Typical skill configuration method, device and equipment based on collection dialogue library and medium

Abstract

The invention relates to the field of manpower and finance, and provides a verbal skill configuration method based on a collection dialogue library, and the method comprises the steps: obtaining a collection voice sample set and corresponding historical collection data from a sample library, classifying the collection voice sample set according to the historical collection data, and obtaining at least one successful collection voice sample; performing keyword processing on the text of the successful collection voice sample to obtain a keyword of a first repayment willingness, extracting a first response skill from the successful collection voice sample according to the keyword, and collecting the first response skill of all the successful collection voice samples to obtain an initial collection dialogue library; revising non-compliant texts in the initial collection dialogue library by using a compliant corpus to obtain a target collection dialogue library; and obtaining voice information of the to-be-collected object, extracting a voice keyword of the voice information, performing matching from the target collection dialogue library according to the voice keyword to obtain a target corresponding dialogue, and feeding back the target corresponding dialogue to the to-be-collected object for collection.

Description

Technical Field

[0001] The present invention relates to the fields of artificial intelligence and finance, and to a method, device, equipment and medium for configuring speech based on a debt collection dialogue library. Background Art

[0002] With the steady development of the national economy, financial credit technologies, such as credit card loans, unsecured loans, and mortgage loans, have entered a stage of large-scale development. In order to survive in the fiercely competitive market, banks or lending financial institutions are pursuing customer scale while the non-performing loan rate is also increasing simultaneously. Therefore, the scale of non-performing asset disposal business of banks or lending financial institutions will become larger and larger, and the amount of business that needs to be collected is also increasing.

[0003] At present, the collection method adopted for the collection business is completed manually, and the collection personnel communicate with the collection personnel through telephone calls, etc. However, the manual collection method cannot directly provide the most effective collection words, resulting in a low collection success rate of the collection business.

[0004] Therefore, there is an urgent need for a method that can improve the collection success rate of collection business. Summary of the invention

[0005] The present invention provides a method, device, equipment and medium for configuring words based on a collection dialogue library, the main purpose of which is to solve the technical problem that it is impossible for humans to provide the most effective collection words, resulting in a low success rate of collection business.

[0006] To achieve the above-mentioned purpose, the present invention provides a method, device, equipment and medium for configuring speech based on a debt collection dialogue library, the method comprising:

[0007] Acquire a collection voice sample set and corresponding historical collection data from a sample library, classify the collection voice sample set according to the historical collection data, and obtain at least one successful collection voice sample;

[0008] Perform keyword processing on the text of the successful collection voice sample to obtain a keyword of the first repayment willingness, extract a first response dialogue from the successful collection voice sample according to the keyword, collect the first response dialogues of all successful collection voice samples, and obtain an initial collection dialogue library;

[0009] Using a preset compliance corpus to revise the non-compliant text in the initial collection dialogue library to obtain a target collection dialogue library;

[0010] Acquire the voice information of the object to be collected, extract the voice keywords of the voice information, match the target response dialogue from the target collection dialogue library according to the voice keywords, and feed back the target response dialogue to the object to be collected for collection.

[0011] Optionally, before acquiring the collection voice sample set and the corresponding historical collection data from the sample library, the method further includes:

[0012] All the collection voice samples in the sample library are classified according to the collection business type corresponding to the collection voice to obtain at least one category of collection voice samples.

[0013] Optionally, before performing keyword processing on the text of the successful collection voice sample, the method further includes:

[0014] Performing noise reduction processing on the successful collection voice sample by using an adaptive filter;

[0015] The successful collection voice samples after noise reduction are converted into collection text samples through a preset speech recognition algorithm.

[0016] Optionally, the converting of the successful debt collection voice sample after noise reduction processing into a debt collection text sample by a preset voice recognition algorithm includes:

[0017] Extracting speech sequences of different tones from the successful debt collection speech samples by using the preset speech recognition algorithm;

[0018] Calculating the timbre similarity between the speech sequences of different timbres and the preset speech features in the speech feature library respectively;

[0019] Determining a first voice feature and a second voice feature according to the timbre similarity;

[0020] The speech sequence corresponding to the first speech feature and the speech sequence corresponding to the second speech feature are converted into the collection text sample.

[0021] Optionally, before generating an initial debt collection dialogue library according to the first response dialogue set, the method further includes:

[0022] A plurality of emotion tags are obtained, and an emotion mapping table is constructed according to the plurality of emotion tags and the first response dialogue.

[0023] Optionally, after acquiring the collection voice sample set and the corresponding historical collection data from the sample library, the method further includes:

[0024] Classifying the collection voice sample set according to the historical collection data to obtain at least one collection failure voice sample;

[0025] The text of the failed collection voice sample is processed with keywords to obtain a second keyword of the second repayment willingness, and a second response dialogue is extracted from the failed collection voice sample according to the second keyword, and the second response dialogue of all failed collection voice samples is collected.

[0026] Optionally, the step of matching the target response dialogue from the target collection dialogue library according to the voice keywords includes:

[0027] The voice information is recognized by a pre-built emotion detection model to obtain an emotion label of the object to be collected;

[0028] The voice information is converted into object text information in real time by the preset voice recognition algorithm, and voice keywords are extracted from the object text information;

[0029] The target response dialogue is obtained by matching the emotion tag and the voice keyword from the target collection dialogue library.

