A collection device based on intelligent mobile terminal

Abstract

The present invention provides a debt collection device based on an intelligent mobile terminal, which belongs to the field of mobile communication technology, and specifically comprises: a speech-to-text conversion module is responsible for acquiring the conversation voice of a debt collector in real time and converting it into speech-to-text; a sensitive word recognition module is responsible for constructing a text record of the debt collector based on speech-to-text, communication records of instant messaging software, and text message records, and performing sensitive word recognition; a compliance evaluation module is responsible for evaluating the compliance of the debt collector; a communication frequency determination module is responsible for determining the communication frequency between the debt collector and the debtor; a repayment willingness determination module is responsible for obtaining the debtor's repayment willingness based on the number of positive keywords, the number of collection times, the weight of the positive keywords, and the debtor's repayment willingness; a result output module is responsible for warning the debt collector or suspending the collection process according to the compliance and determining the debtor's collection frequency according to the debtor's repayment willingness, thereby further reducing the compliance risk in the collection and the pertinence and efficiency of the collection.

Description

Technical Field

[0001] The present invention belongs to the field of mobile communication technology, and in particular relates to a debt collection device based on an intelligent mobile terminal. Background Art

[0002] In order to identify the illegal behaviors of debt collectors during the debt collection process, the invention patent authorization announcement number CN109660679B "Monitoring method, device, equipment and storage medium for debt collection seat end" collects voice information generated by debt collection seats; determines the target debt collection seat to be assessed according to the received assessment request, and obtains the target voice information and work status information of the target debt collection seat from the voice information; processes the target voice information according to the preset scoring rules to obtain the quality inspection score of the target debt collector, and calculates the work saturation of the target debt collector according to the work status information; obtains the monitoring result of the target debt collector according to the quality inspection score and the work saturation, but there are the following technical problems:

[0003] 1. The sources of monitoring data information are limited. In the actual collection process, the collection personnel may also contact the debtor through instant messaging software, text messages, mobile phones, etc. to collect the debts. If the above call records, communication records, and text message records cannot be monitored and analyzed in real time, there is also the possibility of illegal collection and other behaviors.

[0004] 2. The evaluation of the debtor's willingness to repay based on the identification results of call records, communication records, text message records, etc. is not considered, and the collection frequency is determined based on the evaluation results. For users with low repayment willingness, if the collection frequency cannot be increased, the collection effect is unlikely to be satisfactory.

[0005] In view of the above technical problems, the present invention provides a debt collection device based on a smart mobile terminal. Summary of the invention

[0006] To achieve the purpose of the present invention, the present invention adopts the following technical solutions:

[0007] According to one aspect of the present invention, a debt collection method based on a smart mobile terminal is provided.

[0008] A debt collection method based on a smart mobile terminal, characterized by comprising:

[0009] S11 obtains the conversation voice of the debt collector in real time, and converts the conversation voice into voice text through the FBank feature and MFCC feature of the conversation voice;

[0010] S12 constructs a text record based on the voice text, the communication record of the instant messaging software, and the text message record, determines the sensitive communication according to the recognition result of the sensitive words in the text record, and determines the weight of the sensitive communication according to the type of sensitive words in the sensitive communication and the number of different types of sensitive words, the communication time, and the proportion of communication sentences including sensitive words, and divides the sensitive communication into general sensitive communication and concerned sensitive communication according to the weight;

[0011] S13 evaluates the compliance of the debt collector based on the number and proportion of general sensitive communications, the number and proportion of sensitive communications, and the proportion and weight of sensitive communications;

[0012] S14 regards historical communications excluding sensitive communications as effective communications, and obtains positive keywords based on the matching results of the text records of the effective communications with the keyword database, and determines the basic weights of the positive keywords according to the collection time of the text records where the positive keywords are located, obtains the debtor's willingness to repay based on the number of positive keywords, the number of communications, and the sum of the basic weights of the positive keywords, and determines the debtor's collection frequency based on the debtor's willingness to repay.

[0013] By converting call speech into speech text based on both FBank features and MFCC features, call speech can be analyzed from multiple angles, ensuring the accuracy and comprehensiveness of the speech-to-text analysis results.

[0014] By identifying sensitive words and evaluating compliance, we can avoid the occurrence of violations caused by debt collectors' illegal debt collection, ensure the compliance and legality of debt collection, and also achieve accurate evaluation and effective management of debt collectors.

[0015] The debtor's willingness to repay is obtained based on the number of the positive keywords, the number of communications, and the basic weights of the positive keywords, and the debtor's collection frequency is determined based on the debtor's willingness to repay, thereby achieving an evaluation of the repayment willingness from multiple angles, ensuring the accuracy and comprehensiveness of the repayment willingness evaluation, and confirming the debtor's collection frequency based on the repayment willingness, which also improves the final collection efficiency and targeting.

