WO2020024389A1 - 逾期账单智能催收方法、装置、计算机设备及存储介质 - Google Patents

逾期账单智能催收方法、装置、计算机设备及存储介质 Download PDF

Info

Publication number
WO2020024389A1
WO2020024389A1 PCT/CN2018/106251 CN2018106251W WO2020024389A1 WO 2020024389 A1 WO2020024389 A1 WO 2020024389A1 CN 2018106251 W CN2018106251 W CN 2018106251W WO 2020024389 A1 WO2020024389 A1 WO 2020024389A1
Authority
WO
WIPO (PCT)
Prior art keywords
overdue
collection
voice
customer
call
Prior art date
Application number
PCT/CN2018/106251
Other languages
English (en)
French (fr)
Inventor
许永夫
Original Assignee
平安科技(深圳)有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 平安科技(深圳)有限公司 filed Critical 平安科技(深圳)有限公司
Publication of WO2020024389A1 publication Critical patent/WO2020024389A1/zh

Links

Images

Classifications

    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06QINFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q40/00Finance; Insurance; Tax strategies; Processing of corporate or income taxes
    • G06Q40/03Credit; Loans; Processing thereof

Definitions

  • the present application relates to the field of information processing technology, and in particular, to a method, device, computer equipment, and storage medium for intelligent collection of overdue bills.
  • the embodiments of the present application provide an intelligent collection method for overdue bills, a device, a computer device, and a storage medium to solve the problems of low collection efficiency and high cost in the current manual collection.
  • An overdue bill intelligent collection method includes:
  • the original case is screened according to the collection rules to obtain an entry case, where the entry case carries overdue customer information, and the overdue customer information includes the contact number of the overdue customer;
  • An overdue bill intelligent collection device includes:
  • the reminder case acquisition module is used to screen the original case according to the collection rules to obtain the reminder case.
  • the reminder case carries overdue customer information, and the overdue customer information includes the contact number of the overdue customer;
  • the overdue customer list acquisition module is used to classify the overdue customer information of the reminder case according to the customer risk classification strategy and obtain the overdue customer list.
  • the overdue customer list includes overdue customer information and corresponds to the overdue customer information. Identification of overdue type;
  • a target collection speech acquisition module configured to obtain a collection speech template corresponding to the overdue type identifier, add the overdue customer information to a position corresponding to the collection speech template, and obtain a target collection speech;
  • the overdue collection result acquisition module is used to call the telephone platform to automatically dial the contact phone of the overdue customer and convert the target collection speech to the target collection voice to obtain the overdue collection result of the overdue customer feedback on the target collection voice.
  • a computer device includes a memory, a processor, and computer-readable instructions stored in the memory and executable on the processor.
  • the processor executes the computer-readable instructions, the following steps are implemented:
  • the original case is screened according to the collection rules to obtain an entry case, where the entry case carries overdue customer information, and the overdue customer information includes the contact number of the overdue customer;
  • One or more non-volatile readable storage media storing computer-readable instructions, and when the computer-readable instructions are executed by one or more processors, the one or more processors implement the following steps:
  • the original case is screened according to the collection rules to obtain an entry case, where the entry case carries overdue customer information, and the overdue customer information includes the contact number of the overdue customer;
  • FIG. 1 is an application scenario diagram of a method for intelligent collection of overdue bills in an embodiment of the present application
  • FIG. 2 is a flowchart of a method for intelligent collection of overdue bills in an embodiment of the present application
  • FIG. 3 is a specific flowchart of step S40 in FIG. 2;
  • step S43 in FIG. 3 is a specific flowchart of step S43 in FIG. 3;
  • FIG. 5 is another flowchart of a method for intelligent collection of overdue bills in an embodiment of the present application.
  • FIG. 6 is another flowchart of a method for intelligent collection of overdue bills in an embodiment of the present application.
  • FIG. 7 is a specific flowchart of step S46 in FIG. 6;
  • step S47 in FIG. 6 is a specific flowchart of step S47 in FIG. 6;
  • FIG. 9 is a schematic diagram of an overdue bill intelligent collection device in an embodiment of the present application.
  • FIG. 10 is a schematic diagram of a computer device according to an embodiment of the present application.
  • the method for intelligent collection of overdue bills can be applied in the application environment shown in FIG. 1.
  • the method for intelligent collection of overdue bills is applied to an intelligent collection system.
  • the intelligent collection system includes a client and a server.
  • the client Communicate with the server over the network.
  • the client is also called the client, which refers to the program that provides local services to the client corresponding to the server.
  • Clients can be installed on, but are not limited to, various personal computers, laptops, smartphones, tablets, and portable wearable devices.
  • the server can be implemented by an independent server or a server cluster composed of multiple servers.
  • a method for intelligent collection of overdue bills is provided.
  • the method is applied to the server in FIG. 1 as an example, and includes the following steps:
  • the original case is screened according to the collection rules to obtain the entry case.
  • the entry case carries overdue customer information, and the overdue customer information includes the overdue customer's contact phone number.
  • the collection rules refer to the rules used to determine whether the original case belongs to a case requiring collection.
  • the original case refers to the case requiring repayment stored by the business system.
  • Involved cases refer to cases that meet the collection rules and did not repay on the repayment date.
  • the intelligent collection system obtains the original case sent by the business system, and the original case carries the repayment date of the original case. After the original collection case is obtained, the intelligent collection system uses the collection rules to screen the original cases and selects the Cases with no repayment on the payment date are determined to be cases of reminders.
  • the collection case obtained by the intelligent collection system carries overdue customer information.
  • the overdue customer information refers to the personal information of the overdue customer, including but not limited to the name of the overdue customer, bank card account number, repayment amount, repayment date, and contact phone number. Among them, overdue customers are customers who did not complete their repayment on the repayment date.
  • the business system uses the RSA algorithm to encrypt the original case before sending the original case to the intelligent collection system.
  • the encryption process is specifically that the intelligent collection system randomly selects two larger prime numbers A and B, and calculates a product N of the prime numbers A and the prime numbers B. Then choose a product of the integer e and (A-1) * (B-1) that is coprime, and e is less than (A-1) * (B-1), and finally according to the formula d * e ⁇ 1 (mod (A- 1) (B-1)) Calculate d.
  • step S10 that is, the intelligent collection system screens the original case according to the collection rules, and before the step of obtaining the collection case, the intelligent collection system uses the private key to decrypt the original case.
  • the RSA algorithm is currently the most influential and commonly used public key encryption algorithm, it has high security and can resist most of the cryptographic attacks known so far. Therefore, the RSA algorithm is used to encrypt and decrypt the original case. Can improve the security of data during transmission.
  • the overdue customer information of the urging cases is divided to obtain the overdue customer list.
  • the overdue customer list includes overdue customer information and overdue type identification corresponding to overdue customer information.
  • the customer risk level classification strategy refers to the strategy of classifying customers' risk levels based on at least one of the customer's overdue payment amount, overdue payment days, and historical overdue repayments.
  • the overdue customer list refers to a list that includes overdue customer information obtained by grading entry cases according to the customer risk classification strategy.
  • the overdue customer list is used to store overdue customer information and an overdue type identifier corresponding to the overdue customer information.
  • the overdue customer information includes, but is not limited to, the name of the overdue customer, bank card account number, repayment amount, repayment date, contact phone number, historical overdue payment days and historical overdue payment times.
  • Overdue type identifier refers to the identifier used to uniquely identify the type of overdue customer.
  • the overdue type identifier may be any one or more combinations of a specific character string, a specific number, or a specific letter.
  • overdue customers in this embodiment include, but are not limited to, high-level customers, low-level customers, habitual customers, lost customers, and old customers.
  • high-end customers refer to customers whose overdue repayment days exceed 60 days.
  • Low-level customers refer to customers who have overdue repayment days not exceeding 60 days.
  • Overdue customers refer to customers whose historical overdue repayments have reached 4 times within 6 months. Missing customers refer to customers who could not be reached using the contact phone number in the case.
  • Old Lai customers refer to customers who have not paid back at the designated repayment time.
  • each overdue customer list corresponds to an overdue type identifier.
  • low-level customers correspond to low-level customer type identifier L
  • high-level customers correspond to high-level customer type identifier H
  • habitual customers correspond to habitual customer type identifier F
  • lost customers correspond to lost customer type identifier O
  • old customers correspond.
  • the intelligent collection system After the intelligent collection system obtains the collection case, it will obtain the historical overdue payment days and the historical overdue repayment times corresponding to the overdue customers from the overdue customer information carried in the collection case. Then based on the pre-set customer risk classification strategy, judge the historical overdue payment days and historical overdue repayments corresponding to overdue customers, determine the overdue customer types corresponding to overdue customers in the reminder case, and generate overdue corresponding to different overdue type identifiers. customer list.
  • the overdue customer list includes overdue customer information and overdue type identification, which facilitates subsequent acquisition of a collection call template corresponding to the overdue type identification through the overdue type identification and completes the collection of repayments for overdue customers corresponding to entry cases.
  • S30 Obtain a collection call template corresponding to the overdue type identifier, add overdue customer information to the position corresponding to the collection call template, and obtain a target collection call.
  • the call collection template refers to a pre-set call template in the smart collection system for collection of overdue customers.
  • each collection speech template is associated with an overdue type identifier, so as to find a corresponding collection speech template based on the overdue type identification.
  • the collection call template is stored in the text in the intelligent collection system to facilitate the subsequent addition of overdue customer information to the collection call template to generate a target collection call.
  • the intelligent collection system will select a collection call template corresponding to the overdue type identifier according to the acquired overdue type identifier, and then the name, bank card account number, repayment amount, and repayment in the overdue customer information The date is added to the collection speech template corresponding to the overdue type identifier to obtain the target collection speech.
  • the target collection phrase refers to the collection phrase for overdue customers of a specific type of overdue customers composed of overdue customer information and a collection of speech templates.
  • the intelligent collection system develops different collection collection templates for different types of overdue customers.
  • the collection call corresponding to low-level customers is: Mr./Ms. Hello, hello! Your XX bank credit card ending in xxxx is overdue xxxx yuan, please repay as soon as possible before xxxx days. If there is no objection and agree to repay please confirm ... wait for customers to answer ... If you have any other questions, please call 95511 customer phone to inquire Thank you for your call!
  • the collection call for high-end customers is: Mr./Ms. Hello, hello! Your XX bank credit card ending in xxxx has expired xxxx yuan. In order to avoid affecting your personal credit, please repay as soon as possible before xxxx day. If there is no objection, please confirm ... wait for the customer to answer ... If the customer has questions directly To transfer to a human agent, press 0.
  • the target collection message is acquired through the overdue customer information and the collection message template corresponding to the overdue type identifier, to avoid the case of customer dissatisfaction due to improper use of the collection policy during manual collection, and it is also the step S40 Provide data sources for telephone collection of overdue customers through the telephone platform.
  • S40 Call the phone platform, automatically dial the contact phone number of overdue customers and convert the target collection speech to target collection speech, and obtain the overdue collection results of overdue customers' feedback on the target collection voice.
  • the intelligent collection system After the intelligent collection system obtains the target collection call, it will obtain the contact number of the overdue customer from the overdue customer information in the overdue customer list. After obtaining the target collection call and the contact phone number of the overdue customer, the intelligent collection system will call the built-in phone platform, which integrates the automatic dialing technology. The phone platform will dial the corresponding contact phone number of the overdue customer through the integrated automatic dialing technology. Terminal device. After the terminal device of the overdue customer is connected, the intelligent collection system will convert the target collection speech to the target collection voice and send it to the overdue customer's terminal device through the phone platform so that the overdue customer can hear through the terminal device. The goal is to collect voice, so as to achieve telephone collection of overdue customers.
  • the terminal device is specifically a mobile phone, a smart watch with a call function, or other devices.
  • the target collection speech is converted into collection speech, and the collection speech is used to collect telephone calls for overdue customers, so that the collection of telephone calls is performed on overdue customers in accordance with a preset standard collection speech to prevent collection personnel from calling In the collection process, add your own emotions, which will adversely affect the overdue collection results.
  • the intelligent collection system can also obtain the overdue customer's call voice obtained by the built-in telephone platform, so as to analyze the overdue customer's call voice and obtain the overdue collection result.
  • the intelligent collection system uses the built-in telephone platform to collect telephone calls from overdue customers, realizing the function of completing telephone collection without manual intervention, and improving collection efficiency.