[0030] In order to solve the above problems, the present invention provides a speech configuration device based on a debt collection dialogue library, the device comprising:

[0031] A classification module, used to obtain a collection voice sample set and corresponding historical collection data from a sample library, classify the collection voice sample set according to the historical collection data, and obtain at least one successful collection voice sample;

[0032] An extraction module is used to perform keyword processing on the text of the successful collection voice sample to obtain a keyword of the first repayment willingness, extract a first response dialogue from the successful collection voice sample according to the keyword, collect the first response dialogues of all successful collection voice samples, and obtain an initial collection dialogue library;

[0033] A revision module, used to revise the non-compliant text in the initial collection dialogue library using a preset compliant corpus to obtain a target collection dialogue library;

[0034] The feedback module is used to obtain voice information of the object to be collected, extract voice keywords from the voice information, match the target response dialogue from the target collection dialogue library according to the voice keywords, and feed back the target response dialogue to the object to be collected for collection.

[0035] In order to solve the above problem, the present invention further provides an electronic device, the electronic device comprising:

[0036] at least one processor; and,

[0037] a memory communicatively connected to the at least one processor; wherein,

[0038] The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the speech method based on the collection dialogue library as described in any one of claims 1 to 7.

[0039] In order to solve the above-mentioned problem, the present invention also provides a computer-readable storage medium, and when the computer program is executed by a processor, the speech method based on the collection dialogue library is implemented.

[0040] In this embodiment, by obtaining successful collection voice samples as the basis for analysis, the collection success rate of the collection business can be improved; by classifying the collection voice samples according to the type of collection business, and analyzing the collection sample voice based on each category, it is highly targeted, which improves the accuracy of the subsequent target dialogue and improves the collection success rate of the collection business; by taking into account the willingness to repay, the selected response dialogue can be more accurate, and the response dialogue selected according to the willingness to repay can use the most accurate response dialogue as the allowed dialogue, which is more likely to exclude the situation where the user does not repay, and improve the collection success rate of the collection business; by summarizing the first response dialogue and the second response dialogue, the first response dialogue is used as the allowed dialogue, and the second response dialogue is used as the allowed dialogue. Second, response dialogues are used as prohibited dialogues to prevent debt collectors from using prohibited dialogues, which may lead to users refusing to repay the debt, and indirectly improve the collection success rate of the debt collection business; the first response dialogue is used to collect debts from customers, which directly improves the collection success rate of the debt collection business; the preset RPA tool is used to automatically obtain the latest laws and regulations and the compliance requirements customized within the enterprise from the Internet in real time, and a preset compliance corpus is generated based on the latest laws and regulations and the compliance requirements customized within the enterprise. The compliance corpus is used to correct the initial collection dialogue library, which can prevent debt collectors from touching on clauses that are not allowed by law and reduce the complaint rate; through the participation of emotional tags, more effective response dialogues can be obtained for debt collection, which greatly improves the collection success rate of the debt collection business. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 This is a flow chart of the method for configuring the words based on the collection dialogue library.

[0042] Figure 2 This is a module diagram of the method for configuring the words based on the collection dialogue library.

[0043] Figure 3 This is a schematic diagram of an electronic device for configuring the method for speech based on the debt collection dialogue library.

[0044] The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0045] It should be understood that the specific embodiments described herein are only used to explain the present invention, and are not used to limit the present invention.

[0046] The embodiments of the present invention can acquire and process relevant data based on artificial intelligence technology and the financial field. 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 obtain the best results.

[0047] The basic technologies of artificial intelligence generally include sensors, dedicated artificial intelligence chips, cloud computing, distributed storage, big data processing technology, operation / interaction systems, mechatronics, etc. Artificial intelligence software technology mainly includes computer vision technology, robotics technology, biological retrieval technology, speech processing technology, natural language processing technology, and machine learning / deep learning.

[0048] The embodiment of the present invention provides a method for configuring speech based on a collection dialogue library, and the execution subject of the method for configuring speech based on the collection dialogue library includes but is not limited to at least one of the electronic devices such as a client and a mobile terminal that can be configured to execute the method provided by the embodiment of the present application. The APP can be a blockchain platform. The mobile terminal includes but is not limited to: a single server, a server cluster, a cloud server or a cloud server cluster, etc.

[0049] In an embodiment of the present invention, the method for configuring speech based on the collection dialogue library first obtains a collection voice sample set and corresponding historical collection data from a sample library, classifies the collection voice sample set according to the historical collection data, and obtains at least one successful collection voice sample; performs keyword processing on the text of the successful collection voice sample to obtain a keyword of the first repayment willingness, extracts a first response dialogue from the successful collection voice sample according to the keyword, collects the first response dialogue of all successful collection voice samples, and obtains an initial collection dialogue library; uses a preset compliant corpus to revise the non-compliant text in the initial collection dialogue library to obtain a target collection dialogue library; obtains voice information of the object to be collected, extracts voice keywords of the voice information, matches the target response dialogue from the target collection dialogue library according to the voice keywords, and feeds the target response dialogue back to the object to be collected for collection.

[0050] To facilitate understanding of this embodiment, a method for configuring speech based on a collection dialogue library disclosed in an embodiment of the present application is first introduced in detail. Figure 1A flowchart of a method for configuring a collection dialogue library based on an embodiment of the present application is shown. Figure 1 As shown, the following steps are included:

[0051] S1. Obtain a collection voice sample set and corresponding historical collection data from a sample library, classify the collection voice sample set according to the historical collection data, and obtain at least one successful collection voice sample.

[0052] In the process of debt collection, since the most effective response dialogue can improve the success rate of debt collection, the present application classifies the debt collection voice sample set to obtain at least one successful debt collection voice sample.

