, k represents the current state serial number, i represents the mediation factor serial number, j=n represents the value n or value range n of the identity parameter j, is the number of choices of mediation factor i for the group with value n or value range n of identity parameter j under the current state k, D is the total number of choices for the group with value n or value range n of identity parameter j under the current state k, and the current state includes good, medium, and poor;

[0029] The probability of selection of each moderator based on multiple data units Build a predictive model.

[0030] In one embodiment of the present application, determining a mediation solution with the highest acceptance probability based on the identity information and a pre-built prediction model includes:

[0031] Substituting the value of each identity parameter in the identity information into the prediction model to obtain an initial adjustment scheme for the value of each identity parameter under multiple states, wherein the initial adjustment scheme includes the probability of each selection of multiple adjustment factors;

[0032] Merging initial mediation solutions corresponding to multiple identity parameters in multiple states to obtain an intermediate mediation solution corresponding to each identity parameter, wherein the intermediate mediation solution includes the total probability of each selection of multiple adjustment factors;

[0033] The choice with the highest total probability is taken as the predicted choice of the corresponding moderator, and a mediation scheme is constructed based on the predicted choices of multiple moderators.

[0034] In one embodiment of the present application, determining a mediation solution with the highest acceptance probability based on the identity information, the state information, and a pre-built prediction model includes:

[0035] Determine the current state in the state information, and substitute the value of each identity parameter in the identity information in the current state into the prediction model to obtain an initial mediation scheme corresponding to each identity parameter in the current state, wherein the initial mediation scheme includes the probability of each selection of multiple adjustment factors;

[0036] Merging multiple initial mediation solutions to obtain an intermediate mediation solution, wherein the intermediate mediation solution includes the total probability of each selection of multiple adjustment factors;

[0037] The choice with the highest total probability is taken as the predicted choice of the corresponding moderator, and a mediation scheme is constructed based on the predicted choices of multiple moderators.

[0038] In one embodiment of the present application, it also includes:

[0039] When identity information exists in the conversation text of the multi-round conversation, extracting the overdue information of the customer based on the identity information, and generating a repayment plan based on the mediation plan and the overdue information;

[0040] The repayment plan is returned to the customer.

[0041] The present application also provides an online mediation solution generation system, comprising:

[0042] The dialogue module is used to perform multiple rounds of dialogue with the customer based on the pre-built background knowledge base and calling the language model when receiving a dialogue request from the customer;

[0043] A first generating module is used to send a pre-built universal mediation solution to the customer when the conversation text of the multi-round conversation does not contain any target information, wherein the target information is the customer's identity information or the customer's status information;

[0044] A second generation module is used to determine a mediation scheme with the highest acceptance probability based on the identity information and a pre-built prediction model when there is identity information but no state information in the dialogue text of the multi-round dialogue, wherein the prediction model includes acceptance probabilities of multiple mediation factors corresponding to multiple identity parameters under multiple states;

[0045] The third generation module is used to determine a mediation solution with the highest acceptance probability based on the identity information, the state information and a pre-built prediction model when there is identity information and state information in the dialogue text of the multi-round dialogue.

[0046] The beneficial effects of the present invention are as follows: an online mediation solution generation method and system of the present invention performs mediation through online dialogue. During the online dialogue, the guide text of each dialogue round is constructed to communicate with the customer, and the customer is guided to enter identity information and status information. Taking into account the complexity of the dialogue, the present application copes with the complex dialogue process by constructing a background knowledge base and calling a large language model. If no useful information is extracted during the dialogue, a general mediation solution is directly output to the customer after the dialogue ends. If identity information is extracted, the corresponding mediation solution is determined using the identity information and a pre-built prediction model and output to the customer. If identity information and status information are extracted, a more accurate personalized mediation solution is output to the customer using the identity information, status information and a pre-built prediction model. The present application adopts an online mediation method with high efficiency. At the same time, a corresponding mediation solution is formulated based on the characteristics of the customer with high accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] The present invention will be further described below in conjunction with the accompanying drawings and embodiments:

[0048] Figure 1 It is an application scenario diagram of an online mediation solution generation method shown in an embodiment of the present application;

[0049] Figure 2 is a flow chart of an online mediation solution generation method shown in an embodiment of the present application;

[0050] Figure 3 It is a structural diagram of an online mediation solution generation system shown in one embodiment of the present application. DETAILED DESCRIPTION

[0051] The following describes the embodiments of the present invention by specific examples, and those skilled in the art can easily understand other advantages and effects of the present invention from the contents disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and the details in this specification can also be modified or changed in various ways based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that the following embodiments and features in the embodiments can be combined with each other without conflict.