  • the original cases are first filtered through the collection rules, the original cases that do not meet the collection rules are deleted, and the collection cases that meet the collection rules are effectively reduced to effectively reduce the business of the intelligent collection system from business
  • the system obtains data transmission volume, saves data transmission time, and improves collection efficiency.
  • the overdue customer information of the collection cases is divided to obtain the overdue customer list corresponding to the different overdue customer types.
  • the overdue customer list includes the overdue type identifier.
  • a collection call template and overdue customer information corresponding to the overdue type identifier are obtained to generate a target collection call for a specific overdue customer, so that the target collection call is targeted.
  • the overdue customer's contact phone number and target collection utterance in the overdue customer information are sent to the built-in phone platform of the intelligent collection system, and the overdue customers are collected through the phone platform to obtain the overdue collection results, so that the call can be completed without manual intervention. Collection function to improve collection efficiency.
  • step S40 the telephone platform is called to automatically dial the contact phone of the overdue customer and convert the target collection speech to the target collection voice, and obtain the overdue collection of the feedback of the target customer to the target collection voice
  • the telephone platform includes the following steps:
  • the built-in phone platform in this embodiment refers to the phone platform built in the intelligent collection system, and the phone platform integrates the automatic dialing technology.
  • This automatic dialing technology is CTI technology.
  • CTI technology is a technology developed from the traditional computer telephone integration (Computer Telephony Integration) technology.
  • CTI technology can apply computer technology to the telephone system and can automatically call the telephone.
  • the instruction information in the ID is used to identify and process, and to establish a related voice channel connection, to transmit a predetermined recording file to the user and to transfer incoming calls.
  • the smart collection system After the smart collection system obtains the contact phone number of the overdue customer, it will send the contact phone number to the built-in phone platform. After obtaining the contact phone number of the overdue customer, the phone platform will use the integrated automatic dialing technology to detect the overdue customer. Phone number for automatic dialing.
  • the intelligent collection system realizes the function of automatically dialing the terminal device corresponding to the contact phone of the overdue customer through the automatic dialing technology in the built-in telephone platform, and realizes the function of completing phone dialing without manual intervention.
  • S42 Obtain a connection signal fed back by the terminal device corresponding to the contact phone. Based on the connection signal, TTS technology is used to convert the target collection speech to the corresponding target collection speech, and the target collection speech is sent to the terminal device.
  • connection signal refers to a signal fed back to the telephone platform by the terminal device corresponding to the contact phone, and the connection signal can be used to identify whether the terminal device corresponding to the contact phone is connected.
  • TTS technology refers to the technology of converting text information generated by the computer itself or external input into spoken Chinese and outputting it.
  • Target collection speech refers to the collection collection of content in target collection speech.
  • the phone platform automatically dials the terminal device corresponding to the contact phone based on the contact phone of the overdue customer.
  • the terminal device sends a connection signal to the phone platform, indicating that The terminal device corresponding to the contact phone has been connected.
  • the telephone platform After the telephone platform obtains the connection signal, it will feed the connection signal to the server of the intelligent collection system.
  • the server of the intelligent collection system will use TTS technology to convert the target collection speech into the corresponding target collection speech.
  • the collection voice is sent to the terminal device, so that the overdue customers who carry the terminal device can hear the target collection voice to realize intelligent voice collection without manual intervention in the collection process, improve collection efficiency, and help reduce collection costs.
  • TTS technology to transform target collection speech into corresponding target collection speech, standardize the collection collection speech, avoiding the improper collection collection staff's tone during manual collection to cause customer dissatisfaction or the collection collection staff's speech to be unstandard, which will affect the collection collection effect.
  • S43 Receive the call voice feedback from the terminal device, perform semantic analysis on the call voice, and obtain the overdue collection result.
  • the intelligent collection system will receive the call voice from the overdue customer feedback from the terminal device through the telephone platform. After the intelligent collection system obtains the call voice feedback from the terminal device, it will perform semantic analysis on the call voice of the overdue customer to determine whether the call voice of the overdue customer contains the preset “OK”, “OK”, “OK” and “Wait a while,” and other keywords that carry one-way consent to repayment. If the call voice contains keywords such as "OK”, “OK”, “Yes” and "Waiting for Repayment" which carry one-way consent to repayment, it means that the overdue customer agrees to repay, and the overdue collection result is willing to repay .
  • the voice of the call does not contain keywords such as "OK”, “OK”, “Yes”, and "Waiting for Repayment", which carry the one-way consent to repayment, it means that the overdue customer does not agree to the repayment and the overdue collection result is not Willing to repay.
  • Step S41-Step S43 the intelligent collection system automatically dials the terminal device corresponding to the contact phone of the overdue customer through the built-in telephone platform. After the dialing is successful, the target collection speech is converted into the target collection voice by TTS technology and sent to the overdue The corresponding terminal device of the customer completes the dialogue with the overdue customer and obtains the overdue collection result, achieving the effect of intelligent overdue bill collection. No manual collection is required, reducing the labor cost of the collection process and improving collection efficiency.
  • step S43 a semantic analysis is performed on the call voice to obtain an overdue collection result, which specifically includes the following steps:
  • S431 Scan the call voice to determine whether the call voice contains a preset target keyword.
  • the target keywords refer to keywords preset in the intelligent collection system for identifying whether overdue customers agree to repay. That is, the target keyword is a keyword carrying one-way consent to repayment.
  • the smart collection system After the smart collection system obtains the voice of the call sent by the terminal device of the overdue customer through the phone platform, it will scan the voice of the voice according to the keyword recognition technology in the smart collection system to obtain the voice characteristics of the voice of the call. Determine whether there is a word matching the target keyword in the call voice.
  • the keyword recognition technology in this embodiment uses a pre-trained HMM (Hidden Markov Model Hidden Markov Model) to perform target keyword recognition on the call voice.
  • the pre-trained HMM model is pre-trained and stored in the intelligent collection system, which is used to identify whether there are "good", "OK", “may", and "wait for it” in the call voice.
  • Word model In this embodiment, a pre-trained HMM model is used to identify whether a target keyword exists in the call voice, which can improve the recognition accuracy rate.
  • the intelligent collection system uses voice-to-text technology to convert the voice of a call into text. Then, the text corresponding to the voice of the speech is segmented by using a word segmentation tool, and words such as stop words (prepositions, prepositions and pronouns), symbols, numbers, and letters are removed to obtain words to be identified.
  • the word to be identified refers to a word obtained after removing stop words, symbols, numbers, and letters from the text corresponding to the call voice.
  • the similarity algorithm is used to calculate the similarity between the recognized words and the target keywords. If the similarity reaches a preset value, it is determined that the target keyword is included in the call voice. among them.
  • the similarity algorithm in this embodiment includes, but is not limited to, a cosine similarity algorithm.
  • the consent sign refers to the sign used to indicate that the overdue customer agrees to repay.
  • the call voice contains the target keywords such as "OK”, “OK”, “Yes”, and “wait for repayment”
  • the result of overdue collection was consent to repayment.
  • the intelligent collection system determines that the target keyword is included in the call voice, it will mark the overdue collection result of the overdue customer with the corresponding consent mark, such as "OK” or "1” .
  • the disagreement sign refers to a sign used to indicate that the overdue customer does not agree with the repayment.
  • the target keywords such as "OK”, “OK”, “Yes”, and “wait for repayment”
  • the overdue collection result obtained afterwards is disagreement on repayment.
  • the intelligent collection system determines that the target keyword is not included in the call voice, it will mark the overdue collection result of the overdue customer with the corresponding disagreement mark, such as "no" or "0 ".
  • step S431-step S433 the intelligent collection system scans the call voice to determine whether the call voice contains the target keywords, determines whether the overdue customer agrees to repay, and achieves the purpose of realizing intelligent collection of overdue bills without manual intervention.
  • the intelligent collection system will match the corresponding consent mark or disagree mark to the overdue collection result, which is convenient for the intelligent collection system to Agree or disagree to perform the corresponding subsequent steps.
  • the intelligent collection system will also send corresponding short messages and emails in the form of short messages and emails after the phone communication with overdue customers is completed and the overdue collection result is agreed to repayment.
  • the overdue customer list also includes the email addresses of overdue customers.
  • the method of intelligent overdue bill collection also includes:
  • the collection information template refers to an information template that is set in advance and stored in the intelligent collection system to prompt overdue customers to pay on time.
  • the collection information template is: XXX Mr./Mrs. Hello! According to the telephone communication, you have confirmed that your credit card ending with Ping An Bank xxxx will complete the repayment amount xxxx yuan before xxxx days; if you have any other questions, please call 95511 customer phone to ask, thank you for your call!
  • Effective collection information refers to collection information for specific overdue customers obtained by adding overdue customer information to the collection information template.
  • the intelligent collection system will obtain a collection information template and overdue customer information, and add the name, bank card account number, repayment amount, and repayment date in the overdue customer information to the collection information template. To get effective collection information for that particular overdue customer.
  • S45 Send the effective collection information to the corresponding terminal device through the contact phone of the overdue customer or the email address of the overdue customer.
  • the intelligent collection system After the intelligent collection system obtains the effective collection information, it will send the effective collection information to the terminal device corresponding to the contact phone of the overdue customer through the contact phone of the overdue customer.
  • the overdue customer list also includes the email address of the overdue customer
  • the intelligent collection system will obtain the corresponding email address of the overdue customer from the overdue customer list, and then send the effective collection information to the overdue customer's mailbox by mail.
  • the terminal device corresponding to the address. Sending effective collection information to overdue customers by SMS or email can further ensure that overdue customers can receive and view the effective collection information and improve the quality of collection.
  • the smart collection system after the smart collection system obtains the overdue collection result with the disagreement sign, it will further confirm the situation of the overdue customer. First, it is confirmed whether the call voice is from the overdue customer himself. If it is overdue, If the customer is in person, follow-up steps will be executed; if it is not the overdue customer, it will be handed over to the offline follow-up staff for collection. Among them, the follow-up staff refers to the staff who, when finding overdue customers themselves, face overdue customers to collect overdue loans. As shown in FIG. 6, after the step of obtaining overdue collection results in step S43, the method for intelligent overdue bill collection further includes:
  • voiceprint recognition is also called speaker recognition, and the identity of the speaker is determined by identifying the voiceprint of the speaker.
  • step S47 is executed to perform emotion recognition on the voice of the call and obtain the emotion recognition result of the overdue customer.
  • the intelligent collection system determines that the speaker of the call voice is the overdue customer, further analysis of the call voice is required to determine the emotion of the overdue customer when receiving the call collection, so that subsequent artificial agents can communicate with the overdue customer by telephone.
  • Steps S46 to S47 After the overdue collection result carrying the disagreement sign is obtained, the intelligent collection system will first identify the voice of the overdue customer's call voice to confirm whether the call voice comes from himself; if the call voice comes from the overdue client himself , Emotion recognition is performed on the voice of the call, and the emotion recognition result is obtained, which is convenient for the artificial agent to adopt an appropriate collection strategy to conduct telephone communication with overdue customers and reduce the customer complaint rate.
  • step S46 performs voiceprint recognition on the call voice to determine whether the speaker of the call voice is the overdue customer himself, and specifically includes the following steps:
  • S461 Preprocess the call voice to obtain the voice to be recognized.
  • the voice to be recognized refers to the voice obtained by preprocessing the voice of the call. Specifically, before performing voiceprint recognition on a call voice, it is necessary to preprocess the call voice, remove noise in the call voice (including mute and environmental noise in the voice call), and obtain the voice to be recognized.
  • the pre-processing process specifically includes: first, pre-emphasis processing of the call voice; then framing and windowing the pre-emphasis processing call voice; and finally, using endpoint detection technology to process the call voice to remove the call Noise in speech to get the speech to be recognized.
  • S462 Perform speech feature extraction on the speech to be recognized, and obtain the speech feature to be identified corresponding to the speech to be identified.
  • the speech to be recognized is processed by a fast Fourier transform, a logarithmic operation, and a discrete cosine transform to obtain the characteristics of the speech to be recognized.
  • the feature to be recognized is an MFCC feature.
  • the voiceprint recognition model is used to perform voiceprint recognition on the voice features to be recognized and the original voice features corresponding to the overdue customer, and to determine whether the speaker of the call voice is the overdue customer himself.