[0053] In this embodiment, each collection business corresponds to a collection voice sample and historical collection data. The sample library includes the collection voice samples and historical collection data corresponding to each collection business. All collection voice samples and the historical collection data corresponding to each collection voice sample are obtained from the sample library. The collection voice sample set is classified according to the repayment results of the historical collection data to obtain successful collection voice samples and failed collection voice samples, and at least one successful collection voice sample is selected.

[0054] In one embodiment, before acquiring the collection voice sample set and the corresponding historical collection data from the sample library, the method further includes:

[0055] All the collection voice samples in the sample library are classified according to the collection business type corresponding to the collection voice to obtain at least one category of collection voice samples.

[0056] In this embodiment, each collection voice sample corresponds to a collection business, and all collection voice samples in the sample library are classified according to the type of the collection business to obtain at least one category of collection voice samples.

[0057] In this embodiment, the collection business refers to the business of the creditor to collect and collect overdue accounts from the borrower when the borrower fails to repay the loan in accordance with the repayment time and method agreed in the contract; the collection voice sample refers to the actual call recording between the collector and the customer, or it can be a simulated conversation simulated by the collection dialogue model for each category of collection business; the historical collection data refers to the collection case results corresponding to each collection business, including the collection cycle, the number of collections and the final repayment result; the collection cycle refers to the time span of the collection from the time point of the start of the collection, and the collection cycle can reflect the rationality of the time span of the collection after the borrower fails to repay the loan on time; the number of collections refers to the number of collections conducted on the collection case during the collection process, and the collection methods include telephone communication, SMS notification, etc.; the repayment result refers to whether the borrower of the case finally successfully repays the loan; the sample library refers to the collection dialogue set corresponding to each type of collection business and the historical collection data of each collection business that are pre-collected and recorded.

[0058] In the implementation scenario of the financial field, for example, a sample library is constructed, and the collection voice samples and historical collection data corresponding to all collection businesses are stored in the sample library. The data in the sample library is classified according to the type of collection business (such as credit card collection business, unsecured loan collection business, mortgage loan collection business, etc.), and the collection voice samples and historical collection data corresponding to the three categories of collection business are obtained. For any category of collection business, the corresponding collection voice samples and the corresponding historical collection data are obtained, and the collection voice samples of successful collection are obtained from the collection voice samples according to the successful repayment results of the historical collection data, and the collection voice samples of failed collection are obtained from the collection voice samples according to the failed repayment results of the historical collection data.

[0059] In this embodiment, the collection voice samples are classified according to the type of collection business, the collection sample voices are analyzed based on each category, and then the successful samples of the collection sample voices in each category are analyzed. This has strong pertinence, improves the accuracy of the subsequent target speech, and improves the collection success rate of the collection business.

[0060] In one embodiment, the classifying the collection voice sample set according to the historical collection data to obtain at least one successful collection voice sample includes:

[0061] Obtaining the repayment result in the historical collection data corresponding to the collection voice sample;

[0062] The collection voice samples are classified according to the repayment results to obtain at least one successful collection voice sample.

[0063] In this embodiment, the repayment result in the historical collection data corresponding to each collection voice sample is obtained, and the collection voice samples with the repayment result of successful repayment are classified to obtain the collection voice samples with successful collection.

[0064] In this embodiment, by obtaining successful collection voice samples as the basis for analysis, the collection success rate of the collection business can be improved; by classifying the collection voice samples according to the type of collection business, and analyzing the collection sample voice based on each category, it has strong pertinence, improves the accuracy of the subsequent target speech, and improves the collection success rate of the collection business.

[0065] S2. Perform keyword processing on the text of the successful collection voice sample to obtain keywords of the first willingness to repay, extract the first response dialogue from the successful collection voice sample according to the keywords, collect the first response dialogues of all successful collection voice samples, and obtain an initial collection dialogue library.

[0066] In this embodiment, successfully collected voice samples are converted into collection text samples through a pre-trained speech recognition algorithm, and the corresponding keywords of the first repayment willingness in each collection sample text are identified, and the similarity of every two collection sample texts is calculated based on the keywords, and every two collection text samples with a similarity greater than a preset threshold are classified into the same category to obtain at least one category of collection text samples, and a preset number of response dialogues are extracted from the successfully collected voice samples based on the keywords of the first repayment willingness of the same category, and the first response dialogues of all successfully collected voice samples are collected to obtain an initial collection dialogue library.

[0067] In this embodiment, the first willingness to repay refers to the degree of effective repayment during the collection process, for example: the customer repays directly, the customer expresses fundraising, the customer expresses the hope for reduction or exemption, etc.; the first response dialogue set refers to the most effective response dialogue set in the collection process; the keyword refers to the keyword corresponding to the first willingness to repay in the successful collection voice sample. For example, the first willingness to repay corresponding to this case is to indicate interest reduction or exemption, then the customer said: "Can the interest be reduced or exempted?" "Interest reduction or exemption" is the keyword corresponding to the first willingness to repay.

[0068] In the implementation scenario of the financial field, multiple successful credit card collection voices are obtained, and each credit card collection voice is converted into a credit card collection sample through a speech recognition model. The debt collector text samples and customer text samples in the credit card collection samples are identified, and the keywords of the first repayment willingness of each customer text sample are identified. The similarity of each customer text sample is calculated according to the keywords. When the similarity is greater than a threshold, they are classified into the same category, and the keywords corresponding to each repayment willingness are extracted. At least one response dialogue is selected from the successful collection voice samples according to the keywords, and all the first response dialogues are collected to obtain an initial collection dialogue library.