[0016] On the other hand, the present invention provides a debt collection device based on a smart mobile terminal, which adopts the above-mentioned debt collection method based on a smart mobile terminal, including a speech-to-text conversion module, a sensitive word recognition module, a compliance assessment module, a communication frequency determination module, a repayment willingness determination module, and a result output module;

[0017] The voice-to-text conversion module is responsible for acquiring the conversation voice of the debt collector in real time and converting the conversation voice into voice-to-text;

[0018] The sensitive word recognition module is responsible for constructing text records based on the voice text, communication records of instant messaging software, and SMS records, and performing sensitive word recognition based on the text records;

[0019] The compliance assessment module is responsible for assessing the compliance of the debt collector;

[0020] The communication frequency determination module is responsible for determining the communication frequency between the debt collector and the debtor;

[0021] The repayment willingness determination module is responsible for obtaining the repayment willingness of the debtor based on the number of the positive keywords, the number of communications, and the sum of the basic weights of the positive keywords;

[0022] The result output module is responsible for warning the debt collector or suspending the debt collection process according to the compliance level and determining the debt collection frequency of the debtor according to the debtor's willingness to repay.

[0023] Other features and advantages will be described in the following description, and partly become apparent from the description, or understood by practicing the invention. The purpose and other advantages of the invention are realized and obtained by the structures particularly pointed out in the description and the drawings.

[0024] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, preferred embodiments are given below and described in detail with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] The above and other features and advantages of the present invention will become more apparent by describing in detail exemplary embodiments thereof with reference to the attached drawings.

[0026] Figure 1 It is a flow chart of a collection method based on a smart mobile terminal;

[0027] Figure 2 is a flowchart of specific steps for converting the call speech into speech text;

[0028] Figure 3 It is a flowchart of the specific steps for determining the weight of sensitive communication;

[0029] Figure 4 It is a flowchart of the specific steps of assessing the compliance of debt collectors;

[0030] Figure 5 It is a flowchart of the specific steps of evaluating the debtor's willingness to repay;

[0031] Figure 6 The present invention is a framework diagram of a debt collection device based on a smart mobile terminal. DETAILED DESCRIPTION

[0032] Example embodiments will now be described more fully with reference to the accompanying drawings. However, example embodiments can be implemented in a variety of forms and should not be construed as limited to the embodiments set forth herein; rather, these embodiments are provided so that the present invention will be comprehensive and complete and fully convey the concepts of the example embodiments to those skilled in the art. The same reference numerals in the figures represent the same or similar structures, and thus their detailed description will be omitted.

[0033] The terms "a", "an", "the", and "said" are used to indicate the presence of one or more elements / components / etc.; the terms "comprising" and "having" are used to express an open-ended inclusive meaning and mean that additional elements / components / etc. may be present in addition to the listed elements / components / etc.

[0034] To solve the above problems, according to one aspect of the present invention, Figure 1 As shown, a debt collection method based on a smart mobile terminal is provided, which is characterized by specifically including:

[0035] S11 obtains the conversation voice of the debt collector in real time, and converts the conversation voice into voice text through the FBank feature and MFCC feature of the conversation voice;

[0036] Specifically, Figure 2 As shown, the specific steps of converting the call voice into voice text are:

[0037] S21 uses a first-order high-pass filter to perform pre-emphasis processing based on the call speech, and converts the pre-emphasized call speech into a spectrum signal, performs modulus square on the spectrum signal to obtain a speech signal power spectrum, and obtains FBank features through a Mel filter group and a logarithmic operation, and based on the FBank features, uses an acoustic model based on the BIRNN algorithm to predict an output sequence, and based on the output sequence, uses a CTC decoder and a beam search to perform decoding to obtain a speech and text recognition result;

[0038] S22 extracts MFCC features based on the call speech to obtain MFCC features, uses an acoustic model based on a BIGRU algorithm to predict an output sequence, and uses a CTC decoder and a beam search to perform decoding based on the output sequence to obtain a speech and text prediction result;

[0039] S23 performs fusion and de-duplication processing based on the speech-to-text recognition result and the speech-to-text prediction result to obtain speech-to-text.

[0040] By converting call speech into speech text based on both FBank features and MFCC features, call speech can be analyzed from multiple angles, ensuring the accuracy and comprehensiveness of the speech-to-text analysis results.