  • the voiceprint recognition model refers to a model that is pre-trained and stored in an intelligent collection system to identify whether a speaker of a call voice is an overdue customer himself.
  • voiceprint recognition models including but not limited to GMM-UBM (Gaussian mixture model-universal background model) model and i-vector (identity-vector, identity authentication vector) model.
  • GMM-UBM Global System for Mobile Communications
  • i-vector identity-vector, identity authentication vector
  • the voiceprint recognition model in this embodiment uses an i-vector model as the voiceprint recognition model.
  • the original voice feature refers to the voice feature corresponding to the speaking voice of the overdue customer stored in advance by the intelligent collection system.
  • the corresponding original voice feature can be obtained, and the original voice feature is stored in association with the overdue customer information so that It can be called directly in the subsequent voiceprint recognition without corresponding processing during the recognition process to improve the recognition efficiency.
  • the intelligent collection system will search for the original voice features corresponding to the overdue customer information pre-stored in the intelligent collection system based on the customer information of the overdue customers. Then, input the unrecognized voice features and original voice features of the overdue customer into a pre-trained voiceprint recognition model, and obtain the i-vector vectors corresponding to the unrecognized voice features and the original voice features; and then calculate the unrecognized voice features.
  • the spatial distance of the i-vector vector corresponding to the original voice feature If the spatial distance is greater than the distance threshold set according to the actual situation, it can be determined that the speaker of the call voice is the overdue customer himself.
  • the space vector distance in this embodiment is calculated using a cosine distance calculation formula.
  • Step S461-Step S463 by acquiring the voice feature to be recognized corresponding to the call voice, and using the voiceprint recognition model to identify the voice feature to be recognized and the original voice feature corresponding to the overdue customer, determine whether the speaker of the call voice is the overdue customer himself.
  • the overdue customer it means that the overdue customer is unwilling to repay, and the subsequent steps need to perform emotional recognition on the voice of the call, so that the agent can use manual communication to understand the reason why the overdue customer did not repay.
  • step S47 a pre-stored emotion recognition model is called for emotion recognition of the call voice, and the emotion recognition result corresponding to the overdue collection result is obtained, which specifically includes the following steps:
  • the emotion recognition model is a model that is pre-trained and stored in the intelligent collection system for identifying the emotions of overdue customers.
  • the emotion recognition model in this embodiment uses a vector machine-based emotion recognition model.
  • the emotions to be confirmed include, but are not limited to, emotions such as happiness, anger, sadness, annoyance and calmness when the speaker identified by the emotion recognition model speaks.
  • the intelligent collection system inputs the voice feature to be recognized into a pre-trained emotion recognition model, and performs emotion recognition on the voice feature to be recognized to Obtain the emotional scores of the various to-be-confirmed emotions of the overdue customer corresponding to the call voice.
  • the emotion recognition model in this embodiment After the emotion recognition model in this embodiment obtains the features to be recognized, it will separately calculate the emotion scores for the emotions of happiness, anger, sadness, annoyance, and calmness waiting for recognition in the recognized features, and obtain the corresponding scores for different emotions to be recognized. Emotional score.
  • the computational complexity of using a vector machine-based emotion recognition model is small, and the final result can be determined according to a small number of support vectors. It helps to catch key samples and remove redundant samples during training. It has good robustness.
  • the emotion recognition result refers to the calculation of the emotion score carried by the emotion to be confirmed, and obtains a result that can finally express the emotion of overdue customers when they are collected.
  • the emotion recognition model compares the emotion scores carried by the emotions to be confirmed, and outputs the emotion to be confirmed with the largest emotion score as the final emotion recognition result.
  • the intelligent collection system will send the emotion recognition result and the corresponding overdue customer information to the artificial agent, so that the artificial agent can use the appropriate collection strategy based on the emotion recognition result to conduct telephone communication to the overdue customer.
  • Step S471-Step S472 input the speech features to be identified into the emotion recognition model, and obtain the emotions to be confirmed corresponding to the speech features to be identified; and obtain the emotion recognition results of overdue customers according to the emotion score carried by the emotions to be confirmed.
  • Use the emotion recognition model to intelligently judge the emotions of overdue customers, improve the accuracy and speed of emotion recognition, and facilitate the artificial agents to use appropriate collection strategies based on emotion recognition results to conduct telephone communication with overdue customers. .
  • the original cases are first filtered through the collection rules, the original cases that do not meet the collection rules are deleted, and the collection cases that meet the collection rules are obtained. Then, according to the pre-set customer risk classification strategy of the intelligent collection system, the overdue customer information of the collection cases is divided, and the overdue customer list corresponding to different overdue customer types is obtained, which facilitates the subsequent use of different collection tactics and overdue for different overdue customer types.
  • Customers communicate by phone.
  • a collection call template and overdue customer information corresponding to the overdue type identifier are obtained to generate a target collection call for a specific overdue customer.
  • the auto-dial technology of the telephone platform is used to collect overdue customers by telephone. To obtain overdue collection results, determine whether overdue customers have repayment willingness, improve collection efficiency and reduce collection costs.
  • the effective collection information is sent to the terminal device of the overdue customer; if the overdue collection result carries the disagreement mark, the voiceprint recognition and emotional recognition of the overdue customer's call voice will be performed, and eventually The emotion recognition result is fed back to the artificial agent, so that the artificial agent collects overdue customers based on the emotion recognition result, and reduces the complaint rate caused by improper use of the collection strategy.
  • an overdue bill intelligent collection device is provided.
  • the overdue bill intelligent collection device corresponds to the overdue bill intelligent collection method in the above embodiment.
  • the search-only search device includes a case collection module 10, an overdue customer list acquisition module 20, a target collection speech acquisition module 30, and an overdue collection result acquisition module 40.
  • the detailed description of each function module is as follows:
  • the reminder case acquisition module 10 is configured to screen the original case according to the collection rules to obtain the reminder case.
  • the reminder case carries overdue customer information, and the overdue customer information includes the overdue customer's contact phone number.
  • the overdue customer list acquisition module 20 is used to classify overdue customer information in the reminder case according to the customer risk classification strategy, and obtain the overdue customer list.
  • the overdue customer list includes overdue customer information and an overdue type identifier corresponding to the overdue customer information.
  • the target collection speech acquisition module 30 is configured to obtain a collection speech template corresponding to the overdue type identifier, add overdue customer information to a position corresponding to the collection speech template, and obtain a target collection speech.
  • the overdue collection result acquisition module 40 is used to call the telephone platform to automatically dial the contact phone of overdue customers and convert the target collection speech to the target collection voice to obtain the overdue collection results of overdue customers' feedback on the target collection voice.
  • the overdue collection result acquisition module 40 includes an automatic dialing unit 41, a target voice acquisition and transmission unit 42, and an overdue collection result acquisition unit 43.
  • the automatic dialing unit 41 is configured to call the telephone platform to automatically dial the contact phone of the overdue customer based on the contact phone of the overdue customer.
  • the target voice acquisition and sending unit 42 is configured to obtain a connection signal fed back from the terminal device corresponding to the contact phone. Based on the connection signal, the TTS technology is used to convert the target collection speech to the corresponding target collection speech, and send the target collection speech to the terminal. device.
  • the overdue collection result acquisition unit 43 is configured to receive a call voice feedback from the terminal device, perform semantic analysis on the call voice, and obtain an overdue collection result.
  • the overdue collection result acquisition unit 43 includes a target keyword judgment unit 431, a first overdue collection result acquisition unit 432, and a second overdue collection result acquisition unit 433.
  • a target keyword judging unit 431 is configured to scan a call voice and determine whether the call voice includes a preset target keyword.
  • the first overdue collection result obtaining unit 432 is configured to obtain the overdue collection result carrying a consent mark if the call voice contains a target keyword.
  • the second overdue collection result acquisition unit 433 is configured to obtain the overdue collection result carrying a disagreement identifier if the target keyword is not included in the call voice.
  • the overdue customer list also includes the email address of the overdue customer.
  • the overdue bill intelligent collection device further includes an effective collection information acquisition unit 44, an effective collection information transmission unit 45, a voiceprint recognition unit 46, and an emotion recognition unit 47.
  • the effective collection information acquisition unit 44 is configured to obtain valid collection information based on the collection information template and overdue customer information if the overdue collection result carries a consent identifier.
  • the effective collection information sending unit 45 is configured to send the effective collection information to the corresponding terminal device through the contact phone of the overdue customer or the email address of the overdue customer.
  • the voiceprint recognition unit 46 is configured to perform voiceprint recognition on the call voice if the overdue collection result carries a disagreement identifier, and determine whether the speaker of the call voice is the overdue customer himself.
  • the emotion recognition unit 47 is configured to: if the speaker of the call voice is the overdue customer himself, call a pre-stored emotion recognition model to perform emotion recognition on the call voice, and obtain an emotion recognition result corresponding to the overdue collection result.
  • the voiceprint recognition unit 46 includes a call voice pending unit 461, a voice feature extraction unit 462, and a voiceprint recognition model judgment unit 463.
  • the call voice pending unit 461 is configured to preprocess the call voice and obtain the voice to be recognized.
  • the voice feature extraction unit 462 is configured to perform voice feature extraction on the voice to be recognized, and obtain a voice feature to be recognized corresponding to the voice to be recognized.
  • the voiceprint model judging unit 463 is configured to perform voiceprint recognition by using the voiceprint recognition model to identify the voice features to be recognized and the original voice features corresponding to the overdue customers, to determine whether the speaker of the call voice is the overdue customer himself.
  • the emotion recognition unit 47 includes an emotion recognition model judgment unit 471 and an emotion recognition result acquisition unit 472.
  • the emotion recognition model judging unit 471 is configured to input speech features to be recognized into the emotion recognition model, and obtain the emotions to be confirmed corresponding to the features to be recognized, and the emotions to be confirmed carry an emotion score.
  • the emotion recognition result acquisition unit 472 is configured to calculate an emotion score carried by the emotion to be confirmed, and obtain an emotion recognition result.
  • each module in the overdue bill intelligent collection device can be implemented in whole or in part by software, hardware, and a combination thereof.
  • the above-mentioned modules may be embedded in the hardware form or independent of the processor in the computer device, or may be stored in the memory of the computer device in the form of software, so that the processor calls and performs the operations corresponding to the above modules.
  • a computer device is provided.
  • the computer device may be a server, and its internal structure diagram may be as shown in FIG. 10.
  • the computer device includes a processor, a memory, a network interface, and a database connected through a system bus.
  • the processor of the computer device is used to provide computing and control capabilities.
  • the memory of the computer device includes a non-volatile storage medium and an internal memory.
  • the non-volatile storage medium stores an operating system, computer-readable instructions, and a database.
  • the internal memory provides an environment for operating the operating system and computer-readable instructions in a non-volatile storage medium.
  • the computer equipment database is used to store the data involved in the overdue bill intelligent collection method.
  • the network interface of the computer device is used to communicate with an external terminal through a network connection.
  • the computer-readable instructions are executed by a processor to implement a method for intelligent collection of overdue bills.
  • a computer device which includes a memory, a processor, and computer-readable instructions stored on the memory and executable on the processor.
  • the processor executes the computer-readable instructions
  • the processor implements the overdue in the foregoing embodiment
  • the steps of the smart bill collection method include, for example, steps S10 to S40 shown in FIG. 2.
  • the functions of each module / unit of the overdue bill intelligent collection device in the foregoing embodiment are implemented, for example, modules 10 to 40 shown in FIG. 9. To avoid repetition, we will not repeat them here.
  • one or more non-readable storage media storing computer-readable instructions are provided, and when the computer-readable instructions are executed by one or more processors, the one or more processors implement The steps of the method for intelligent collection of overdue bills in the foregoing embodiment include, for example, steps S10 to S40 shown in FIG. 2.
  • the one or more processors are caused to implement the functions of each module / unit of the overdue bill intelligent collection device in the foregoing embodiment, for example, modules 10 to 10 shown in FIG. 9 Module 40. To avoid repetition, we will not repeat them here.
  • Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory.
  • Volatile memory can include random access memory (RAM) or external cache memory.
  • RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous chain Synchlink DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).
  • SRAM static RAM
  • DRAM dynamic RAM
  • SDRAM synchronous DRAM
  • DDRSDRAM dual data rate SDRAM
  • ESDRAM enhanced SDRAM
  • SLDRAM synchronous chain Synchlink DRAM
  • Rambus direct RAM
  • DRAM direct memory bus dynamic RAM
  • RDRAM memory bus dynamic RAM