[0069] In one embodiment, before performing keyword processing on the text of the successful collection voice sample, the method further includes:

[0070] Performing noise reduction processing on the successful collection voice sample by using an adaptive filter;

[0071] The successful collection voice samples after noise reduction are converted into collection text samples through the preset speech recognition model.

[0072] In this embodiment, the adaptive filter includes the Least Mean Square algorithm, i.e., the LMS algorithm, etc.; the preset speech recognition algorithm includes a hidden Markov model and a vector quantization algorithm of a non-parametric model, and the successful collection voice sample can be converted into a collection text sample through the hidden Markov model, or the successful collection voice sample can be converted into a collection text sample through the vector quantization algorithm of a non-parametric model.

[0073] In one embodiment, the method of converting the successful debt collection voice sample after noise reduction processing into a debt collection text sample by using a preset speech recognition model includes:

[0074] Extracting speech sequences of different tones from the successful debt collection speech samples by using the preset speech recognition algorithm;

[0075] Calculating the timbre similarity between the speech sequences of different timbres and the preset speech features in the speech feature library respectively;

[0076] Determining a first voice feature and a second voice feature according to the timbre similarity;

[0077] The speech sequence corresponding to the first speech feature and the speech sequence corresponding to the second speech feature are converted into the collection text sample.

[0078] In this embodiment, the speech sequences of different timbres refer to the words spoken by the debt collector and the customer. For example, during the debt collection process, the debt collector and the customer establish a telephone contact channel and store the call information. By extracting the speech series of the debt collector and the customer, the speech sequence of the debt collector and the speech sequence of the customer can be obtained respectively; the speech feature library includes the timbre features of all debt collectors; the first speech feature refers to the voice feature of the debt collector, and the second speech feature refers to the voice feature of the customer.

[0079] In this embodiment, different voice sequences in the collection voice sample are extracted through a preset voice recognition algorithm to obtain two voice sequences, and the timbre features corresponding to the two voice sequences are extracted respectively. The similarities of the two timbre features with the preset voice features in the voice feature library are calculated respectively to obtain two timbre similarities and compare the two similarities. The one with high timbre similarity is taken as the first voice feature, and the one with low timbre similarity is taken as the second voice feature. The voice sequences corresponding to the first voice feature and the second voice feature are respectively converted into the collection sample text through the preset voice recognition algorithm, and the collection sample text includes the voice information of the collector and the voice information of the customer.

[0080] In this embodiment, by identifying the timbre features in the sample voice of the collection and calculating the similarity with the preset voice features in the voice feature library, the voice information of the collection personnel and the voice information of the customer can be accurately obtained, which provides a basis for subsequent analysis and increases the success rate of collection.

[0081] In one embodiment, before generating an initial collection dialogue library according to the first response dialogue set, the method further includes:

[0082] A plurality of emotion tags are obtained, and an emotion mapping table is constructed according to the plurality of emotion tags and the first response dialogue.

[0083] In this embodiment, the emotion label refers to a label of an emotional state, such as an anger label, a depression label, an anxiety label, etc.; the emotion mapping table refers to a table that records the correspondence between the emotion label and the first response term.

[0084] In one embodiment, after acquiring the collection voice sample set and the corresponding historical collection data from the sample library, the method further includes:

[0085] Classifying the collection voice sample set according to the historical collection data to obtain at least one collection failure voice sample;

[0086] The text of the failed collection voice sample is processed with keywords to obtain keywords of the second willingness to repay, and the second response dialogue is extracted from the failed collection voice sample according to the second keyword, and the second response dialogue of all failed collection voice samples is collected.

[0087] In this embodiment, all the collection voice samples and the historical collection data corresponding to each collection voice sample are obtained from the sample library, and the collection failure voice samples are obtained from the collection voice samples according to the repayment results of the historical collection data. The collection failure voice samples are converted into collection text samples through a pre-trained speech recognition algorithm, and the keywords corresponding to the second repayment willingness in each collection sample text are identified. The similarity of every two collection sample texts is calculated according to the keywords, and every two collection text samples with a similarity greater than a preset threshold are classified into the same category to obtain at least one category of collection text samples. A preset number of response dialogues are extracted from the collection failure voice samples according to the keywords of the second repayment willingness of the same category, and the second response dialogues of all collection failure voice samples are collected.

[0088] In this embodiment, the second response dialogue set refers to the most ineffective response dialogue set in the collection process; the second repayment willingness refers to the degree to which the customer is unable to communicate, such as hanging up the phone, unwillingness to communicate, etc.

[0089] In the implementation scenario of the financial field, for example, the data in the sample library is classified according to the type of collection business (such as credit card collection business, unsecured loan collection business, mortgage loan collection business, etc.), and the collection voice samples and historical collection data corresponding to the three categories of collection business are obtained. For any category of collection business, the corresponding collection voice sample and the corresponding historical collection data are obtained, and the collection voice sample of failed collection is obtained from the collection voice sample according to the historical collection data, and the corresponding keywords of the second repayment willingness in each collection sample text are identified, and the similarity of each two collection sample texts is calculated according to the keywords, and each two collection text samples with a similarity greater than a preset threshold are classified into the same category to obtain at least one category of collection text samples, and a preset number of response dialogues are extracted from the failed collection voice samples according to the keywords of the second repayment willingness of the same category, and the second response dialogues of all failed collection voice samples are collected to obtain the most ineffective second response dialogue set.

[0090] Referring to the above processing of the sample voice text of successful collection, the same scheme is used to process the sample voice of failed collection, and a preset number of second response dialogues are selected from the sample voice text of failed collection as the most ineffective response dialogues.