[0052] It should be noted that the illustrations provided in the following embodiments are only schematic illustrations of the basic concept of the present invention, and thus the drawings only show the layers related to the present invention rather than being drawn according to the number, shape and size of the layers in actual implementation. In actual implementation, the type, quantity and proportion of each layer may be changed arbitrarily, and the layer layout may also be more complicated.

[0053] In the following description, numerous details are discussed to provide a more thorough explanation of embodiments of the present invention; however, it is apparent to one skilled in the art that embodiments of the present invention may be practiced without these specific details.

[0054] Figure 1 is an application scenario diagram of an online mediation solution generation method shown in an embodiment of the present application, such as Figure 1 As shown, in this application, the client 110 can perform online mediation by accessing the server 120. The server 120 calls the language model by accessing the NLP (Natural Language Processing) server 130. At the same time, the server 120 sets up a background database internally, responds to the professional questions of the customer through the structured data in the background database, and calls the language model to respond to other questions of the customer. In this application, the server 120 extracts useful information from multiple rounds of dialogue and inputs it into the pre-built prediction model, so as to predict the mediation plan with a higher probability of customer acceptance.

[0055] Figure 2 is a flow chart of a method for generating an online mediation solution shown in an embodiment of the present application. Figure 1 As shown: An online mediation solution generation method of this embodiment may include steps S110 to S140:

[0056] S110, upon receiving a conversation request from a customer, based on a pre-built background knowledge base, calling a large language model to perform multiple rounds of conversations with the customer;

[0057] This application can be used to mediate between banks and overdue loan users, or for other purposes;

[0058] Taking the mediation between banks and users with overdue loans as an example, after receiving notification from the bank, the user can access the online mediation platform by logging into the relevant server and then conduct multiple rounds of conversations.

[0059] The multi-round dialogue in this application is requested by the client and responded by the server, thus officially entering into the dialogue process.

[0060] Considering the complexity of the conversation, this application calls the language big model service by linking the API interface to cope with the complex and changeable conversation process. In addition, taking the mediation of banks and overdue loan users as an example, customers may ask a lot of questions about bank policies, loan product information and related legal provisions. Therefore, this application also builds a background knowledge base with bank policies, loan product information and related legal provisions as background materials to fill in the background knowledge.

[0061] Specifically, with the server as the execution end, the process of multiple rounds of dialogue is as follows:

[0062] S1, determining the current conversation turn, and sending a pre-built guide text to the customer based on the current conversation turn;

[0063] In this application, for each round of dialogue, a corresponding guiding text is preset.

[0064] For example, the introductory text for the first round of conversation is mainly to make a brief introduction and let the customer enter personal information. The text example is:

[0065] "Hello, I am mediation robot XX. Please enter your personal information and press X to end."

[0066] The above guide text can be text or voice. If voice is used to perform multiple rounds of dialogue, the voice / text conversion service needs to be called for conversion. For example, the online business websocket polling access voice triggers the real-time speech-to-text (ASR) service based on the funAsr model.

[0067] S2, when receiving a reply text from the customer, extracting information from the reply text; when no reply text is received from the customer after the target time, taking the next conversation round as the current conversation round, and returning to step S1 until all conversation rounds are completed;

[0068] The customer's reply text may directly provide corresponding information based on the guide text, or it may input irrelevant information. Therefore, information extraction is required to determine whether the desired text information exists.

[0069] For example, the reply text in the first round of conversation needs to be extracted to see if there is any identity information. Identity information can be a name or an ID number. Since the identity information entered by the customer is usually incomplete, it is necessary to rely on the entered identity information to perform a search in the information database provided by the bank to obtain complete personal information.

[0070] Specifically, in order to protect the privacy of information, banks generally encrypt customer information. In this application, after obtaining part of the customer's identity information, it is matched in the information database. When any customer is hit, the corresponding request for the customer's secret key is received, and when the secret key fed back by the bank system is received, it is decrypted to obtain the complete customer information.