Landscapes

  • Business, Economics & Management (AREA)
  • Accounting & Taxation (AREA)
  • Finance (AREA)
  • Engineering & Computer Science (AREA)
  • Development Economics (AREA)
  • Economics (AREA)
  • Marketing (AREA)
  • Strategic Management (AREA)
  • Technology Law (AREA)
  • Physics & Mathematics (AREA)
  • General Business, Economics & Management (AREA)
  • General Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Telephonic Communication Services (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

一种逾期账单智能催收方法、装置、计算机设备及存储介质,该方法包括:根据客户风险等级划分策略对入催案件的逾期客户信息进行划分,获取逾期客户名单,逾期客户名单包括逾期客户信息和与逾期客户信息相对应的逾期类型标识(S20);获取与逾期类型标识相对应的催收话术模板,将逾期客户信息添加到催收话术模板对应的位置,获取目标催收话术(S30);调用电话平台,自动拨通逾期客户的联系电话并将目标催收话术转换成目标催收语音,获取逾期客户对目标催收语音反馈的逾期催收结果(S40)。该方法提高催收工作效率,降低催收成本和因催收策略使用不当引起的客户投诉率。

Description

逾期账单智能催收方法、装置、计算机设备及存储介质
本申请以2018年8月2日提交的申请号为201810870736.2,名称为“逾期账单智能催收方法、装置、计算机设备及存储介质”的中国发明专利申请为基础,并要求其优先权。
技术领域
本申请涉及信息处理技术领域,尤其涉及一种逾期账单智能催收方法、装置、计算机设备及存储介质。
背景技术
随着人们消费意识和消费习惯的转变,以及国家政策的倡导,消费贷款等借贷业务发展迅速,随之而来出现的贷款逾期还款问题也越来越突出。目前业内对于逾期未还款的贷款催收大多是基于人工催收手段来处理金融行业的逾期贷款的。该种人工催收方式不仅需要耗费大量成本投入到催收人员的培训,还存在催收工作效率较低的问题。另外,催收人员在催收过程中会存在催收策略使用不当引起客户投诉,甚至法律诉讼风险的问题,给公司带来巨大损失,不利于贷款业务稳定高效地开展。
发明内容
本申请实施例提供一种逾期账单智能催收方法、装置、计算机设备及存储介质,以解决当前人工催收时存在的催收效率低且成本较高的问题。
一种逾期账单智能催收方法,包括:
根据催收规则对原始案件进行筛选,获取入催案件,所述入催案件携带有逾期客户信息,所述逾期客户信息包括逾期客户的联系电话;
根据客户风险等级划分策略对所述入催案件的逾期客户信息进行划分,获取逾期客户名单,所述逾期客户名单包括逾期客户信息和与所述逾期客户信息相对应的逾期类型标识;
获取与所述逾期类型标识相对应的催收话术模板,将所述逾期客户信息添加到所述催收话术模板对应的位置,获取目标催收话术;
调用电话平台,自动拨通逾期客户的联系电话并将所述目标催收话术转换成目标催收语音,获取逾期客户对所述目标催收语音反馈的逾期催收结果。
一种逾期账单智能催收装置,包括:
入催案件获取模块,用于根据催收规则对原始案件进行筛选,获取入催案件,所述入催案件携带有逾期客户信息,所述逾期客户信息包括逾期客户的联系电话;
逾期客户名单获取模块,用于根据客户风险等级划分策略对所述入催案件的逾期客户 信息进行划分,获取逾期客户名单,所述逾期客户名单包括逾期客户信息和与所述逾期客户信息相对应的逾期类型标识;
目标催收话术获取模块,用于获取与所述逾期类型标识相对应的催收话术模板,将所述逾期客户信息添加到所述催收话术模板对应的位置,获取目标催收话术;
逾期催收结果获取模块,用于调用电话平台,自动拨通逾期客户的联系电话并将所述目标催收话术转换成目标催收语音,获取逾期客户对所述目标催收语音反馈的逾期催收结果。
一种计算机设备,包括存储器、处理器以及存储在所述存储器中并可在所述处理器上运行的计算机可读指令,所述处理器执行所述计算机可读指令时实现如下步骤:
根据催收规则对原始案件进行筛选,获取入催案件,所述入催案件携带有逾期客户信息,所述逾期客户信息包括逾期客户的联系电话;
根据客户风险等级划分策略对所述入催案件的逾期客户信息进行划分,获取逾期客户名单,所述逾期客户名单包括逾期客户信息和与所述逾期客户信息相对应的逾期类型标识;
获取与所述逾期类型标识相对应的催收话术模板,将所述逾期客户信息添加到所述催收话术模板对应的位置,获取目标催收话术;
调用电话平台,自动拨通逾期客户的联系电话并将所述目标催收话术转换成目标催收语音,获取逾期客户对所述目标催收语音反馈的逾期催收结果。
一个或多个存储有计算机可读指令的非易失性可读存储介质,所述计算机可读指令被一个或多个处理器执行时,使得所述一个或多个处理器实现如下步骤:
根据催收规则对原始案件进行筛选,获取入催案件,所述入催案件携带有逾期客户信息,所述逾期客户信息包括逾期客户的联系电话;
根据客户风险等级划分策略对所述入催案件的逾期客户信息进行划分,获取逾期客户名单,所述逾期客户名单包括逾期客户信息和与所述逾期客户信息相对应的逾期类型标识;
获取与所述逾期类型标识相对应的催收话术模板,将所述逾期客户信息添加到所述催收话术模板对应的位置,获取目标催收话术;
调用电话平台,自动拨通逾期客户的联系电话并将所述目标催收话术转换成目标催收语音,获取逾期客户对所述目标催收语音反馈的逾期催收结果。
本申请的一个或多个实施例的细节在下面的附图及描述中提出。本申请的其他特征和优点将从说明书、附图以及权利要求书变得明显。
附图说明
为了更清楚地说明本申请实施例的技术方案,下面将对本申请实施例的描述中所需要使用的附图作简单地介绍,显而易见地,下面描述中的附图仅仅是本申请的一些实施例, 对于本领域普通技术人员来讲,在不付出创造性劳动性的前提下,还可以根据这些附图获得其他的附图。
图1是本申请一实施例中逾期账单智能催收方法的一应用场景图;
图2是本申请一实施例中逾期账单智能催收方法的一流程图;
图3是图2中步骤S40的一具体流程图;
图4是图3中步骤S43的一具体流程图;
图5是本申请一实施例中逾期账单智能催收方法的另一流程图;
图6是本申请一实施例中逾期账单智能催收方法的另一流程图;
图7是图6中步骤S46的一具体流程图;
图8是图6中步骤S47的一具体流程图;
图9本申请一实施例中逾期账单智能催收装置的一示意图;
图10是本申请一实施例中计算机设备的一示意图。
具体实施方式
下面将结合本申请实施例中的附图,对本申请实施例中的技术方案进行清楚、完整地描述,显然,所描述的实施例是本申请一部分实施例,而不是全部的实施例。基于本申请中的实施例,本领域普通技术人员在没有作出创造性劳动前提下所获得的所有其他实施例,都属于本申请保护的范围。
本申请实施例提供的逾期账单智能催收方法,可应用在如图1的应用环境中,该逾期账单智能催收方法应用在智能催收系统中,该智能催收系统包括客户端和服务器,其中,客户端通过网络与服务器进行通信。其中,客户端又称为用户端,是指与服务器相对应,为客户提供本地服务的程序。客户端可安装在但不限于各种个人计算机、笔记本电脑、智能手机、平板电脑和便携式可穿戴设备。服务器可以用独立的服务器或者是多个服务器组成的服务器集群来实现。
在一实施例中,如图2所示,提供一种逾期账单智能催收方法,以该方法应用在图1中的服务器为例进行说明,包括如下步骤:
S10:根据催收规则对原始案件进行筛选,获取入催案件,入催案件携带有逾期客户信息,逾期客户信息包括逾期客户的联系电话。
其中,催收规则指用于判断原始案件是否属于需要进行催收的案件的规则。原始案件指业务系统存储的需要还款的案件。入催案件指满足催收规则,在应还款日没有还款的案件。
具体地,智能催收系统获取业务系统发送的原始案件,原始案件携带有该原始案件的应还款日,智能催收系统在获取原始案件后,采用催收规则对原始案件进行筛选,筛选出在应还款日没有还款的案件,确定为入催案件。智能催收系统获取的入催案件携带有逾期客户信息,该逾期客户信息指逾期客户的个人信息,包括但不限于逾期客户的姓名、银行卡账号、还款金额、还款日期和联系电话。其中,逾期客户指在应还款日没有完成还款的 客户。
进一步地,为了防止从业务系统将原始案件发送到智能催收系统的过程中,客户信息泄露的情况发生,业务系统会在将原始案件发送给智能催收系统之前,采用RSA算法对原始案件进行加密。加密过程具体为,智能催收系统随机选择两个较大的素数A和B,计算素数A和素数B的乘积N。然后选择一个整数e与(A-1)*(B-1)的积互质,且e小于(A-1)*(B-1),最后根据公式d*e≡1(mod(A-1)(B-1))计算d。则(N,e)称为公钥,(N,d)称为私钥。智能催收系统保留私钥,将公钥发送给业务系统,业务系统使用这个公钥对原始案件中的内容进行加密,生成对应的密文。相应地,在步骤S10之前,即智能催收系统根据催收规则对原始案件进行筛选,获取入催案件的步骤之前,智能催收系统使用私钥对原始案件进行解密。
由于RSA算法是目前最有影响力和最常用的公钥加密算法,安全性高,能够抵抗到目前为止已知的绝大多数的密码攻击,因此,使用RSA算法对原始案件进行加密和解密,可以提高数据在传输过程中的安全性。
S20:根据客户风险等级划分策略对入催案件的逾期客户信息进行划分,获取逾期客户名单,逾期客户名单包括逾期客户信息和与逾期客户信息相对应的逾期类型标识。
其中,客户风险等级划分策略指根据客户逾期还款金额、逾期还款天数和历史逾期还款次数等指标中的至少一个对客户进行风险等级划分的策略。逾期客户名单指根据客户风险等级划分策略对入催案件进行等级划分后得到的包括逾期客户信息的名单。该逾期客户名单用于存储逾期客户信息和与逾期客户信息对应的逾期类型标识。其中,逾期客户信息包括但不限于逾期客户的姓名、银行卡账号、还款金额、还款日期、联系电话、历史逾期还款天数和历史逾期还款次数。逾期类型标识指用于唯一识别逾期客户类型的标识。该逾期类型标识可以是特定的字符串、特定的数字或者特定的字母中的任何一种或者多种组合。
本实施例中的逾期客户类型包括但不限于高阶客户、低阶客户、惯逾客户、失联客户和老赖客户。其中,高阶客户指逾期还款天数超过60天的客户。低阶客户指逾期还款天数不超过60天的客户。惯逾客户指6个月内的历史逾期还款次数累计达到4次的客户。失联客户指根据入催案件中的联系电话无法联系到的客户。老赖客户指到了指定还款时间一直不还钱的客户。为了方便区分,每一种逾期客户名单对应一个逾期类型标识。例如:低阶客户对应低阶客户类型标识L,高阶客户对应高阶客户类型标识H,惯逾客户对应惯逾客户类型标识F,失联客户对应失联客户类型标识O,老赖客户对应老赖客户类型标识Z。
具体地,智能催收系统获取入催案件后,会从入催案件携带的逾期客户信息中获取逾期客户对应的历史逾期还款天数和历史逾期还款次数。然后基于预先设置的客户风险等级划分策略对逾期客户对应的历史逾期还款天数和历史逾期还款次数进行判断,确定入催案件中逾期客户对应的逾期客户类型,生成不同逾期类型标识对应的逾期客户名单。该逾期 客户名单中包括逾期客户信息和逾期类型标识,方便后续通过逾期类型标识获取与逾期类型标识对应的催收话术模板,完成对入催案件对应的逾期客户进行还款催收。
S30:获取与逾期类型标识相对应的催收话术模板,将逾期客户信息添加到催收话术模板对应的位置,获取目标催收话术。
其中,催收话术模板指智能催收系统中预先设置的用于对逾期客户进行还款催收的话术模板。其中,每一催收话术模板与一逾期类型标识相关联,以便基于逾期类型标识查找到对应的催收话术模板。本实施例中,催收话术模板以文本形式存储在智能催收系统中,方便后续将逾期客户信息添加到催收话术模板中生成目标催收话术。
具体地,获取逾期类型标识后,智能催收系统会根据获取的逾期类型标识选取与逾期类型标识对应的催收话术模板,然后将逾期客户信息中的姓名、银行卡账号、还款金额和还款日期添加到与该逾期类型标识相对应的催收话术模板中,获取目标催收话术。其中,目标催收话术指有逾期客户信息和催收话术模板组成的针对特定逾期客户类型的逾期客户的催收话术。
智能催收系统为了更加方便地服务客户,针对不同逾期客户类型的逾期客户,制定不同的催收话术模板。例如:低阶客户对应的催收话术为:XXX先生/女士,你好!您XX银行尾号为xxxx的信用卡逾期xxxx元,请尽快在xxxx日之前还款,如果没有异议并同意还款请确认……等待客户回答……,如果还有其他问题请拨打95511客户电话询问,谢谢您的来电!
高阶客户对应的催收话术为:XXX先生/女士,你好!您XX银行尾号为xxxx的信用卡已经逾期xxxx元,为避免影响您的个人征信,请尽快在xxxx日之前还款,如果没有异议请确认……等待客户回答……,如果客户有疑问直接转接到人工坐席处理,请按0。