[0091] It is worth mentioning that, in the present embodiment, all first response dialogue sets are summarized as allowed dialogues, and all second response dialogue sets are summarized as prohibited dialogues to generate an initial collection dialogue library.

[0092] In this embodiment, the participation of repayment willingness is utilized to make the selected response dialogue more accurate. The response dialogue selected based on repayment willingness can use the most accurate response dialogue as the permitted dialogue, which can largely eliminate the situation where the user does not repay, and improve the collection success rate of the collection service.

[0093] S3. Using a preset compliance corpus, the non-compliant text in the initial debt collection dialogue library is revised to obtain a target debt collection dialogue library.

[0094] In this embodiment, the latest laws and regulations as well as the compliance requirements customized within the enterprise are automatically obtained from the Internet in real time through a preset RPA tool. A preset compliance corpus is generated based on the latest laws and regulations as well as the compliance requirements customized within the enterprise. The prohibited words contained in the preset compliance corpus are matched with the allowed words in the initial corpus. When the match is successful, the allowed words are modified to prohibited words to obtain the target collection dialogue library.

[0095] In this embodiment, the self-developed RPA tool refers to a self-developed robotic automation process tool that is used to automatically obtain the latest laws and regulations in real time; laws and regulations refer to the regulations and norms formulated and issued by government departments that constrain the entire society; the compliance requirements customized within the enterprise refer to the internal rules and regulations and compliance standards specified according to the requirements of its own business characteristics.

[0096] In this embodiment, the compliance corpus is formed by automatically acquiring the latest laws, regulations and internal company rules and regulations, and the initial collection dialogue library is continuously corrected using the compliance corpus to obtain the target collection dialogue library, which can reduce the impact on production capacity and customer complaint rate.

[0097] In this embodiment, the first response dialogue and the second response dialogue are summarized, the first response dialogue is used as the allowed dialogue, and the second response dialogue is used as the prohibited dialogue, so as to avoid the situation where the collection personnel use the prohibited dialogue and the user is unwilling to repay, thereby indirectly improving the collection success rate of the collection business; using the first response dialogue to collect debts from customers directly improves the collection success rate of the collection business; through the preset RPA tool, the latest laws and regulations and the compliance requirements customized within the enterprise are automatically obtained from the Internet in real time, and a preset compliance corpus is generated according to the latest laws and regulations and the compliance requirements customized within the enterprise. The compliance corpus is used to correct the initial collection dialogue library, which can avoid the collection personnel from touching the terms that are not allowed by law and reduce the complaint rate.

[0098] S4. Acquire voice information of the object to be collected, extract voice keywords from the voice information, match the target response dialogue from the target collection dialogue library according to the voice keywords, and feed back the target response dialogue to the object to be collected for collection.

[0099] In this embodiment, when the collection personnel receives a business collection request from the superior platform, they obtain the basic information of the object to be collected and the bill to be collected according to the business collection request, establish a connection with the object to be collected (such as a telephone connection), obtain the voice of the object to be collected in real time, and translate the voice of the object to be collected into the object text of the object to be collected, recognize the voice of the object to be collected through a pre-built emotion detection model, obtain the emotion label of the object to be collected, extract the voice keywords of the voice information, match the target response dialogue from the target collection dialogue library according to the emotion label and the voice keyword, and feed back the target response dialogue to the collection personnel for the collection personnel to select the most appropriate target response dialogue for collection, or directly feed back to the object to be collected for collection.

[0100] In this embodiment, the platform includes the collection management system of banks, financial enterprises and other institutions and the collection business allocation and management platform provided by the supervisory agency; the business collection request refers to a specific collection business request sent by the platform to the collection personnel; the basic information includes the name, contact information, employment information, credit information, personal identity information, etc. of the collection object; the bills to be collected include loan information, overdue days, loan type, etc.; the object text refers to the voice of the collection object converted into text through voice recognition technology.

[0101] In a financial implementation scenario, for example, when a debt collector receives a debt collection service from a bank, he obtains name A, age 30, mobile phone number 138xxxx3696, an employee of Company X, and a loan of 100,000 yuan in unsecured loans, which is overdue for 10 days. The debt collector contacts user A by dialing phone number 138xxxx3696. During the communication with the customer, he obtains the voice of customer A, translates the voice of customer A into text, identifies the emotional label of the voice of customer A, obtains 3 allowed scripts and 3 prohibited scripts from the target collection dialogue library based on the text and emotional label of customer A, and selects one of the 3 allowed scripts for collection to avoid duplication of the script with the 3 prohibited scripts.

[0102] In one embodiment, the step of matching the target response dialogue from the target collection dialogue library according to the voice keywords includes:

[0103] The voice information is recognized by a pre-built emotion detection model to obtain an emotion label of the object to be collected;

[0104] The voice information is converted into object text information in real time by the preset voice recognition algorithm, and voice keywords are extracted from the object text information;

[0105] The target response dialogue is obtained by matching the emotion tag and the voice keyword from the target collection dialogue library.

[0106] In this embodiment, the pre-built emotion detection model can be a convolutional neural network. After obtaining the voice of the object to be collected, the voice of the object to be collected is detected by the emotion detection model to obtain the emotion label of the object to be collected, and the voice of the object to be collected is converted into object text in real time, and voice keywords are extracted from the object text. The target response dialogue is obtained by matching the emotion label and the voice keywords from the target collection dialogue library.