[0071] In the second round of dialogue, the guiding text usually guides the customer to provide status information to obtain the customer's recent financial situation. Generally speaking, the probability of overdue payments for customers in normal conditions is low, so obtaining the customer's financial situation has a certain reference value for the subsequent formulation of corresponding mediation plans.

[0072] Therefore, the reply text of the second round of dialogue needs to determine whether there is status information.

[0073] In one embodiment of the present application, information extraction from the reply text includes:

[0074] The reply text is segmented to obtain a plurality of words; the segmentation algorithm in this embodiment may adopt a string-based matching method or a statistical-based segmentation method (such as an N-gram model, a hidden Markov model (HMM), a conditional random field (CRF), etc.);

[0075] The recognition model is called to perform entity recognition on the multiple words. When the multiple words include an entity matching the current dialogue round, it is determined that the response text contains target information of the current dialogue round; otherwise, it is determined that the response text does not contain target information of the current dialogue round, wherein the recognition model is obtained by marking identity words or status words as training data and training an artificial neural network.

[0076] Specifically, this application is based on deep learning models: such as long short-term memory network (LSTM), BiLSTM+CRF, etc. to identify the identity information and status information in the reply text. For example, the reply text contains the name, ID number, and the description of the current status in the reply text, such as "unemployed", "divorced", "business failure", etc., which can all be accurately identified. Finally, the customer's status is summarized into three levels: good, medium, and poor. Different levels can correspond to different mediation plan strategies.

[0077] In some cases, some customers are unwilling to reveal their own status information. In this application, the emotional state of the reply text can also be used to analyze the customer's status. For example, a pre-built emotional word library is used to match multiple words, wherein the emotional word library contains multiple emotional tendency words, and the emotional weight of each emotional tendency word is pre-marked, for example, the weight of negative words is -1, the weight of neutral words is 0, and the weight of positive words is 1. The emotional word library is used to determine the emotional tendency weight of each word in the reply text, and finally the average value is calculated. If the average value is less than -0.3, the status is judged to be poor, the average value is greater than 0.3, the status is judged to be good, and the average value is between -0.3-0.3, and it is judged to be medium.

[0078] S3, when the target information of the current dialogue round exists in the reply text, the next dialogue round is used as the current dialogue round, and the process returns to step S1 until all dialogue rounds are completed, wherein the target information is the customer identity or status feature;

[0079] If the corresponding target information exists in the reply text of the current dialogue round, the next dialogue round is entered. For example, after obtaining the customer identity information, the dialogue round of obtaining the customer status information is entered. At this time, the process returns to step S1 and sends the guide text of the round of obtaining the customer status information to the customer.

[0080] S4, when the target information of the current dialogue round does not exist in the reply text and the question entity exists in the reply text, extract the keywords in the reply text, match the keywords with a pre-built background knowledge base, and when structured data matching the keywords exists in the background knowledge base, return the text in the structured data to the customer; when structured data matching the keywords does not exist in the background knowledge base, forward the reply text to the called language model for processing, and return to step S2.

[0081] When the corresponding target information does not exist in the reply text, it is necessary to determine whether the customer asks a counter-question. The customer may ask counter-questions about the loan product itself, the bank's repayment policy, and relevant legal knowledge. At this time, it is further determined whether there is a question entity in the reply text, such as "how", "why", "how", "?", etc. The present application determines whether there is a question entity by character matching. If it exists, it is confirmed that the customer asks a counter-question at this time. When there is no question entity in the reply text, it means that there is neither target information nor a counter-question in the reply text at this time, and it is judged to be an invalid text. At this time, the pre-constructed guide text is sent to the customer again based on the current dialogue turn, and return to step S2. To prompt the customer.

[0082] When the customer asks a counter-question, the application needs to determine what the customer wants to ask in the reply text. The application determines the key entities in the reply text by keyword extraction.

[0083] Key entities can be extracted by character matching or by training a recognition model.

[0084] For example, if the reply text is "How is the credit limit of XXX given?", the keywords "XXX" and "credit limit" can be extracted here. Therefore, the keywords can be matched in the background knowledge base to obtain the corresponding knowledge points and provide them to the customer.