本实施例中,通过逾期客户信息和与逾期类型标识相对应的催收话术模板,获取目标催收话术,避免人工催收时,由于催收策略使用不当引起客户不满的情况发生,同时为步骤S40中通过电话平台对逾期客户进行电话催收提供数据来源。
S40:调用电话平台,自动拨通逾期客户的联系电话并将目标催收话术转换成目标催收语音,获取逾期客户对目标催收语音反馈的逾期催收结果。
具体地,智能催收系统获取目标催收话术后,会从逾期客户名单中的逾期客户信息中获取逾期客户的联系电话。在获取目标催收话术和逾期客户的联系电话后,智能催收系统会调用内置的电话平台,该电话平台集成了自动拨号技术,电话平台会通过集成的自动拨号技术拨打逾期客户的联系电话对应的终端设备,在逾期客户的终端设备接通后,智能催收系统会将目标催收话术转换为目标催收语音,通过电话平台发送给逾期客户的终端设备,以使逾期客户可通过该终端设备听到该目标催收语音,从而实现对逾期客户进行电话催收。本实施例中,该终端设备具体为手机、具有通话功能的智能手表或其他设备。
本实施例中,将目标催收话术转化成催收语音,利用该催收语音对逾期客户进行电话催收,使得电话催收按照预设的标准的催收话术对逾期客户进行电话催收,避免催收人员 在电话催收过程中,加入自己的情绪,对逾期催收结果造成不良影响。
同时,智能催收系统还可获取内置的电话平台所获取到的逾期客户的通话语音,以便对逾期客户的通话语音进行分析,获取逾期催收结果。智能催收系统通过内置的电话平台对逾期客户进行电话催收,实现不需要人工干预即可完成电话催收的功能,提高催收工作效率。
本申请实施例所提供的逾期账单智能催收方法中,首先通过催收规则对原始案件进行筛选,将不符合催收规则的原始案件删除,获取符合催收规则的入催案件,有效减少智能催收系统从业务系统获取数据传输量,节省数据传输时间,提高催收效率。然后,根据智能催收系统预先设置的客户风险等级划分策略对入催案件的逾期客户信息进行划分,获取不同逾期客户类型对应的逾期客户名单,为了方便识别,逾期客户名单中包括逾期类型标识。接着,根据逾期类型标识获取与逾期类型标识对应的催收话术模板和逾期客户信息,生成针对特定逾期客户的目标催收话术,使得目标催收话术具有针对性。最后将逾期客户信息中逾期客户的联系电话和目标催收话术发送给智能催收系统内置的电话平台,通过电话平台对逾期客户进行电话催收,获取逾期催收结果,实现不需要人工干预即可完成电话催收的功能,提高催收工作效率。
在一实施例中,如图3所示,步骤S40,调用电话平台,自动拨通逾期客户的联系电话并将目标催收话术转换成目标催收语音,获取逾期客户对目标催收语音反馈的逾期催收结果,具体包括如下步骤:
S41:基于逾期客户的联系电话,调用电话平台对逾期客户的联系电话进行自动拔号。
本实施例中内置的电话平台指智能催收系统中内置的电话平台,该电话平台中集成了自动拨号技术。该自动拨号技术为CTI技术,其中,CTI技术是从传统的计算机电话集成(Computer Telephony Integration)技术发展而来的一种技术,CTI技术可以将计算机技术应用到电话系统中,能够自动地对电话中的指令信息进行识别处理,并通过建立有关的话路连接,而向用户传送预定的录音文件和转接来话的技术。
具体地,智能催收系统在获取逾期客户的联系电话后,会将该联系电话发送给内置的电话平台,电话平台在获取到逾期客户的联系电话后,会通过集成的自动拨号技术对对逾期客户的联系电话进行自动拔号。本实施例中,智能催收系统通过内置的电话平台中的自动拨号技术实现了自动拨打逾期客户的联系电话对应的终端设备,实现不需要人工干预即可完成电话拨号的功能。
S42:获取联系电话对应的终端设备反馈的接通信号,基于接通信号,采用TTS技术将目标催收话术转化成对应的目标催收语音,将目标催收语音发送给终端设备。
其中,接通信号指联系电话对应的终端设备在接通后反馈给电话平台的信号,该接通信号可用于识别联系电话对应的终端设备是否接通。TTS技术指将计算机自己产生或者外部输入的文字信息转变为汉语口语并输出的技术。目标催收语音指将目标催收话术中的内容转化成的催收语音。
具体地,电话平台基于逾期客户的联系电话,对该联系电话对应的终端设备进行自动拨号,当该联系电话对应的终端设备接通后,该终端设备会发送一个接通信号给电话平台,表示该联系电话对应的终端设备已经接通。电话平台在获取接通信号后,会将该接通信号反馈给智能催收系统的服务器,智能催收系统的服务器会通过TTS技术将目标催收话术转化成对应的目标催收语音,通过电话平台将目标催收语音发送给终端设备,以使携带该终端设备的逾期客户可听到该目标催收语音,以实现智能语音催收,无需人工干预催收过程,提高催收效率,并有助于降低催收成本。
进一步地,通过采用TTS技术将目标催收话术转化成对应的目标催收语音,使得催收语音标准化,避免人工催收时催收人员语气不当引起客户不满或者催收人员说话语音不标准而影响催收效果的情况发生。
S43:接收终端设备反馈的通话语音,对通话语音进行语义分析,获取逾期催收结果。
具体地,智能催收系统会通过电话平台接收终端设备反馈的来自逾期客户的通话语音。智能催收系统在获取到终端设备反馈的通话语音后,会对逾期客户的通话语音进行语义分析,判断逾期客户的通话语音中是否含有预先设置的“好的”、“OK”、“可以”和“等会就还”等携带同意还款单向的关键词。若通话语音中含有“好的”、“OK”、“可以”和“等会就还”等携带同意还款单向的关键词,则表示逾期客户同意还款,逾期催收结果为愿意还款。若通话语音中不含有“好的”、“OK”、“可以”和“等会就还”等携带同意还款单向的关键词,则表示逾期客户不同意还款,逾期催收结果为不愿意还款。
步骤S41-步骤S43,智能催收系统通过内置的电话平台对逾期客户的联系电话对应的终端设备进行自动拨号,在拨号成功后,通过TTS技术将目标催收话术转化成目标催收语音,发送给逾期客户对应的终端设备,完成和逾期客户的对话,获取逾期催收结果,达到逾期账单智能催收的效果,不需要人工催收,降低催收过程的人工成本,提高催收效率。
在一实施例中,如图4所示,步骤S43中,对通话语音进行语义分析,获取逾期催收结果,具体包括如下步骤:
S431:对通话语音进行扫描,判断通话语音是否包含预先设置的目标关键词。
其中,目标关键词指在智能催收系统中预先设置的用于识别逾期客户是否同意还款的关键词。即该目标关键词是携带同意还款单向的关键词。
具体地,智能催收系统通过电话平台获取到逾期客户的终端设备发送的通话语音后,会根据智能催收系统中的关键词识别技术对通话语音进行逐字扫描,获取通话语音的语音特征。判断通话语音中是否存在和目标关键词匹配的词语。本实施例中的关键词识别技术采用预先训练好的HMM(Hidden Markov Model隐式马尔科夫模型)对通话语音进行目标关键词识别。其中,预先训练好的HMM模型是预先训练好的存储在智能催收系统中,用于识别通话语音中是否存在“好的”、“OK”、“可以”和“等会就还”等目标关键词的模型。本实施例中,采用预先训练好的HMM模型识别通话语音中是否存在目标关键词,可以提高识别准确率。
或者,智能催收系统通过语音转换文字技术,将通话语音转化为文字。然后,采用分词工具对该通话语音对应的文字进行切分,去除停用词(分词、介词和代词等)、符号、数字和字母等词,获取待识别词。该待识别词指通话语音对应的文字中去除停用词、符号、数字和字母之后获取的词。最后,采用相似度算法对待识别词和目标关键词进行相似度计算,若相似度达到预设值,则确定通话语音中包含目标关键词。其中。本实施例中的相似度算法包括但不限于余弦相似度算法。
S432:若通话语音中包含有目标关键词,则获取携带有同意标识的逾期催收结果。
其中,同意标识指用于表示逾期客户同意还款的标识。具体地,当通话语音中包含“好的”、“OK”、“可以”和“等会就还”等目标关键词时,则表示逾期客户同意还款,与该逾期客户进行电话沟通后获取的逾期催收结果为同意还款。为了方便智能催收系统根据逾期催收结果执行后续步骤,智能催收系统在确定通话语音中包含目标关键词后,会对该逾期客户的逾期催收结果标注对应的同意标识,如“OK”或者“1”。
S433:若通话语音中不包含有目标关键词,则获取携带有不同意标识的逾期催收结果。
其中,不同意标识指用于表示逾期客户不同意还款的标识。具体地,当通话语音中不包含“好的”、“OK”、“可以”和“等会就还”等目标关键词时,则表示逾期客户不同意还款,与该逾期客户进行电话沟通后获取的逾期催收结果为不同意还款,智能催收系统在确定通话语音中不包含目标关键词后,会对该逾期客户的逾期催收结果标注对应的不同意标识,如“no”或者“0”。
步骤S431-步骤S433,智能催收系统通过对通话语音进行扫描,判断通话语音中是否含有目标关键词,确定逾期客户是否同意还款,达到无需人工干预即可实现逾期账单智能催收的目的。为了方便智能催收系统根据逾期催收结果执行后续步骤,智能催收系统在确定逾期客户语音通话中是否存在目标关键词后,会对逾期催收结果匹配对应的同意标识或者不同意标识,方便智能催收系统根据同意标识或者不同意标识执行对应的后续步骤。
在一实施例中,为了进一步保证电话沟通的有效性,智能催收系统在与逾期客户电话沟通完毕且逾期催收结果为同意还款后,还会通过短信和邮件的形式,发送对应的短信和邮件来提醒逾期客户进行还款。因此,逾期客户名单中还包括逾期客户的邮箱地址,如图5所示,在步骤S43,获取逾期催收结果的步骤之后,逾期账单智能催收方法还包括:
S44:若逾期催收结果携带有同意标识,则基于催收信息模板和逾期客户信息,获取有效催收信息。
其中,催收信息模板指预先设置好并存储在智能催收系统中的,用于提示逾期客户按时还款的信息模板。例如催收信息模板为:XXX先生/女士,您好!根据电话沟通,您已确认您平安银行尾号为xxxx的信用卡将在xxxx日之前完成还款金额xxxx元;如果还有其他问题请拨打95511客户电话询问,谢谢您的来电!
有效催收信息指将逾期客户信息添加到催收信息模板中得到的针对特定逾期客户的催收信息。
具体地,当逾期催收结果携带有同意标识时,智能催收系统会获取催收信息模板和逾期客户信息,将逾期客户信息中的姓名、银行卡账号、还款金额和还款日期添加到催收信息模板中,获取针对该特定逾期客户的有效催收信息。
S45:将有效催收信息通过逾期客户的联系电话或者逾期客户的邮箱地址发送给对应的终端设备。
智能催收系统在获取有效催收信息后,会通过逾期客户的联系电话将有效催收信息发送给该逾期客户的联系电话对应的终端设备。另外,若逾期客户名单中还包括逾期客户的邮箱地址时,智能催收系统会从逾期客户名单中获取该逾期客户对应的邮箱地址,然后以邮件的形式将有效催收信息发送给该逾期客户的邮箱地址对应的终端设备。采用短信或者邮件两种方式发送有效催收信息给逾期客户,可以进一步确保逾期客户能收到并查看该有效催收信息,提高催收质量。
在一实施例中,当智能催收系统在获取到携带有不同意标识的逾期催收结果后,会对该逾期客户进行进一步地的情况确认,首先确认该通话语音是不是来自逾期客户本人,若是逾期客户本人,则执行后续步骤;若不是逾期客户本人,则交给线下跟催工作人员催收。其中,跟催工作人员指通过找到逾期客户本人,当面对逾期客户进行逾期贷款催收的工作人员。如图6所示,在步骤S43,获取逾期催收结果的步骤之后,逾期账单智能催收方法还包括:
S46:若逾期催收结果携带不同意标识,则对通话语音进行声纹识别,确定通话语音的说话人是否为逾期客户本人。
其中,声纹识别也称为说话人识别,通过辨认说话人的声纹,确定说话人的身份。
具体地,如果获取的逾期催收结果携带有不同意标识,则表示逾期客户不同意还款。此时,智能催收系统会使用内置的声纹识别技术对逾期客户的通话语音进行识别,确定该通话语音是否为逾期客户本人的声音。若确定通话语音不是逾期客户本人的声音,则表示该联系电话对应的联系人发生变化,需要使用线下跟催的方法联系逾期客户本人,以催促该逾期客户本人及时还清逾期贷款。若确定通话语音是逾期客户本人的声音,执行步骤S47,对通话语音进行情绪识别,获取逾期客户的情绪识别结果。
S47:若通话语音的说话人为逾期客户本人,则调用预先存储的情绪识别模型对通话语音进行情绪识别,获取与逾期催收结果相对应的情绪识别结果。
具体地,若智能催收系统确定通话语音的说话人是逾期客户本人时,则需要对通话语音进行进一步分析,确定逾期客户本人在接收电话催收时的情绪,以便后续人工坐席与逾期客户进行电话沟通时,采用照顾逾期客户情绪的催收策略和逾期客户进行电话沟通,避免催收人员在催收过程中因为催收策略使用不当引起客户投诉的问题发生。
步骤S46-步骤S47,智能催收系统在获取携带有不同意标识的逾期催收结果后,会首先对逾期客户的通话语音进行声纹识别,确认该通话语音是否来自本人;若通话语音来自逾期客户本人,则对通话语音进行情绪识别,获取情绪识别结果,方便人工坐席采取恰当 的催收策略与逾期客户进行电话沟通,降低客户投诉率。
在一实施例中,如图7所示,步骤S46,对通话语音进行声纹识别,确定通话语音的说话人是否为逾期客户本人,具体包括如下步骤:
S461:对通话语音进行预处理,获取待识别语音。
其中,待识别语音指对通话语音进行预处理后得到语音。具体地,对通话语音进行声纹识别之前,需要对通话语音进行预处理,去除通话语音中的噪声(包括语音通话中的静音和环境噪音),获取待识别语音。其中,预处理的过程具体包括:首先,对通话语音进行预加重处理;然后对预加重处理后的通话语音进行分帧和加窗运算;最后,使用端点检测技术对通话语音进行处理,去除通话语音中的噪声,以获取待识别语音。