[0107] In this embodiment, by obtaining successful collection voice samples as the basis for analysis, the collection success rate of the collection business can be improved; by classifying the collection voice samples according to the type of collection business, and analyzing the collection sample voice based on each category, it is highly targeted, which improves the accuracy of the subsequent target dialogue and improves the collection success rate of the collection business; by taking into account the willingness to repay, the selected response dialogue can be more accurate, and the response dialogue selected according to the willingness to repay can use the most accurate response dialogue as the allowed dialogue, which is more likely to exclude the situation where the user does not repay, and improve the collection success rate of the collection business; by summarizing the first response dialogue and the second response dialogue, the first response dialogue is used as the allowed dialogue, and the second response dialogue is used as the allowed dialogue. Second, response dialogues are used as prohibited dialogues to prevent debt collectors from using prohibited dialogues, which may lead to users refusing to repay the debt, and indirectly improve the collection success rate of the debt collection business; the first response dialogue is used to collect debts from customers, which directly improves the collection success rate of the debt collection business; the preset RPA tool is used to automatically obtain the latest laws and regulations and the compliance requirements customized within the enterprise from the Internet in real time, and a preset compliance corpus is generated based on the latest laws and regulations and the compliance requirements customized within the enterprise. The compliance corpus is used to correct the initial collection dialogue library, which can prevent debt collectors from touching on clauses that are not allowed by law and reduce the complaint rate; through the participation of emotional tags, more effective response dialogues can be obtained for debt collection, which greatly improves the collection success rate of the debt collection business.

[0108] Reference Figure 2 As shown, it is a functional module diagram of the speech configuration device 100 based on the debt collection dialogue library of the present invention.

[0109] The classification module 110 is used to obtain a collection voice sample set and corresponding historical collection data from a sample library, classify the collection voice sample set according to the historical collection data, and obtain at least one successful collection voice sample.

[0110] In one embodiment, before acquiring the collection voice sample set and the corresponding historical collection data from the sample library, the method further includes:

[0111] All the collection voice samples in the sample library are classified according to the collection business type corresponding to the collection voice to obtain at least one category of collection voice samples.

[0112] The extraction module 120 is used to perform keyword processing on the text of the successful collection voice sample to obtain keywords of the first repayment willingness, extract the first response dialogue from the successful collection voice sample according to the keywords, collect the first response dialogue of all successful collection voice samples, and obtain an initial collection dialogue library.

[0113] In one embodiment, before performing keyword processing on the text of the successful collection voice sample, the method further includes:

[0114] Performing noise reduction processing on the successful collection voice sample by using an adaptive filter;

[0115] The successful collection voice samples after noise reduction are converted into collection text samples through a preset speech recognition algorithm.

[0116] In one embodiment, the method of converting the successful debt collection voice sample after noise reduction processing into a debt collection text sample by using a preset voice recognition algorithm includes:

[0117] Extracting speech sequences of different tones from the successful debt collection speech samples by using the preset speech recognition algorithm;

[0118] Calculating the timbre similarity between the speech sequences of different timbres and the preset speech features in the speech feature library respectively;

[0119] Determining a first voice feature and a second voice feature according to the timbre similarity;

[0120] The speech sequence corresponding to the first speech feature and the speech sequence corresponding to the second speech feature are converted into the collection text sample.

[0121] In one embodiment, before generating an initial collection dialogue library according to the first response dialogue set, the method further includes:

[0122] A plurality of emotion tags are obtained, and an emotion mapping table is constructed according to the plurality of emotion tags and the first response dialogue.

[0123] In one embodiment, after acquiring the collection voice sample set and the corresponding historical collection data from the sample library, the method further includes:

[0124] Classifying the collection voice sample set according to the historical collection data to obtain at least one collection failure voice sample;

[0125] The text of the failed collection voice sample is processed with keywords to obtain a second keyword of the second repayment willingness, and a second response dialogue is extracted from the failed collection voice sample according to the second keyword, and the second response dialogue of all failed collection voice samples is collected.

[0126] A revision module 130, configured to revise the non-compliant text in the initial collection dialogue library using a preset compliant corpus to obtain a target collection dialogue library;

[0127] The feedback module 140 is used to obtain voice information of the object to be collected, extract voice keywords from the voice information, match the target response dialogue from the target collection dialogue library according to the voice keywords, and feed back the target response dialogue to the object to be collected for collection.

[0128] In one embodiment, the step of matching the target response dialogue from the target collection dialogue library according to the voice keywords includes:

[0129] The voice information is recognized by a pre-built emotion detection model to obtain an emotion label of the object to be collected;

[0130] The voice information is converted into object text information in real time by the preset voice recognition algorithm, and voice keywords are extracted from the object text information;

[0131] The target response dialogue is obtained by matching the emotion tag and the voice keyword from the target collection dialogue library.

[0132] Reference Figure 3 FIG. 1 is a schematic diagram of a preferred embodiment of an electronic device 1 of the present invention.

[0133] The electronic device 1 includes but is not limited to: a memory 11, a processor 12, a display 13 and a network interface 14. The electronic device 1 is connected to a network through the network interface 14 to obtain raw data. The network may be a wireless or wired network such as an intranet, the Internet, a global system of mobile communication (GSM), a wideband code division multiple access (WCDMA), a 4G network, a 5G network, Bluetooth, Wi-Fi, a call network, etc.