[0085] Wherein, the method for constructing the background knowledge base includes:

[0086] (1) Obtaining background knowledge materials, wherein the background knowledge materials include the bank's policy documents, loan product information, and relevant legal provisions; the background knowledge materials may be in docx, pdf, html, and other formats, which can quickly obtain text content.

[0087] (2) segmenting the background knowledge material into multiple paragraphs;

[0088] When dividing paragraphs, the division can be performed based on paragraph marks. However, in this embodiment, for the coherence of the paragraph content, the division is performed using a machine learning method.

[0089] (3) calling the language big model to extract keywords and summarize the multiple paragraphs, and vectorizing the keywords, paragraph summaries, texts and corresponding titles of the multiple paragraphs to obtain structured data;

[0090] (4) Constructing a background knowledge base based on the structured processing. Specifically, paragraph summary embedding, title embedding, text embedding, keywords, and summary are stored in elasticsearch, the original file path of the text is stored in mysql, and the embedding type data is stored in milvus.

[0091] For example, the following paragraph:

[0092] Document: How is the XXX quota determined? The XXX quota is automatically determined by the system after a comprehensive evaluation based on multi-dimensional evaluation criteria.

[0093] The title is XXX FAQ;

[0094] Extract keywords, XXX, amount, evaluation;

[0095] Summary: Du Xiaoman's credit limit is a multi-dimensional evaluation system, and the system automatically evaluates;

[0096] Then use the existing embedding model shaw / dmeta-embedding-zh to calculate the embedding vector values ​​of keywords, titles, summaries, and texts, and store them in the corresponding databases in turn.

[0097] Taking the question "How is the credit limit of XXX given?" as an example, the keywords "XXX" and "credit limit" were extracted in the previous text. Through phrase matching, we can get the text content corresponding to "extracted keywords, XXX, credit limit, evaluation", that is, "The XXX credit limit is automatically given by the system after a comprehensive evaluation based on multi-dimensional evaluation standards."

[0098] The above question is about background knowledge. If the reply text question is not about background knowledge and the keywords cannot be extracted, it will be transferred to the language model for processing, and the existing language model will be used to answer the customer's question, and then return to step S2 to extract relevant information from the customer's reply again.

[0099] The above process is repeated over and over again, which can effectively handle customer conversations in various situations and extract as much valid information as possible.

[0100] S120, when the conversation text of the multi-round conversation does not contain any target information, sending a pre-built universal mediation solution to the customer, wherein the target information is the customer's identity information or the customer's status information;

[0101] If the customer does not reply or keeps replying invalid text, after the conversation ends, it is impossible to predict the solution through relevant feature information. Then the pre-built general adjustment solution is directly output, for example: 15,000 yuan is recommended to be divided into 6 to 12 installments. The above installment plan is determined in advance with the bank, that is, if the customer accepts the plan, it means that the mediation is successful.

[0102] S130, when identity information exists in the conversation text of the multi-round conversation but state information does not exist, determining a mediation scheme with the highest acceptance probability based on the identity information and a pre-built prediction model, wherein the prediction model includes acceptance probabilities of multiple mediation factors corresponding to multiple identity parameters under multiple states;

[0103] S140, when identity information and state information exist in the dialogue text of the multi-round dialogue, determine a mediation solution with the highest acceptance probability based on the identity information, the state information and a pre-built prediction model.

[0104] In step S130 and step S140, if identity information is extracted from multiple rounds of dialogue, or identity information and status information are extracted, then a mediation solution with the highest probability can be determined based on a pre-built prediction model.

[0105] The prediction model in this application is constructed based on the questionnaire data, and explains the mediation selection tendency of various groups from a statistical perspective. For example, people with better economic status are generally more inclined to accept the mediation plan, young people are more inclined to repay in installments, and office workers are more inclined to pay more installments than freelancers or self-employed individuals.

[0106] The big data in the questionnaire data is used to reflect the selection tendency of multiple groups of people. Finally, the various identity parameters of the customers are substituted into the model to obtain the mediation plan with the highest probability corresponding to each identity parameter. Finally, from multiple mediation plans, the highest probability option can be determined.