S462:对待识别语音进行语音特征提取,获取待识别语音对应的待识别语音特征。
具体地,获取待识别语音后,将待识别语音通过快速傅里叶变换、对数运算和离散余弦变换处理后,获取待识别语音特征。该待识别语音特征为MFCC特征。
S463:采用声纹识别模型对待识别语音特征和逾期客户对应的原始语音特征进行声纹识别,确定通话语音的说话人是否为逾期客户本人。
其中,声纹识别模型指预先训练好的存储在智能催收系统中的用于识别通话语音的说话人是否为逾期客户本人的模型。现有多种成熟的声纹识别模型,包括但不限于GMM-UBM(Gaussian mixture model-universal background model,混合模型-通用背景模型)模型和i-vector(identity-vector,身份认证向量)模型。本实施例中的声纹识别模型采用i-vector模型作为声纹识别模型。原始语音特征指智能催收系统预先存储的该逾期客户的说话语音对应的语音特征。可以理解地,该逾期客户的说话语音经过与步骤S461-S462相对应的预处理和特征提取之后,即可获取到对应的原始语音特征,并将该原始语音特征与逾期客户信息关联存储,以便后续进行声纹识别时可直接调用,无需在识别过程中进行相应的处理,以提高识别效率。
具体地,当获取待识别语音特征后,智能催收系统会根据逾期客户的客户信息查找智能催收系统中预先存储的与该逾期客户信息对应的原始语音特征。然后,将该逾期客户的待识别语音特征和原始语音特征输入到预先训练好的声纹识别模型中,获取待识别语音特征和原始语音特征各自对应的i-vector向量;再计算待识别语音特征和原始语音特征各自对应的i-vector向量的空间距离,若空间距离大于根据实际情况设置的距离阈值,则可以确定通话语音的说话人为逾期客户本人。本实施例中的空间向量距离采用余弦距离计算公式计算。
步骤S461-步骤S463,通过获取通话语音对应的待识别语音特征,采用声纹识别模型对待识别语音特征和逾期客户对应的原始语音特征进行识别,确定通话语音的说话人是否为逾期客户本人,若为逾期客户本人,则表示逾期客户不愿意还款,需要通过后续步骤对通话语音进行情绪识别,方便坐席人员采用人工沟通的方式了解逾期客户不还款的原因。
在一实施例中,如图8所示,步骤S47,调用预先存储的情绪识别模型对通话语音进 行情绪识别,获取与逾期催收结果相对应的情绪识别结果,具体包括如下步骤:
S471:将待识别语音特征输入到情绪识别模型中,获取待识别语音特征对应的待确认情绪,待确认情绪携带有情绪得分。
其中,情绪识别模型是预先训练好并存储在智能催收系统中的用于识别逾期客户情绪的模型。本实施例中的情绪识别模型采用基于向量机的情绪识别模型。待确认情绪包括但不限于情绪识别模型识别出的说话人说话时的高兴、愤怒、悲伤、烦噪和平静等情绪。具体地,在根据待识别语音特征确认通话语音的说话人为逾期客户本人的声音后,智能催收系统将待识别语音特征输入到预先训练好的情绪识别模型中,对待识别语音特征进行情绪识别,以获取通话语音对应的逾期客户在说话时的各种待确认情绪的情绪得分。本实施例中的情绪识别模型在获取待识别语音特征后,会分别对待识别语音特征中的高兴、愤怒、悲伤、烦噪和平静等待确认情绪进行情绪得分计算,获取不同的待确认情绪对应的情绪得分。采用基于向量机的情绪识别模型计算复杂度较小,可根据少数支持向量决定最终结果,在训练过程中有助于抓住关键样本,剔除冗余样本,具有较好的鲁棒性。
S472:对待确认情绪携带的情绪得分进行计算,获取情绪识别结果。
其中,情绪识别结果指对待确认情绪携带的情绪得分进行计算,获取能最终表示逾期客户在被催收时情绪的结果。具体地,在获取通话语音中不同的待确认情绪的情绪得分后,情绪识别模型会比较各待确认情绪携带的情绪得分,将情绪得分最大的待确认情绪作为最终的情绪识别结果输出。智能催收系统会将该情绪识别结果和对应的逾期客户信息发送给人工坐席,便于人工坐席根据情绪识别结果采用合适的催收策略对逾期客户进行电话沟通。
步骤S471-步骤S472,将待识别语音特征输入到情绪识别模型中,获取待识别语音特征对应的待确认情绪;根据待确认情绪携带的情绪得分,获取逾期客户的情绪识别结果。通过情绪识别模型对逾期客户的情绪进行智能判断,提高情绪识别的准确性和判断速度,方便人工坐席根据情绪识别结果采用合适的催收策略对逾期客户进行电话沟通。。
本申请实施例所提供的逾期账单智能催收方法中,首先通过催收规则对原始案件进行筛选,将不符合催收规则的原始案件删除,获取符合催收规则的入催案件。然后根据智能催收系统预先设置的客户风险等级划分策略对入催案件的逾期客户信息进行划分,获取不同逾期客户类型对应的逾期客户名单,方便后续针对不同逾期客户类型采用不同的催收话术与逾期客户进行电话沟通。接着根据逾期客户名单中的逾期类型标识获取与逾期类型标识对应的催收话术模板和逾期客户信息,生成针对特定逾期客户的目标催收话术,通过电话平台的自动拨号技术对逾期客户进行电话催收,获取逾期催收结果,确定逾期客户是否有还款意愿,提高催收工作效率,降低催收成本。若逾期催收结果中携带有同意标识,则发送有效催收信息给逾期客户的终端设备;若逾期催收结果中携带有不同意标识,则对逾期客户的通话语音进行声纹识别和情绪识别,最终将情绪识别结果反馈给人工坐席,使得人工坐席根据情绪识别结果对逾期客户进行电话催收,降低因为催收策略使用不当引起客 户投诉的投诉率。
应理解,上述实施例中各步骤的序号的大小并不意味着执行顺序的先后,各过程的执行顺序应以其功能和内在逻辑确定,而不应对本申请实施例的实施过程构成任何限定。
在一实施例中,提供一种逾期账单智能催收装置,该逾期账单智能催收装置与上述实施例中逾期账单智能催收方法一一对应。如图9所示,该只能催搜装置包括入催案件获取模块10、逾期客户名单获取模块20、目标催收话术获取模块30和逾期催收结果获取模块40。各功能模块详细说明如下:
入催案件获取模块10,用于根据催收规则对原始案件进行筛选,获取入催案件,入催案件携带有逾期客户信息,逾期客户信息包括逾期客户的联系电话。
逾期客户名单获取模块20,用于根据客户风险等级划分策略对入催案件的逾期客户信息进行划分,获取逾期客户名单,逾期客户名单包括逾期客户信息和与逾期客户信息相对应的逾期类型标识。
目标催收话术获取模块30,用于获取与逾期类型标识相对应的催收话术模板,将逾期客户信息添加到催收话术模板对应的位置,获取目标催收话术。
逾期催收结果获取模块40,用于调用电话平台,自动拨通逾期客户的联系电话并将目标催收话术转换成目标催收语音,获取逾期客户对目标催收语音反馈的逾期催收结果。
进一步地,逾期催收结果获取模块40包括自动拨号单元41、目标语音获取发送单元42和逾期催收结果获取单元43。
自动拨号单元41,用于基于逾期客户的联系电话,调用电话平台对逾期客户的联系电话进行自动拔号。
目标语音获取发送单元42,用于获取联系电话对应的终端设备反馈的接通信号,基于接通信号,采用TTS技术将目标催收话术转化成对应的目标催收语音,将目标催收语音发送给终端设备。
逾期催收结果获取单元43,用于接收终端设备反馈的通话语音,对通话语音进行语义分析,获取逾期催收结果。
进一步地,逾期催收结果获取单元43包括目标关键词判断单元431、第一逾期催收结果获取单元432和第二逾期催收结果获取单元433。
目标关键词判断单元431,用于对通话语音进行扫描,判断通话语音是否包含预先设置的目标关键词。
第一逾期催收结果获取单元432,用于若通话语音中包含有目标关键词,则获取携带有同意标识的逾期催收结果。
第二逾期催收结果获取单元433,用于若通话语音中不包含有目标关键词,则获取携带有不同意标识的逾期催收结果。
进一步地,逾期客户名单中还包括逾期客户的邮箱地址,该逾期账单智能催收装置还包括有效催收信息获取单元44、有效催收信息发送单元45、声纹识别单元46和情绪识别 单元47。
有效催收信息获取单元44,用于若逾期催收结果携带有同意标识,则基于催收信息模板和逾期客户信息,获取有效催收信息。
有效催收信息发送单元45,用于将有效催收信息通过逾期客户的联系电话或者逾期客户的邮箱地址发送给对应的终端设备。
声纹识别单元46,用于若逾期催收结果携带不同意标识,则对通话语音进行声纹识别,确定通话语音的说话人是否为逾期客户本人。
情绪识别单元47,用于若通话语音的说话人为逾期客户本人,则调用预先存储的情绪识别模型对通话语音进行情绪识别,获取与逾期催收结果相对应的情绪识别结果。
进一步地,声纹识别单元46包括通话语音待处理单元461、语音特征提取单元462和声纹识别模型判断单元463。
通话语音待处理单元461,用于对通话语音进行预处理,获取待识别语音。
语音特征提取单元462,用于对待识别语音进行语音特征提取,获取待识别语音对应的待识别语音特征。
声纹模型判断单元463,用于采用声纹识别模型对待识别语音特征和逾期客户对应的原始语音特征进行声纹识别,确定通话语音的说话人是否为逾期客户本人。
进一步地,情绪识别单元47包括情绪识别模型判断单元471和情绪识别结果获取单元472。
情绪识别模型判断单元471,用于将待识别语音特征输入到情绪识别模型中,获取待识别语音特征对应的待确认情绪,待确认情绪携带有情绪得分。
情绪识别结果获取单元472,用于对待确认情绪携带的情绪得分进行计算,获取情绪识别结果。
关于逾期账单智能催收装置的具体限定可以参见上文中对于逾期账单智能催收方法的限定,在此不再赘述。上述逾期账单智能催收装置中的各个模块可全部或部分通过软件、硬件及其组合来实现。上述各模块可以硬件形式内嵌于或独立于计算机设备中的处理器中,也可以以软件形式存储于计算机设备中的存储器中,以便于处理器调用执行以上各个模块对应的操作。
在一个实施例中,提供了一种计算机设备,该计算机设备可以是服务器,其内部结构图可以如图10所示。该计算机设备包括通过系统总线连接的处理器、存储器、网络接口和数据库。其中,该计算机设备的处理器用于提供计算和控制能力。该计算机设备的存储器包括非易失性存储介质、内存储器。该非易失性存储介质存储有操作系统、计算机可读指令和数据库。该内存储器为非易失性存储介质中的操作系统和计算机可读指令的运行提供环境。该计算机设备的数据库用于存储逾期账单智能催收方法中涉及到的数据。该计算机设备的网络接口用于与外部的终端通过网络连接通信。该计算机可读指令被处理器执行时以实现一种逾期账单智能催收方法。
在一个实施例中,提供了一种计算机设备,包括存储器、处理器及存储在存储器上并可在处理器上运行的计算机可读指令,处理器执行计算机可读指令时实现上述实施例中逾期账单智能催收方法的步骤,例如图2所示的步骤S10至步骤S40。或者,处理器执行计算机可读指令时实现上述实施例中逾期账单智能催收装置的各模块/单元的功能,例如图9所示的模块10至模块40。为避免重复,这里不再赘述。
在一个实施例中,提供了一个或多个存储有计算机可读指令的非易失性可读存储介质,计算机可读指令被一个或多个处理器执行时,使得一个或多个处理器实现上述实施例逾期账单智能催收方法的步骤,例如图2所示的步骤S10至步骤S40。或者,计算机可读指令被一个或多个处理器执行时,使得一个或多个处理器实现上述实施例中逾期账单智能催收装置的各模块/单元的功能,例如图9所示的模块10至模块40。为避免重复,这里不再赘述。
本领域普通技术人员可以理解实现上述实施例方法中的全部或部分流程,是可以通过计算机可读指令来指令相关的硬件来完成,所述的计算机可读指令可存储于一非易失性计算机可读取存储介质中,该计算机可读指令在执行时,可包括如上述各方法的实施例的流程。其中,本申请所提供的各实施例中所使用的对存储器、存储、数据库或其它介质的任何引用,均可包括非易失性和/或易失性存储器。非易失性存储器可包括只读存储器(ROM)、可编程ROM(PROM)、电可编程ROM(EPROM)、电可擦除可编程ROM(EEPROM)或闪存。易失性存储器可包括随机存取存储器(RAM)或者外部高速缓冲存储器。作为说明而非局限,RAM以多种形式可得,诸如静态RAM(SRAM)、动态RAM(DRAM)、同步DRAM(SDRAM)、双数据率SDRAM(DDRSDRAM)、增强型SDRAM(ESDRAM)、同步链路(Synchlink)DRAM(SLDRAM)、存储器总线(Rambus)直接RAM(RDRAM)、直接存储器总线动态RAM(DRDRAM)、以及存储器总线动态RAM(RDRAM)等。
所属领域的技术人员可以清楚地了解到,为了描述的方便和简洁,仅以上述各功能单元、模块的划分进行举例说明,实际应用中,可以根据需要而将上述功能分配由不同的功能单元、模块完成,即将所述装置的内部结构划分成不同的功能单元或模块,以完成以上描述的全部或者部分功能。
以上所述实施例仅用以说明本申请的技术方案,而非对其限制;尽管参照前述实施例对本申请进行了详细的说明,本领域的普通技术人员应当理解:其依然可以对前述各实施例所记载的技术方案进行修改,或者对其中部分技术特征进行等同替换;而这些修改或者替换,并不使相应技术方案的本质脱离本申请各实施例技术方案的精神和范围,均应包含在本申请的保护范围之内。