[0134] Among them, the memory 11 includes at least one type of readable medium, and the readable medium includes flash memory, hard disk, multimedia card, card-type memory (for example, SD or DX memory, etc.), random access memory (RAM), static random access memory (SRAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), programmable read-only memory (PROM), magnetic memory, disk, optical disk, etc. In some embodiments, the memory 11 can be an internal storage unit of the electronic device 1, such as a hard disk or memory of the electronic device 1. In other embodiments, the memory 11 can also be an external storage device of the electronic device 1, such as a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (FlashCard), etc. equipped with the electronic device 1. Of course, the memory 11 can also include both the internal storage unit of the electronic device 1 and its external storage device. In this embodiment, the memory 11 is generally used to store the operating system and various application software installed on the electronic device 1, such as the program code of the speech configuration 10 of the collection dialogue library, etc. In addition, the memory 11 can also be used to temporarily store various types of data that have been output or are to be output.

[0135] In some embodiments, the processor 12 may be a central processing unit (CPU), a controller, a microcontroller, a microprocessor, or other data processing chip. The processor 12 is generally used to control the overall operation of the electronic device 1, such as performing control and processing related to data interaction or communication. In this embodiment, the processor 12 is used to run the program code stored in the memory 11 or process data, such as running the program code of the speech configuration 10 of the collection dialogue library.

[0136] The display 13 may be referred to as a display screen or a display unit. In some embodiments, the display 13 may be an LED display, a liquid crystal display, a touch-sensitive liquid crystal display, and an organic light-emitting diode (OLED) touch device. The display 13 is used to display information processed in the electronic device 1 and to display a visual working interface, such as displaying the results of data statistics.

[0137] The network interface 14 may optionally include a standard wired interface, a wireless interface (such as a WI-FI interface), and the network interface 14 is generally used to establish a communication connection between the electronic device 1 and other electronic devices.

[0138] Figure 3Only the electronic device 1 having components 11 - 14 and a speech configuration 10 of a debt collection dialogue library is shown, but it should be understood that it is not required to implement all of the components shown, and more or fewer components may be implemented instead.

[0139] Optionally, the electronic device 1 may further include a user interface, which may include a display (Display), an input unit such as a keyboard (Keyboard), and the optional user interface may also include a standard wired interface and a wireless interface. Optionally, in some embodiments, the display may be an LED display, a liquid crystal display, a touch-sensitive liquid crystal display, and an organic light-emitting diode (Organic Light-Emitting Diode, OLED) touch device, etc. Among them, the display may also be appropriately referred to as a display screen or a display unit, which is used to display information processed in the electronic device 1 and to display a visual user interface.

[0140] The electronic device 1 may further include a radio frequency (RF) circuit, a sensor, an audio circuit, etc., which will not be described in detail here.

[0141] In the above embodiment, when the processor 12 executes the speech configuration 10 of the debt collection dialogue library stored in the memory 11, the following steps can be implemented:

[0142] Acquire a collection voice sample set and corresponding historical collection data from a sample library, classify the collection voice sample set according to the historical collection data, and obtain at least one successful collection voice sample;

[0143] Perform keyword processing on the text of the successful collection voice sample to obtain a keyword of the first repayment willingness, extract a first response dialogue from the successful collection voice sample according to the keyword, collect the first response dialogues of all successful collection voice samples, and obtain an initial collection dialogue library;

[0144] Using a preset compliance corpus to revise the non-compliant text in the initial collection dialogue library to obtain a target collection dialogue library;

[0145] Acquire the voice information of the object to be collected, extract the voice keywords of the voice information, match the target response dialogue from the target collection dialogue library according to the voice keywords, and feed back the target response dialogue to the object to be collected for collection.

[0146] For a detailed description of the above steps, please refer to Figure 1 Description of a flowchart of an embodiment of an asynchronous credit granting method.

[0147] In addition, an embodiment of the present invention further proposes a computer-readable medium, which may be non-volatile or volatile. The computer-readable medium may be any one or any combination of a hard disk, a multimedia card, an SD card, a flash memory card, an SMC, a read-only memory (ROM), an erasable programmable read-only memory (EPROM), a portable compact disk read-only memory (CD-ROM), a USB memory, etc. The computer-readable medium includes a data storage area and a program storage area, the data storage area stores data created according to the use of the blockchain node, and the program storage area stores the speech configuration 10 of the collection dialogue library. When the speech configuration 10 of the collection dialogue library is executed by the processor, the following operations are implemented:

[0148] Acquire a collection voice sample set and corresponding historical collection data from a sample library, classify the collection voice sample set according to the historical collection data, and obtain at least one successful collection voice sample;

[0149] Perform keyword processing on the text of the successful collection voice sample to obtain a keyword of the first repayment willingness, extract a first response dialogue from the successful collection voice sample according to the keyword, collect the first response dialogues of all successful collection voice samples, and obtain an initial collection dialogue library;

[0150] Using a preset compliance corpus to revise the non-compliant text in the initial collection dialogue library to obtain a target collection dialogue library;

[0151] Acquire the voice information of the object to be collected, extract the voice keywords of the voice information, match the target response dialogue from the target collection dialogue library according to the voice keywords, and feed back the target response dialogue to the object to be collected for collection.

[0152] The specific implementation of the computer-readable medium of the present invention is substantially the same as the specific implementation of the asynchronous credit granting method described above, and will not be described in detail herein.

[0153] In another embodiment, the asynchronous credit granting method provided by the present invention can further ensure the privacy and security of all the above data, and all the above data can also be stored in a blockchain node. For example, feature dimensions and feature embeddings can all be stored in blockchain nodes.

[0154] It should be noted that the blockchain referred to in the present invention is a new application model of computer technologies such as distributed data storage, peer-to-peer transmission, consensus mechanism, encryption algorithm, etc. Blockchain is essentially a decentralized database, a string of data blocks generated by cryptographic methods. Each data block contains a batch of network transaction information, which is used to verify the validity of its information (anti-counterfeiting) and generate the next block. Blockchain can include the blockchain underlying platform, platform product service layer, and application service layer.