[0107] In one embodiment of the present application, the method for constructing the prediction model includes:

[0108] (1) obtaining questionnaire data of the target population, wherein the questionnaire data includes the current status and identity information of the respondents and the selection of multiple mediation factors, wherein the mediation factors include the repayment method, repayment period, and number of repayment installments;

[0109] Among them, a questionnaire survey was conducted for the target group who had overdue behavior or was in overdue status. A large amount of questionnaire data was obtained. The questionnaire included the current status, identity information and the selection of multiple mediation factors. For example, the repayment method selection includes one-time repayment and installment repayment. When repaying by installment, the repayment period can be selected as half a year, one year or more, and the number of repayment installments can be selected as 3 installments, 6 installments, 12 installments or more.

[0110] (2) dividing the questionnaire data multiple times based on the current state and identity information to obtain multiple data units, wherein each data unit includes a selection of multiple mediation factors corresponding to a value or value range of a single identity parameter in the current state;

[0111] For example, the current status is divided into good, medium, and poor. In the identity information, gender is divided into male and female, occupation is divided into working and self-employed, and age is divided into three age groups: 20-30, 30-50, and above 50.

[0112] The questionnaire data is divided into three data sets based on the status of good, medium and poor, and then each data set is divided into 2 data units based on male and female, and each data set is divided into three data units based on the three age groups of 20-30, 30-50, and above 50, etc.

[0113] (3) Calculate the probability of selecting each mediation factor in each data unit Wherein, k represents the current state serial number, i represents the mediation factor serial number, j=n represents the value n or value range n of the identity parameter j, is the number of choices of mediation factor i for the group with value n or value range n of identity parameter j under the current state k, D is the total number of choices for the group with value n or value range n of identity parameter j under the current state k, and the current state includes good, medium, and poor;

[0114] Each data unit represents the selection bias of a population in one of the states. This application uses the probability of selection of each mediator To reflect the selection bias of the crowd.

[0115] (4) The probability of selection of each moderator in multiple data units Build a predictive model.

[0116] Finally, the selection biases of multiple populations are used to construct a prediction model.

[0117] Therefore, the final process of predicting the mediation plan for customers based on the prediction model is as follows.

[0118] In one embodiment of the present application, determining a mediation solution with the highest acceptance probability based on the identity information and a pre-built prediction model includes:

[0119] (1) Substituting the value of each identity parameter in the identity information into the prediction model to obtain an initial adjustment scheme for the value of each identity parameter under multiple states, wherein the initial adjustment scheme includes the probability of each selection of multiple adjustment factors;

[0120] The initial mediation scheme includes the selection probabilities of multiple moderators under multiple states;

[0121] For example, for people aged 20-30, when the economic situation is poor, there is a 20% probability of choosing more than 12 periods, while when the economic situation is good, there is only a 5% probability of choosing more than 12 periods.

[0122] (2) merging the initial mediation solutions corresponding to the multiple identity parameters under the multiple states to obtain an intermediate mediation solution corresponding to each identity parameter, wherein the intermediate mediation solution includes the total probability of each selection of the multiple adjustment factors;

[0123] The combined intermediate mediation scheme not only sums the selection probabilities of multiple adjustment factors in the three states, but also sums the selection probabilities of multiple adjustment factors corresponding to each identity parameter. For example, customer A, 25 years old, male. Belonging to the 20-30 age group, male group, the initial mediation schemes include:

[0124] People aged 20-30: Good: Factor 1: Option A: 20%, Option B: 80%;

[0125] Middle: Factor 1: Option A: 40%, Option B: 60%;

[0126] Difference: Factor 1: Option A: 60%, Option B: 40%;

[0127] Male group: Good: Factor 1: Option A: 35%, Option B: 65%;

[0128] Middle: Factor 1: Option A: 40%, Option B: 60%;

[0129] Difference: Factor 1: Option A: 45%, Option B: 55%;

[0130] After the merger, in the intermediate mediation plan, for mediation factor one, option A: 240%, option B: 360%; other mediation factors are calculated according to the above method.

[0131] (3) The choice with the highest total probability is taken as the predicted choice of the corresponding regulatory factor, and a mediation plan is constructed based on the predicted choices of multiple regulatory factors.

[0132] Finally, the option with the highest probability is taken as the predicted option and forms the final mediation plan.