Claims (20)

  1. 一种逾期账单智能催收方法,其特征在于,包括:
    根据催收规则对原始案件进行筛选,获取入催案件,所述入催案件携带有逾期客户信息,所述逾期客户信息包括逾期客户的联系电话;
    根据客户风险等级划分策略对所述入催案件的逾期客户信息进行划分,获取逾期客户名单,所述逾期客户名单包括逾期客户信息和与所述逾期客户信息相对应的逾期类型标识;
    获取与所述逾期类型标识相对应的催收话术模板,将所述逾期客户信息添加到所述催收话术模板对应的位置,获取目标催收话术;
    调用电话平台,自动拨通逾期客户的联系电话并将所述目标催收话术转换成目标催收语音,获取逾期客户对所述目标催收语音反馈的逾期催收结果。
  2. 如权利要求1所述的逾期账单智能催收方法,其特征在于,所述调用电话平台,自动拨通逾期客户的联系电话并将所述目标催收话术转换成目标催收语音,获取逾期客户对所述目标催收语音反馈的逾期催收结果,包括:
    基于所述逾期客户的联系电话,调用电话平台对所述逾期客户的联系电话进行自动拔号;
    获取所述联系电话对应的终端设备反馈的接通信号,基于所述接通信号,采用TTS技术将所述目标催收话术转化成对应的目标催收语音,将所述目标催收语音发送给所述终端设备;
    接收所述终端设备反馈的通话语音,对所述通话语音进行语义分析,获取逾期催收结果。
  3. 如权利要求2所述的逾期账单智能催收方法,其特征在于,所述对所述通话语音进行语义分析,获取逾期催收结果,包括:
    对所述通话语音进行扫描,判断所述通话语音是否包含预先设置的目标关键词;
    若所述通话语音中包含有所述目标关键词,则获取携带有同意标识的逾期催收结果;
    若所述通话语音中不包含有所述目标关键词,则获取携带有不同意标识的逾期催收结果。
  4. 如权利要求1所述的逾期账单智能催收方法,其特征在于,所述逾期客户名单中还包括逾期客户的邮箱地址;
    在所述获取逾期催收结果的步骤之后,所述逾期账单智能催收方法还包括:
    若所述逾期催收结果携带有同意标识,则基于催收信息模板和逾期客户信息,获取有效催收信息;
    将有效催收信息通过所述逾期客户的联系电话或者逾期客户的邮箱地址发送给对应的终端设备。
  5. 如权利要求1所述的逾期账单智能催收方法,其特征在于,在所述获取逾期催收结果的步骤之后,所述逾期账单智能催收方法还包括:
    若所述逾期催收结果携带不同意标识,则对所述通话语音进行声纹识别,确定所述通话语音的说话人是否为逾期客户本人;
    若所述通话语音的说话人为逾期客户本人,则调用预先存储的情绪识别模型对所述通话语音进行情绪识别,获取与所述逾期催收结果相对应的情绪识别结果。
  6. 如权利要求5所述的逾期账单智能催收方法,其特征在于,所述对所述通话语音进行声纹识别,确定所述通话语音的说话人是否为逾期客户本人,包括:
    对所述通话语音进行预处理,获取待识别语音;
    对所述待识别语音进行语音特征提取,获取所述待识别语音对应的待识别语音特征;
    采用声纹识别模型对所述待识别语音特征和所述逾期客户对应的原始语音特征进行声纹识别,确定所述通话语音的说话人是否为逾期客户本人。
  7. 如权利要求5或6所述的逾期账单智能催收方法,其特征在于,所述调用预先存储的情绪识别模型对所述通话语音进行情绪识别,获取与所述逾期催收结果相对应的情绪识别结果,包括:
    将所述待识别语音特征输入到所述情绪识别模型中,获取所述待识别语音特征对应的待确认情绪,所述待确认情绪携带有情绪得分;
    对所述待确认情绪携带的情绪得分进行计算,获取情绪识别结果。
  8. 一种逾期账单智能催收装置,其特征在于,包括:
    入催案件获取模块,用于根据催收规则对原始案件进行筛选,获取入催案件,所述入催案件携带有逾期客户信息,所述逾期客户信息包括逾期客户的联系电话;
    逾期客户名单获取模块,用于根据客户风险等级划分策略对所述入催案件的逾期客户信息进行划分,获取逾期客户名单,所述逾期客户名单包括逾期客户信息和与所述逾期客户信息相对应的逾期类型标识;
    目标催收话术获取模块,用于获取与所述逾期类型标识相对应的催收话术模板,将所述逾期客户信息添加到所述催收话术模板对应的位置,获取目标催收话术;
    逾期催收结果获取模块,用于调用电话平台,自动拨通逾期客户的联系电话并将所述目标催收话术转换成目标催收语音,获取逾期客户对所述目标催收语音反馈的逾期催收结果。
  9. 一种计算机设备,包括存储器、处理器以及存储在所述存储器中并可在所述处理器上运行的计算机可读指令,其特征在于,所述处理器执行所述计算机可读指令时实现如下步骤:
    根据催收规则对原始案件进行筛选,获取入催案件,所述入催案件携带有逾期客户信息,所述逾期客户信息包括逾期客户的联系电话;
    根据客户风险等级划分策略对所述入催案件的逾期客户信息进行划分,获取逾期客户 名单,所述逾期客户名单包括逾期客户信息和与所述逾期客户信息相对应的逾期类型标识;
    获取与所述逾期类型标识相对应的催收话术模板,将所述逾期客户信息添加到所述催收话术模板对应的位置,获取目标催收话术;
    调用电话平台,自动拨通逾期客户的联系电话并将所述目标催收话术转换成目标催收语音,获取逾期客户对所述目标催收语音反馈的逾期催收结果。
  10. 如权利要求9所述的计算机设备,其特征在于,所述调用电话平台,自动拨通逾期客户的联系电话并将所述目标催收话术转换成目标催收语音,获取逾期客户对所述目标催收语音反馈的逾期催收结果,包括:
    基于所述逾期客户的联系电话,调用电话平台对所述逾期客户的联系电话进行自动拔号;
    获取所述联系电话对应的终端设备反馈的接通信号,基于所述接通信号,采用TTS技术将所述目标催收话术转化成对应的目标催收语音,将所述目标催收语音发送给所述终端设备;
    接收所述终端设备反馈的通话语音,对所述通话语音进行语义分析,获取逾期催收结果。
  11. 如权利要求10所述的计算机设备,其特征在于,所述对所述通话语音进行语义分析,获取逾期催收结果,包括:
    对所述通话语音进行扫描,判断所述通话语音是否包含预先设置的目标关键词;
    若所述通话语音中包含有所述目标关键词,则获取携带有同意标识的逾期催收结果;
    若所述通话语音中不包含有所述目标关键词,则获取携带有不同意标识的逾期催收结果。
  12. 如权利要求9所述的计算机设备,其特征在于,所述逾期客户名单中还包括逾期客户的邮箱地址;
    在所述获取逾期催收结果的步骤之后,所述处理器执行所述计算机可读指令时还实现如下步骤:
    若所述逾期催收结果携带有同意标识,则基于催收信息模板和逾期客户信息,获取有效催收信息;
    将有效催收信息通过所述逾期客户的联系电话或者逾期客户的邮箱地址发送给对应的终端设备。
  13. 如权利要求9所述的计算机设备,其特征在于,在所述获取逾期催收结果的步骤之后,所述处理器执行所述计算机可读指令时还实现如下步骤:
    若所述逾期催收结果携带不同意标识,则对所述通话语音进行声纹识别,确定所述通话语音的说话人是否为逾期客户本人;
    若所述通话语音的说话人为逾期客户本人,则调用预先存储的情绪识别模型对所述通 话语音进行情绪识别,获取与所述逾期催收结果相对应的情绪识别结果。
  14. 如权利要求13所述的计算机设备,其特征在于,所述对所述通话语音进行声纹识别,确定所述通话语音的说话人是否为逾期客户本人,包括:
    对所述通话语音进行预处理,获取待识别语音;
    对所述待识别语音进行语音特征提取,获取所述待识别语音对应的待识别语音特征;
    采用声纹识别模型对所述待识别语音特征和所述逾期客户对应的原始语音特征进行声纹识别,确定所述通话语音的说话人是否为逾期客户本人。
  15. 一个或多个存储有计算机可读指令的非易失性可读存储介质,其特征在于,所述计算机可读指令被一个或多个处理器执行时,使得所述一个或多个处理器实现如下步骤:
    根据催收规则对原始案件进行筛选,获取入催案件,所述入催案件携带有逾期客户信息,所述逾期客户信息包括逾期客户的联系电话;
    根据客户风险等级划分策略对所述入催案件的逾期客户信息进行划分,获取逾期客户名单,所述逾期客户名单包括逾期客户信息和与所述逾期客户信息相对应的逾期类型标识;
    获取与所述逾期类型标识相对应的催收话术模板,将所述逾期客户信息添加到所述催收话术模板对应的位置,获取目标催收话术;
    调用电话平台,自动拨通逾期客户的联系电话并将所述目标催收话术转换成目标催收语音,获取逾期客户对所述目标催收语音反馈的逾期催收结果。
  16. 如权利要求15所述的非易失性可读存储介质,其特征在于,所述调用电话平台,自动拨通逾期客户的联系电话并将所述目标催收话术转换成目标催收语音,获取逾期客户对所述目标催收语音反馈的逾期催收结果,包括:
    基于所述逾期客户的联系电话,调用电话平台对所述逾期客户的联系电话进行自动拔号;
    获取所述联系电话对应的终端设备反馈的接通信号,基于所述接通信号,采用TTS技术将所述目标催收话术转化成对应的目标催收语音,将所述目标催收语音发送给所述终端设备;
    接收所述终端设备反馈的通话语音,对所述通话语音进行语义分析,获取逾期催收结果。
  17. 如权利要求16所述的非易失性可读存储介质,其特征在于,所述对所述通话语音进行语义分析,获取逾期催收结果,包括:
    对所述通话语音进行扫描,判断所述通话语音是否包含预先设置的目标关键词;
    若所述通话语音中包含有所述目标关键词,则获取携带有同意标识的逾期催收结果;
    若所述通话语音中不包含有所述目标关键词,则获取携带有不同意标识的逾期催收结果。
  18. 如权利要求15所述的非易失性可读存储介质,其特征在于,所述逾期客户名单中 还包括逾期客户的邮箱地址;
    在所述获取逾期催收结果的步骤之后,所述计算机可读指令被一个或多个处理器执行时,使得所述一个或多个处理器还实现如下步骤:
    若所述逾期催收结果携带有同意标识,则基于催收信息模板和逾期客户信息,获取有效催收信息;
    将有效催收信息通过所述逾期客户的联系电话或者逾期客户的邮箱地址发送给对应的终端设备。
  19. 如权利要求15所述的非易失性可读存储介质,其特征在于,在所述获取逾期催收结果的步骤之后,所述计算机可读指令被一个或多个处理器执行时,使得所述一个或多个处理器还实现如下步骤:
    若所述逾期催收结果携带不同意标识,则对所述通话语音进行声纹识别,确定所述通话语音的说话人是否为逾期客户本人;
    若所述通话语音的说话人为逾期客户本人,则调用预先存储的情绪识别模型对所述通话语音进行情绪识别,获取与所述逾期催收结果相对应的情绪识别结果。
  20. 如权利要求19所述的非易失性可读存储介质,其特征在于,所述对所述通话语音进行声纹识别,确定所述通话语音的说话人是否为逾期客户本人,包括:
    对所述通话语音进行预处理,获取待识别语音;
    对所述待识别语音进行语音特征提取,获取所述待识别语音对应的待识别语音特征;
    采用声纹识别模型对所述待识别语音特征和所述逾期客户对应的原始语音特征进行声纹识别,确定所述通话语音的说话人是否为逾期客户本人。
PCT/CN2018/106251 2018-08-02 2018-09-18 逾期账单智能催收方法、装置、计算机设备及存储介质 WO2020024389A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201810870736.2 2018-08-02
CN201810870736.2A CN109064315A (zh) 2018-08-02 2018-08-02 逾期账单智能催收方法、装置、计算机设备及存储介质

Publications (1)

Publication Number Publication Date
WO2020024389A1 true WO2020024389A1 (zh) 2020-02-06

Family

ID=64832876

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2018/106251 WO2020024389A1 (zh) 2018-08-02 2018-09-18 逾期账单智能催收方法、装置、计算机设备及存储介质

Country Status (2)