[0041] S12 constructs a text record based on the voice text, the communication record of the instant messaging software, and the text message record, determines the sensitive communication according to the recognition result of the sensitive words in the text record, and determines the weight of the sensitive communication according to the type of sensitive words in the sensitive communication and the number of different types of sensitive words, the communication time, and the proportion of communication sentences including sensitive words, and divides the sensitive communication into general sensitive communication and concerned sensitive communication according to the weight;

[0042] Specifically, the text records include voice text, communication records of instant messaging software, text information of SMS records, and the collection time corresponding to the text information.

[0043] To give a specific example, the format of the text record is the collection time: 8:00 am on October 11, 2022, and the collection text message: Dear sir, your personal loan has been overdue for 96 days and has affected your credit. It is recommended that you complete the repayment as soon as possible.

[0044] Specifically, keyword extraction is performed based on the text record to obtain text record keywords, and sensitive words are identified according to matching results between the text record keywords and a sensitive word library, where the types of sensitive words include general sensitive words and special sensitive words.

[0045] To give a specific example, keyword extraction based on TF-IDF is essentially a statistical method used to evaluate the importance of a word to a document set or a document in a corpus; the importance of a word increases in direct proportion to the number of times it appears in the document, but decreases in inverse proportion to the frequency of its appearance in the corpus.

[0046] Specifically, the specific steps for keyword extraction based on TF-IDF are:

[0047] Step 1: Calculate word frequency. Considering that articles vary in length, we standardize word frequency to facilitate comparison of different articles.

[0048] Step 2: Calculate the inverse document frequency. If a word is more common, the denominator will be larger and the inverse document frequency will be smaller and closer to 0.

[0049] Step 3: Calculate TF-IDF and determine keywords based on TF-IDF. TF-IDF is proportional to the number of times a word appears in a document and inversely proportional to the number of times the word appears in the entire language.

[0050] To give a specific example, sensitive words such as prosecution, ID number, exposure, etc. that violate the privacy of the debtor or abuse legal provisions are general sensitive words, while sensitive words such as beating, gang beating, threat, etc. that seriously violate legal provisions are special sensitive words.

[0051] Specific examples, such as Figure 3 As shown, the specific steps of determining the weight of the sensitive communication are:

[0052] S31: determining the number of general sensitive words and the number of special sensitive words in the sensitive communication according to the number of sensitive words in the sensitive communication, and determining whether the sensitive communication belongs to the sensitive communication of concern according to the number of general sensitive words and the number of special sensitive words in the sensitive communication; if so, determining the weight of the sensitive communication according to the number of general sensitive words and the number of special sensitive words; if not, proceeding to the next step;

[0053] S32: taking the communication sentences including sensitive words in the sensitive communication as sensitive communication sentences, and determining whether the sensitive communication belongs to the sensitive communication of concern according to the proportion of the sensitive communication sentences, if so, proceeding to step S33, if not, proceeding to step S34;

[0054] S33 divides the sensitive communication sentences into special sensitive communication sentences and general sensitive communication sentences according to the number of general sensitive words and the number of special sensitive words in the sensitive communication sentences, and determines whether the ratio of the special sensitive communication sentences and the ratio of the general sensitive communication sentences meet the requirements. If so, proceed to step S34; if not, determine the weight of sensitive communication according to the ratio of the special sensitive communication sentences and the ratio of the general sensitive communication sentences;

[0055] S34 determines the weight of the sensitive communication by the number of general sensitive words and the number of special sensitive words in the sensitive communication, the ratio of special sensitive communication sentences and the ratio of general sensitive communication sentences, and in combination with the communication time of the sensitive communication.

[0056] Specifically, the weight of sensitive communication is determined by the number of general sensitive words and the number of special sensitive words, which specifically includes:

[0057] Determining the weights of the general sensitive words in the sensitive communication according to the number of the general sensitive words;

[0058] Determine the weight of the special sensitive words in the sensitive communication according to the number of the special sensitive words;

[0059] The weight of the sensitive communication is determined according to the weight of the general sensitive words of the sensitive communication and the weight of the special sensitive words.

[0060] For example, the weight of the sensitive communication is determined by using a prediction model based on the PSO-SVR algorithm, wherein the specific steps of constructing the prediction model are:

[0061] Step 1: Based on historical data, construct a training sample set of collection personnel’s compliance;

[0062] Step 2: Determine the PSO population size sizepop, the maximum number of iterations maxgen, the local optimization index c1, the global optimization index c2, the inertia weight ω, the optimal fitness threshold eps, the penalty coefficient C, and the position threshold of the kernel parameter g [X Cmin ,X Cmax ]、[X gmin ,X gmax ], the relationship coefficient θ between the velocity and position threshold, etc.; the particle position information is set as the potential optimal solution of the SVR parameters (C, g), and initialized in a random manner.