[0133] In one embodiment of the present application, determining a mediation solution with the highest acceptance probability based on the identity information, the state information, and a pre-built prediction model includes:

[0134] (1) determining a current state in the state information, and substituting the value of each identity parameter in the identity information in the current state into the prediction model to obtain an initial mediation scheme corresponding to each identity parameter in the current state, wherein the initial mediation scheme includes the probability of each selection of multiple mediation factors;

[0135] (2) merging the multiple initial mediation schemes to obtain an intermediate mediation scheme, wherein the intermediate mediation scheme includes the total probability of each selection of the multiple adjustment factors;

[0136] (3) The choice with the highest total probability is taken as the predicted choice of the corresponding regulatory factor, and a mediation plan is constructed based on the predicted choices of multiple regulatory factors.

[0137] The process is similar to the above, except that mediation solutions in different states are no longer merged. For example:

[0138] People aged 20-30: Good: Factor 1: Option A: 20%, Option B: 80%;

[0139] Male group: Good: Factor 1: Option A: 35%, Option B: 65%;

[0140] In the final solution, for factor one, option A: 55%, option B: 145%; it is speculated that for mediation factor one, customers will choose option B.

[0141] Finally, when identity information exists in the conversation text of multiple rounds of conversations, the customer's overdue information is extracted based on the identity information, and a repayment plan is generated based on the mediation plan and the overdue information; and the repayment plan is returned to the customer.

[0142] In addition, if the client hangs up or loses connection for any reason, the business model interrupts the current program to save computing resources. Or if the business model detects the end of the current conversation, it ends the program after playing the last polite words.

[0143] An online mediation solution generation method of the present invention performs mediation through online dialogue. During the online dialogue, the guide text of each dialogue round is constructed to communicate with the customer, and the customer is guided to enter identity information and status information. Taking into account the complexity of the dialogue, the present application copes with the complex dialogue process by constructing a background knowledge base and calling a large language model. If no useful information is extracted during the dialogue, a general mediation solution is directly output to the customer after the dialogue ends. If identity information is extracted, the corresponding mediation solution is determined using the identity information and a pre-built prediction model and output to the customer. If identity information and status information are extracted, a more accurate personalized mediation solution is output to the customer using the identity information, status information and a pre-built prediction model. The present application adopts an online mediation method with high efficiency. At the same time, a corresponding mediation solution is formulated based on the characteristics of the customer with high accuracy.

[0144] like Figure 3 As shown, the present application also provides an online mediation solution generation system, comprising:

[0145] The dialogue module is used to perform multiple rounds of dialogue with the customer based on the pre-built background knowledge base and calling the language model when receiving a dialogue request from the customer;

[0146] A first generating module is used to send a pre-built universal mediation solution to the customer when the conversation text of the multi-round conversation does not contain any target information, wherein the target information is the customer's identity information or the customer's status information;

[0147] A second generation module is used to determine a mediation scheme with the highest acceptance probability based on the identity information and a pre-built prediction model when there is identity information but no state information in the dialogue text of the multi-round dialogue, wherein the prediction model includes acceptance probabilities of multiple mediation factors corresponding to multiple identity parameters under multiple states;

[0148] The third generation module is used to determine a mediation solution with the highest acceptance probability based on the identity information, the state information and a pre-built prediction model when there is identity information and state information in the dialogue text of the multi-round dialogue.

[0149] An online mediation solution generation system of the present invention performs mediation through online dialogue. During the online dialogue, the system communicates with the customer by constructing a guide text for each dialogue round, and guides the customer to enter identity information and status information. Taking into account the complexity of the dialogue, the present application copes with the complex dialogue process by constructing a background knowledge base and calling a large language model. If no useful information is extracted during the dialogue, a general mediation solution is directly output to the customer after the dialogue ends. If identity information is extracted, the corresponding mediation solution is determined using the identity information and a pre-built prediction model and output to the customer. If identity information and status information are extracted, a more accurate personalized mediation solution is output to the customer using the identity information, status information and a pre-built prediction model. The present application adopts an online mediation method with high efficiency. At the same time, a corresponding mediation solution is formulated based on the characteristics of the customer with high accuracy.

[0150] This embodiment further provides a computer-readable storage medium on which a computer program is stored. When the program is executed by a processor, any one of the methods in this embodiment is implemented, wherein the method is the execution logic of this system.