Country Link
CN (1) CN109064315A (zh)
WO (1) WO2020024389A1 (zh)

Cited By (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111738824A (zh) * 2020-06-16 2020-10-02 中国建设银行股份有限公司 一种账务数据处理方式筛选方法、装置及系统
CN112053097A (zh) * 2020-09-30 2020-12-08 北京百度网讯科技有限公司 一种贷款催收方法、装置、电子设备及存储介质
CN113159901A (zh) * 2021-04-29 2021-07-23 天津狮拓信息技术有限公司 融资租赁业务会话的实现方法和装置
CN113343058A (zh) * 2021-05-31 2021-09-03 平安普惠企业管理有限公司 语音会话监管方法、装置、计算机设备及存储介质
CN113657901A (zh) * 2021-07-23 2021-11-16 上海钧正网络科技有限公司 欠费用户的催收管理方法、系统、终端及介质
CN113724065A (zh) * 2021-08-31 2021-11-30 平安普惠企业管理有限公司 基于流程指导的辅助催收方法、装置、设备及存储介质
CN113724066A (zh) * 2021-08-31 2021-11-30 平安普惠企业管理有限公司 基于语义分析的权益分配方法、装置、设备及存储介质
CN116452323A (zh) * 2023-04-28 2023-07-18 广州市良策网络技术有限公司 风险评估方法、系统、设备及存储介质
CN117319559A (zh) * 2023-11-24 2023-12-29 杭州度言软件有限公司 一种基于智能语音机器人的催收方法与系统

Families Citing this family (36)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109815489A (zh) * 2019-01-02 2019-05-28 深圳壹账通智能科技有限公司 催收信息生成方法、装置、计算机设备和存储介质
CN109859032A (zh) * 2019-01-22 2019-06-07 深圳壹账通智能科技有限公司 基于智能语音的账款催收方法、装置、设备和存储介质
CN109858917A (zh) * 2019-02-13 2019-06-07 苏州意能通信息技术有限公司 一种基于人工智能的反欺诈系统及其方法
CN109949805B (zh) * 2019-02-21 2021-03-23 江苏苏宁银行股份有限公司 基于意图识别及有限状态自动机的智能催收机器人及催收方法
CN110009480A (zh) * 2019-03-06 2019-07-12 平安科技(深圳)有限公司 司法催收路径的推荐方法、装置、介质、电子设备
CN110084636A (zh) * 2019-03-19 2019-08-02 平安普惠企业管理有限公司 提高电话销售效率方法、装置、计算机设备及存储介质
CN110072201B (zh) * 2019-04-15 2022-02-15 平安科技(深圳)有限公司 短信发送方法、装置、计算机设备及存储介质
CN110166639A (zh) * 2019-04-15 2019-08-23 中国平安人寿保险股份有限公司 语音催收方法、装置、计算机设备及存储介质
CN110264335A (zh) * 2019-05-16 2019-09-20 上海维信荟智金融科技有限公司 基于逾期天数的剩余本金电催分配方法及系统
CN110336786B (zh) * 2019-05-23 2022-02-11 平安科技(深圳)有限公司 消息发送方法、装置、计算机设备和存储介质
CN110265062A (zh) * 2019-06-13 2019-09-20 上海指旺信息科技有限公司 基于情绪检测的智能贷后催收方法及装置
CN110322899B (zh) * 2019-06-18 2023-09-22 平安银行股份有限公司 用户智能分类方法、服务器及存储介质
CN110362712A (zh) * 2019-06-18 2019-10-22 中国平安财产保险股份有限公司 文档数据的延续处理方法、装置、设备及介质
CN110378575B (zh) * 2019-06-28 2022-04-01 上海上湖信息技术有限公司 逾期事件回款催收方法及装置、计算机可读存储介质
CN110570295A (zh) * 2019-07-25 2019-12-13 深圳壹账通智能科技有限公司 资源催收方法、装置、计算机设备和存储介质
CN110782335B (zh) * 2019-09-19 2023-08-15 平安科技(深圳)有限公司 基于人工智能处理信贷数据的方法、装置及存储介质
CN110807699B (zh) * 2019-10-12 2022-06-03 上海上湖信息技术有限公司 逾期事件的回款催收方法及装置、计算机可读存储介质
CN110929005A (zh) * 2019-10-18 2020-03-27 平安科技(深圳)有限公司 基于情绪分析的任务跟催方法、装置、设备及存储介质
CN110991806A (zh) * 2019-11-06 2020-04-10 上海百事通信息技术股份有限公司 一种债务清收管理系统
CN110910231A (zh) * 2019-11-06 2020-03-24 上海百事通信息技术股份有限公司 一种债务清收管理平台
CN111028074B (zh) * 2019-12-06 2024-03-19 深圳乐信软件技术有限公司 逾期账单的更新和查询方法、系统、服务器和存储介质
CN111199470A (zh) * 2019-12-12 2020-05-26 北京淇瑀信息科技有限公司 一种贷后管理频次控制方法、装置及电子设备
CN111192136A (zh) * 2019-12-24 2020-05-22 中信百信银行股份有限公司 一种信贷业务的催收方法、装置、电子设备及存储介质
CN111243605A (zh) * 2020-01-13 2020-06-05 北京三快在线科技有限公司 业务处理方法、装置、设备及存储介质
CN111260466A (zh) * 2020-01-20 2020-06-09 北京合信力科技有限公司 一种触达任务的处理方法及装置
CN111294470B (zh) * 2020-02-05 2021-06-29 北京三快在线科技有限公司 呼叫处理方法、装置、设备及存储介质
CN111461505A (zh) * 2020-03-18 2020-07-28 黑龙江正泽信服务外包有限公司 一种催收信息智能匹配的人工标注系统
CN111833076A (zh) * 2020-07-07 2020-10-27 信雅达系统工程股份有限公司 智能催收系统及方法
CN111835926A (zh) * 2020-07-23 2020-10-27 中国平安财产保险股份有限公司 基于语音交互的智能语音外呼方法、装置、设备及介质
CN112217877A (zh) * 2020-09-23 2021-01-12 上海维信荟智金融科技有限公司 基于微服务架构的催收分案方法及系统
CN112435117A (zh) * 2020-11-27 2021-03-02 平安普惠企业管理有限公司 贷款逾期案件的催收方法、装置、电子设备及存储介质
CN113284497A (zh) * 2020-12-31 2021-08-20 一汽资本控股有限公司 一种向客户进行催收的方法、装置以及智能催收系统
CN112836037A (zh) * 2021-03-26 2021-05-25 中国工商银行股份有限公司 话术推荐方法及装置
CN113824828B (zh) * 2021-10-29 2024-03-08 河北科燃信息科技股份有限公司 一种拨话方法、装置、电子设备及计算机可读存储介质
CN114399379A (zh) * 2022-01-11 2022-04-26 平安普惠企业管理有限公司 基于人工智能的催收行为识别方法、装置、设备及介质
CN116341879B (zh) * 2023-05-26 2024-05-31 杭州度言软件有限公司 一种逾期资产清收智能分案方法与系统

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US8301523B1 (en) * 2007-10-29 2012-10-30 West Corporation System, method and computer readable medium for providing notifications
CN107194807A (zh) * 2017-06-29 2017-09-22 喀什博雅成信网络科技有限公司 一种贷款智能催收系统及方法
CN107240013A (zh) * 2017-04-19 2017-10-10 中国建设银行股份有限公司 一种信用卡催收还款的方法和装置
CN107507088A (zh) * 2017-08-29 2017-12-22 深圳市佰仟金融服务有限公司 催收管理方法及服务器
CN108090826A (zh) * 2017-11-13 2018-05-29 平安科技(深圳)有限公司 一种电话催收方法及终端设备

Family Cites Families (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106572067B (zh) * 2015-10-12 2020-05-12 阿里巴巴集团控股有限公司 语音流传送的方法及系统
CN106683672B (zh) * 2016-12-21 2020-04-03 竹间智能科技(上海)有限公司 一种基于情感和语义的智能对话方法及系统

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US8301523B1 (en) * 2007-10-29 2012-10-30 West Corporation System, method and computer readable medium for providing notifications
CN107240013A (zh) * 2017-04-19 2017-10-10 中国建设银行股份有限公司 一种信用卡催收还款的方法和装置
CN107194807A (zh) * 2017-06-29 2017-09-22 喀什博雅成信网络科技有限公司 一种贷款智能催收系统及方法
CN107507088A (zh) * 2017-08-29 2017-12-22 深圳市佰仟金融服务有限公司 催收管理方法及服务器
CN108090826A (zh) * 2017-11-13 2018-05-29 平安科技(深圳)有限公司 一种电话催收方法及终端设备

Cited By (15)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111738824A (zh) * 2020-06-16 2020-10-02 中国建设银行股份有限公司 一种账务数据处理方式筛选方法、装置及系统
CN112053097A (zh) * 2020-09-30 2020-12-08 北京百度网讯科技有限公司 一种贷款催收方法、装置、电子设备及存储介质
CN113159901A (zh) * 2021-04-29 2021-07-23 天津狮拓信息技术有限公司 融资租赁业务会话的实现方法和装置
CN113159901B (zh) * 2021-04-29 2024-06-04 天津狮拓信息技术有限公司 融资租赁业务会话的实现方法和装置
CN113343058B (zh) * 2021-05-31 2024-02-27 璟滔信息科技(昆山)有限公司 语音会话监管方法、装置、计算机设备及存储介质
CN113343058A (zh) * 2021-05-31 2021-09-03 平安普惠企业管理有限公司 语音会话监管方法、装置、计算机设备及存储介质
CN113657901A (zh) * 2021-07-23 2021-11-16 上海钧正网络科技有限公司 欠费用户的催收管理方法、系统、终端及介质
CN113657901B (zh) * 2021-07-23 2024-04-16 上海钧正网络科技有限公司 欠费用户的催收管理方法、系统、终端及介质
CN113724066A (zh) * 2021-08-31 2021-11-30 平安普惠企业管理有限公司 基于语义分析的权益分配方法、装置、设备及存储介质
CN113724065B (zh) * 2021-08-31 2023-12-29 华道数据处理(苏州)有限公司 基于流程指导的辅助催收方法、装置、设备及存储介质
CN113724065A (zh) * 2021-08-31 2021-11-30 平安普惠企业管理有限公司 基于流程指导的辅助催收方法、装置、设备及存储介质
CN116452323B (zh) * 2023-04-28 2024-01-30 广州市良策网络技术有限公司 风险评估方法、系统、设备及存储介质
CN116452323A (zh) * 2023-04-28 2023-07-18 广州市良策网络技术有限公司 风险评估方法、系统、设备及存储介质
CN117319559A (zh) * 2023-11-24 2023-12-29 杭州度言软件有限公司 一种基于智能语音机器人的催收方法与系统
CN117319559B (zh) * 2023-11-24 2024-02-02 杭州度言软件有限公司 一种基于智能语音机器人的催收方法与系统

Also Published As

Publication number Publication date
CN109064315A (zh) 2018-12-21

Similar Documents

Publication Publication Date Title
WO2020024389A1 (zh) 逾期账单智能催收方法、装置、计算机设备及存储介质
CN111028827B (zh) 基于情绪识别的交互处理方法、装置、设备和存储介质
US10630840B1 (en) Systems for transitioning telephony-based and in-person servicing interactions to and from an artificial intelligence (AI) chat session
CN109949805B (zh) 基于意图识别及有限状态自动机的智能催收机器人及催收方法
US10410636B2 (en) Methods and system for reducing false positive voice print matching
US20230334034A1 (en) Systems and methods for providing searchable customer call indexes
US8145562B2 (en) Apparatus and method for fraud prevention
WO2021184837A1 (zh) 诈骗电话的识别方法、装置、存储介质及终端
CN111696558A (zh) 智能外呼方法、装置、计算机设备及存储介质
EP2840767A1 (en) Speaker recognition and voice tagging for improved service
Shim et al. Phonetic analytics technology and big data: real-world cases
CN113064983B (zh) 语义检测方法、装置、计算机设备及存储介质
CN114398512A (zh) 一种基于大数据的通信运营商行业客户语音画像分析方法
EP4055592B1 (en) Systems and methods for customer authentication based on audio-of-interest
TWI714090B (zh) 機器人電話行銷系統及其計算機裝置與回應訊息產生方法
CN111833068A (zh) 基于声纹识别的身份核验系统及方法
US10446138B2 (en) System and method for assessing audio files for transcription services
US11841932B2 (en) System and method for updating biometric evaluation systems
CN113127673B (zh) 一种声纹数据库的构建方法及其数据调用方法
CN111916073B (zh) 机器人外呼控制方法和装置、服务器、计算机可读存储介质
RU2790946C1 (ru) Способ и система анализа голосовых вызовов на предмет выявления и предотвращения социальной инженерии
CN112765331B (zh) 对话知识模板构建方法、装置、计算机设备及存储介质
WO2023173966A1 (zh) 语音识别方法、终端设备及计算机可读存储介质
US20230385893A1 (en) Ambient listening system for sales assistance
Roshan et al. Capturing important information from an audio conversation

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 18928531

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 18928531

Country of ref document: EP

Kind code of ref document: A1