[0063] Step 3: Take the mean square error (MSE) of the SVR model training as the PSO fitness function, and continuously update the optimal positions of particles and populations.

[0064] Step 4: If the maximum number of iterations is reached or the optimal fitness value is met, the optimal value of the SVR parameter (C, g) is output and the process goes to Step 5; otherwise, the particle velocity and position are updated and the process goes to Step 3.

[0065] Step 5: Substitute the optimal values ​​of parameters (C, g) into the SVR model, train the prediction model based on the training set data, and estimate the compliance of the collection personnel based on the cross-sectional prediction information.

[0066] For example, during the optimization process, the PSO parameters are set to: maxgen = 200, sizepop = 20, c1 = 1.5, c2 = 1.7, ω = 1, r1 = r2 = 1, eps = 0.0001, X Cmin =0.1,X Cmax =100,X gmin =0.01, Xgmax=1000, e=0.6.

[0067] Specifically, the PSO algorithm has the advantages of simple model and fast convergence speed, but it also has the problems of premature maturity and easy to fall into local extreme values. The inertia factor ω has a great influence on the performance of the algorithm. The larger the ω, the stronger the global search ability of the PSO algorithm and the weaker the local search ability; conversely, the smaller the ω, the weaker the global search ability of the PSO algorithm and the stronger the local search ability. Therefore, in the algorithm iteration process, the value of ω should change with the iteration. A larger ω value should be used in the initial iteration to ensure the global search ability. As the iteration proceeds, the ω value should be gradually reduced to improve the local search ability of the algorithm. Therefore, the ω value can be associated with the number of iterations:

[0068]

[0069] Where:ω max is the maximum value that ω can take, d is the current number of iterations, and T is the total number of iterations.

[0070] At the same time, in order to prevent the PSO algorithm from falling into a local extreme value, a random variation factor is added to the above formula to ensure that the algorithm can jump out of the local extreme value with a certain probability. Finally, the value of the inertia factor ω of the proposed improved PSO algorithm is shown as follows:

[0071]

[0072] Here, rand(0.9, 1) is a random number ranging from 0.9 to 1.

[0073] Specifically, the compliance degree of the collection personnel ranges from 0 to 1, wherein the smaller the compliance degree of the collection personnel, the worse the evaluation result of the compliance degree of the collection personnel, and a compliance reminder is output to the collection personnel or the collection process is suspended based on the compliance degree of the collection personnel.

[0074] S13 evaluates the compliance of the debt collector based on the number and proportion of general sensitive communications, the number and proportion of sensitive communications, and the proportion and weight of sensitive communications;

[0075] Specific examples, such as Figure 4 As shown, the specific steps of evaluating the compliance of the debt collector are:

[0076] The compliance of the debt collector is evaluated based on the number and proportion of general sensitive communications, the number and proportion of sensitive communications, and the proportion and weight of sensitive communications.

[0077] Obtain the number and proportion of sensitive communications of the collection personnel, and determine whether the compliance of the collection personnel meets the requirements according to the number and proportion of sensitive communications of the collection personnel. If so, determine that the collection personnel is compliant; if not, proceed to the next step;

[0078] Obtain the number and proportion of general sensitive communications and the number and proportion of sensitive communications of concern by the debt collector, and determine whether the compliance of the debt collector meets the requirements based on the number and proportion of general sensitive communications and the number and proportion of sensitive communications of concern by the debt collector; if so, determine that the debt collector is compliant; if not, proceed to the next step;

[0079] Determine the compliance of the general sensitive communication of the debt collector based on the number and proportion of the general sensitive communication of the debt collector, combined with the weight of the general sensitive communication of the debt collector and the compliance of the general sensitive communication of the debt collector;

[0080] Determine the compliance of the collection personnel's sensitive communications based on the number and proportion of the collection personnel's sensitive communications and the weight of the collection personnel's sensitive communications;

[0081] The compliance of the collection staff is determined based on the collection staff's compliance with sensitive communications and general sensitive communications, combined with the number and proportion of the collection staff's sensitive communications and the weight of the collection staff's sensitive communications.

[0082] Specifically, the compliance degree of the collection personnel ranges from 0 to 1, wherein the smaller the compliance degree of the collection personnel, the worse the evaluation result of the compliance degree of the collection personnel, and a compliance reminder is output to the collection personnel or the collection process is suspended based on the compliance degree of the collection personnel.