[0151] This embodiment also provides an electronic terminal, including: a processor and a memory;

[0152] The memory is used to store a computer program, and the processor is used to execute the computer program stored in the memory, so that the terminal executes any one of the methods in this embodiment.

[0153] The computer-readable storage medium in this embodiment can be understood by ordinary technicians in this field: all or part of the steps of implementing the above-mentioned method embodiments can be completed by hardware related to the computer program. The aforementioned computer program can be stored in a computer-readable storage medium. When the program is executed, the steps of the above-mentioned method embodiments are executed; and the aforementioned storage medium includes: ROM, RAM, magnetic disk or optical disk and other media that can store program codes.

[0154] The electronic terminal provided in this embodiment includes a processor, a memory, a transceiver and a communication interface. The memory and the communication interface are connected to the processor and the transceiver and complete communication with each other. The memory is used to store computer programs, the communication interface is used to communicate, and the processor and the transceiver are used to run computer programs so that the electronic terminal executes each step of the above method.

[0155] In this embodiment, the memory may include a random access memory (RAM), and may also include a non-volatile memory (non-volatile memory), such as at least one disk memory.

[0156] The above-mentioned processor can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components.

[0157] In the above-mentioned embodiments, although the present invention has been described in conjunction with the specific embodiments of the present invention, many replacements, modifications and variations of these embodiments will be apparent to those of ordinary skill in the art based on the foregoing description. The embodiments of the present invention are intended to cover all such replacements, modifications and variations falling within the broad scope of the appended claims.

[0158] The above embodiments are merely illustrative of the principles and effects of the present invention, and are not intended to limit the present invention. Anyone familiar with the art may modify or alter the above embodiments without departing from the spirit and scope of the present invention. Therefore, all equivalent modifications or alterations made by a person of ordinary skill in the art without departing from the spirit and technical concept disclosed by the present invention shall still be covered by the claims of the present invention.

Claims

1. A method for generating an online mediation solution, characterized in that: Includes steps: When receiving a conversation request from a customer, it uses the pre-built background knowledge base and the language model to perform multiple rounds of conversations with the customer. When the conversation text of the multi-round conversation does not contain any target information, a pre-built universal mediation solution is sent to the customer, wherein the target information is the customer's identity information or the customer's status information; When identity information exists in the conversation text of the multi-round conversation but state information does not exist, a mediation scheme with the highest acceptance probability is determined based on the identity information and a pre-built prediction model, wherein the prediction model includes acceptance probabilities of multiple mediation factors corresponding to multiple identity parameters under multiple states; When identity information and state information exist in the dialogue text of multiple rounds of dialogue, a mediation solution with the highest acceptance probability is determined based on the identity information, the state information and a pre-built prediction model.

2. The method for generating an online mediation solution according to claim 1, characterized in that: Based on the pre-built background knowledge base and calling the language model, multiple rounds of conversations with customers are performed, including: S1, determining the current conversation turn, and sending a pre-built guide text to the customer based on the current conversation turn; S2, when receiving a reply text from the customer, extracting information from the reply text; when no reply text is received from the customer after the target time, taking the next conversation round as the current conversation round, and returning to step S1 until all conversation rounds are completed; S3, when the target information of the current dialogue round exists in the reply text, the next dialogue round is used as the current dialogue round, and the process returns to step S1 until all dialogue rounds are completed, wherein the target information is the customer identity or status feature; S4, when the target information of the current dialogue round does not exist in the reply text and the question entity exists in the reply text, extract the keywords in the reply text, match the keywords with a pre-built background knowledge base, and when structured data matching the keywords exists in the background knowledge base, return the text in the structured data to the customer; when structured data matching the keywords does not exist in the background knowledge base, forward the reply text to the called language model for processing, and return to step S2.

3. The method for generating an online mediation solution according to claim 2, characterized in that: When there is no question entity in the reply text, the pre-constructed guide text is sent to the customer again based on the current dialogue turn, and the process returns to step S2.

4. The method for generating an online mediation solution according to claim 1, characterized in that: Extracting information from the reply text includes: Segmenting the reply text to obtain multiple words; The recognition model is called to perform entity recognition on the multiple words. When the multiple words include an entity matching the current dialogue round, it is determined that the response text contains target information of the current dialogue round; otherwise, it is determined that the response text does not contain target information of the current dialogue round, wherein the recognition model is obtained by marking identity words or status words as training data and training an artificial neural network.