[0155] It should be noted that the serial numbers of the above embodiments of the present invention are only for description and do not represent the advantages and disadvantages of the embodiments. And the terms "including", "comprising" or any other variants thereof in this article are intended to cover non-exclusive inclusion, so that a process, device, article or method including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, device, article or method. In the absence of further restrictions, an element defined by the sentence "including a ..." does not exclude the presence of other identical elements in the process, device, article or method including the element.

[0156] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above embodiments can be implemented by means of software plus a necessary general hardware platform, and of course by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art can be embodied in the form of a software product, which is stored in a medium (such as ROM / RAM, magnetic disk, optical disk) as described above, and includes a number of instructions for a terminal device (which can be a mobile phone, a computer, an electronic device, or a network device, etc.) to execute the methods described in each embodiment of the present invention.

[0157] The above are only preferred embodiments of the present invention, and are not intended to limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made using the contents of the present invention's specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection of the present invention.

Claims

1. A method for configuring words based on a debt collection dialogue library, characterized in that: The method comprises: Acquire a collection voice sample set and corresponding historical collection data from a sample library, classify the collection voice sample set according to the historical collection data, and obtain at least one successful collection voice sample; Perform keyword processing on the text of the successful collection voice sample to obtain a keyword of the first repayment willingness, extract a first response dialogue from the successful collection voice sample according to the keyword, collect the first response dialogues of all successful collection voice samples, and obtain an initial collection dialogue library; Using a preset compliance corpus to revise the non-compliant text in the initial collection dialogue library to obtain a target collection dialogue library; Acquire the voice information of the object to be collected, extract the voice keywords of the voice information, match the target response dialogue from the target collection dialogue library according to the voice keywords, and feed back the target response dialogue to the object to be collected for collection.

2. The method for configuring speech based on the debt collection dialogue library according to claim 1, characterized in that: Before acquiring the collection voice sample set and the corresponding historical collection data from the sample library, the method further includes: All the collection voice samples in the sample library are classified according to the collection business type corresponding to the collection voice to obtain at least one category of collection voice samples.

3. The method for configuring speech based on the debt collection dialogue library according to claim 1, characterized in that: Before performing keyword processing on the text of the successful collection voice sample, the method further includes: Performing noise reduction processing on the successful collection voice sample by using an adaptive filter; The successful collection voice samples after noise reduction are converted into collection text samples through a preset speech recognition algorithm.

4. The method for configuring speech based on the debt collection dialogue library according to claim 3, characterized in that: The method of converting the successful debt collection voice sample after noise reduction processing into a debt collection text sample by using a preset voice recognition algorithm includes: Extracting speech sequences of different tones from the successful debt collection speech samples by using the preset speech recognition algorithm; Calculating the timbre similarity between the speech sequences of different timbres and the preset speech features in the speech feature library respectively; Determining a first voice feature and a second voice feature according to the timbre similarity; The speech sequence corresponding to the first speech feature and the speech sequence corresponding to the second speech feature are converted into the collection text sample.

5. The method for configuring speech based on the debt collection dialogue library according to claim 4, characterized in that: Before generating the initial debt collection dialogue library according to the first response dialogue set, the method further includes: A plurality of emotion tags are obtained, and an emotion mapping table is constructed according to the plurality of emotion tags and the first response dialogue.

6. The method for configuring speech based on the debt collection dialogue library according to claim 1, characterized in that: After acquiring the collection voice sample set and the corresponding historical collection data from the sample library, the method further includes: Classifying the collection voice sample set according to the historical collection data to obtain at least one collection failure voice sample; The text of the failed collection voice sample is processed with keywords to obtain a second keyword of the second repayment willingness, and a second response dialogue is extracted from the failed collection voice sample according to the second keyword, and the second response dialogue of all failed collection voice samples is collected.

7. The method for configuring speech based on the debt collection dialogue library according to claim 1, characterized in that: The step of matching the target response dialogue from the target collection dialogue library according to the voice keywords includes: The voice information is recognized by a pre-built emotion detection model to obtain an emotion label of the object to be collected; The voice information is converted into object text information in real time by the preset voice recognition algorithm, and voice keywords are extracted from the object text information; The target response dialogue is obtained by matching the emotion tag and the voice keyword from the target collection dialogue library.

8. A device for configuring speech based on a debt collection dialogue library, characterized in that: The device comprises: A classification module, used to obtain a collection voice sample set and corresponding historical collection data from a sample library, classify the collection voice sample set according to the historical collection data, and obtain at least one successful collection voice sample; An extraction module is used to perform keyword processing on the text of the successful collection voice sample to obtain a keyword of the first repayment willingness, extract a first response dialogue from the successful collection voice sample according to the keyword, collect the first response dialogues of all successful collection voice samples, and obtain an initial collection dialogue library; A revision module, used to revise the non-compliant text in the initial collection dialogue library using a preset compliant corpus to obtain a target collection dialogue library; The feedback module is used to obtain voice information of the object to be collected, extract voice keywords from the voice information, match the target response dialogue from the target collection dialogue library according to the voice keywords, and feed back the target response dialogue to the object to be collected for collection.

9. An electronic device, characterized in that: The electronic device comprises: at least one processor; and, a memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the method for configuring speech based on the collection dialogue library as described in any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the method for configuring speech based on a collection dialogue library as described in any one of claims 1 to 7 is implemented.

Images