5. The method for generating an online mediation solution according to claim 1, characterized in that: The method for constructing the background knowledge base includes: Obtaining background knowledge materials, wherein the background knowledge materials include the bank's policy documents, loan product information, and relevant legal provisions; Segmenting the background knowledge material to obtain a plurality of paragraphs; Calling the language big model to extract keywords and summarize the multiple paragraphs, and vectorizing the keywords, paragraph summaries, texts and corresponding titles of the multiple paragraphs to obtain structured data; A background knowledge base is constructed based on the structured processing.

6. The method for generating an online mediation solution according to claim 1, characterized in that: The method for constructing the prediction model includes: Obtaining questionnaire data of the target population, wherein the questionnaire data includes the current status and identity information of the respondents and the selection of multiple mediation factors, wherein the mediation factors include the repayment method, repayment period, and number of repayment installments; Dividing the questionnaire data multiple times based on the current state and identity information to obtain multiple data units, wherein each data unit includes a selection of multiple mediation factors corresponding to a value or a value range of a single identity parameter in a current state; Calculate the probability of selecting each moderator in each data unit Wherein, k represents the current state serial number, i represents the mediation factor serial number, j=n represents the value n or value range n of the identity parameter j, is the number of choices of mediation factor i for the group with value n or value range n of identity parameter j under the current state k, D is the total number of choices for the group with value n or value range n of identity parameter j under the current state k, and the current state includes good, medium, and poor; The probability of selection of each moderator based on multiple data units Build a predictive model.

7. The method for generating an online mediation solution according to claim 6, characterized in that: Determine a mediation solution with the highest probability of acceptance based on the identity information and a pre-built prediction model, including: Substituting the value of each identity parameter in the identity information into the prediction model to obtain an initial adjustment scheme for the value of each identity parameter in multiple states, wherein the initial adjustment scheme includes the probability of each selection of multiple adjustment factors; Merging initial mediation solutions corresponding to multiple identity parameters in multiple states to obtain an intermediate mediation solution corresponding to each identity parameter, wherein the intermediate mediation solution includes a total probability of each selection of multiple adjustment factors; The choice with the highest total probability is taken as the predicted choice of the corresponding moderator, and a mediation scheme is constructed based on the predicted choices of multiple moderators.

8. The method for generating an online mediation solution according to claim 6, characterized in that: Determining a mediation solution with the highest acceptance probability based on the identity information, the state information, and a pre-built prediction model includes: Determine the current state in the state information, and substitute the value of each identity parameter in the identity information in the current state into the prediction model to obtain an initial mediation scheme corresponding to each identity parameter in the current state, wherein the initial mediation scheme includes the probability of each selection of multiple adjustment factors; Merging multiple initial mediation solutions to obtain an intermediate mediation solution, wherein the intermediate mediation solution includes the total probability of each selection of multiple adjustment factors; The choice with the highest total probability is taken as the predicted choice of the corresponding moderator, and a mediation scheme is constructed based on the predicted choices of multiple moderators.

9. The method for generating an online mediation solution according to claim 1, characterized in that: Also includes: When identity information exists in the conversation text of the multi-round conversation, extracting the overdue information of the customer based on the identity information, and generating a repayment plan based on the mediation plan and the overdue information; The repayment plan is returned to the customer.

10. An online mediation solution generation system, characterized in that: include: The dialogue module is used to perform multiple rounds of dialogue with the customer based on the pre-built background knowledge base and calling the language model when receiving a dialogue request from the customer; A first generating module is used to send a pre-built universal mediation solution to the customer when the conversation text of the multi-round conversation does not contain any target information, wherein the target information is the customer's identity information or the customer's status information; A second generation module is used to determine a mediation scheme with the highest acceptance probability based on the identity information and a pre-built prediction model when there is identity information but no state information in the dialogue text of the multi-round dialogue, wherein the prediction model includes acceptance probabilities of multiple mediation factors corresponding to multiple identity parameters under multiple states; The third generation module is used to determine the mediation scheme with the highest acceptance probability based on the identity information, the state information and a pre-built prediction model when there is identity information and state information in the dialogue text of the multi-round dialogue.

Images