[0083] S14 obtains positive keywords based on the matching results of the text records and the keyword database, and determines the basic weight of the positive keywords according to the collection time of the text records where the positive keywords are located, obtains the debtor's willingness to repay based on the number of positive keywords, the number of communications, and the sum of the basic weights of the positive keywords, and determines the debtor's collection frequency based on the debtor's willingness to repay.

[0084] For example, the specific steps for constructing the sum of the basic weights of the keywords are:

[0085] S41 determines the time difference between the collection time and the current time based on the collection time of the text record where the positive keyword is located, and obtains the basic weight value of the positive keyword based on the time difference;

[0086] S42 obtains the sum of the basic weights of the positive keywords based on the sum of the basic weights of all positive keywords.

[0087] For example, Figure 5 As shown in Figure 1, the specific steps for evaluating the debtor's willingness to repay are:

[0088] S51: taking historical communications excluding sensitive communications as effective communications, and classifying the effective communications into positive effective communications, neutral effective communications, and negative effective communications according to the number of positive keywords in the effective communications;

[0089] S52 obtains the number of negative effective communications and the ratio of the number of negative effective communications to the number of effective communications, and determines the repayment willingness of the negative effective communications of the debtor in combination with the communication time of the negative effective communications, and determines whether the repayment willingness of the debtor is low through the repayment willingness of the negative effective communications. If so, it is determined that the debtor has a low repayment willingness and the collection frequency is increased. If not, proceed to the next step;

[0090] S53: determining the repayment willingness of the debtor for the positive effective communication according to the number of positive effective communications and the ratio of the number of positive effective communications to the number of effective communications, and in combination with the communication time of the positive effective communication;

[0091] S54 determines the repayment willingness of the debtor with no tendency to communicate effectively based on the number of communications without the tendency to communicate effectively and the ratio of the number of communications without the tendency to communicate effectively to the number of communications with effective communication, and in combination with the communication time of the positive effective communication;

[0092] S55 determines the debtor's willingness to repay through the debtor's negative effective communication willingness to repay, positive effective communication willingness to repay, and neutral effective communication willingness to repay, combined with the debtor's communication times and the basic weights of positive keywords.

[0093] Specifically, the repayment willingness is determined by using a model based on the SVR algorithm, wherein the steps of constructing the model are: for the training data set x i (i=1,2,…,N), SVR can be expressed as

[0094] yi-f(·)≤ε+c,ε≥0

[0095] Where: c is the penalty factor, f(·) is the fitting function, y i is the true value and ε is the relaxation factor.

[0096] The linear function shown in is fitted to its distribution characteristics, and the Lagrange multiplier is introduced to obtain the fitting function f(·) in the form shown in formula (2):

[0097]

[0098] , so SVR uses the kernel function R(x i , x) converts the nonlinear problem in the original space into a linear problem in the high-dimensional space. At this time, the fitting function of the nonlinear problem is as follows:

[0099]

[0100] The kernel function used in the present invention is a Gaussian kernel function, and its specific form is shown below, where σ is a kernel parameter.

[0101]

[0102] SVR has strong adaptability and generalization ability for small sample and nonlinear time series fitting problems, but the prediction performance of SVR is greatly affected by the kernel parameter σ and the penalty factor c. If the parameters are not set properly, the prediction performance of SVR will be significantly reduced; the cross-validation parameter setting method currently widely used has the problem of strong subjectivity and complex calculation.

[0103] To give a specific example, the debtor's willingness to repay ranges from 0 to 1, where the greater the debtor's willingness to repay, the greater the debtor's willingness to repay.

[0104] To give a specific example, when the debtor's willingness to repay is less than a preset willingness value, the debtor's collection frequency is increased. The preset willingness value is determined based on the debtor's overdue amount and overdue time. The larger the debtor's overdue amount and the longer the overdue time, the larger the preset willingness value.

[0105] By identifying sensitive words and evaluating compliance, we can avoid the occurrence of violations caused by debt collectors' illegal debt collection, ensure the compliance and legality of debt collection, and also achieve accurate evaluation and effective management of debt collectors.

[0106] By constructing a prediction model based on the number of positive keywords, the number of collection times, and the weight of the positive keywords, the debtor's willingness to repay is obtained, and the debtor's collection frequency is determined according to the debtor's willingness to repay, thereby achieving an evaluation of the repayment willingness from multiple angles, ensuring the accuracy and comprehensiveness of the repayment willingness evaluation, and confirming the debtor's collection frequency based on the repayment willingness, which also improves the final collection efficiency and targeting.

[0107] like Figure 5 As shown, the present invention provides a debt collection device based on a smart mobile terminal, which adopts the above-mentioned debt collection method based on a smart mobile terminal, including a speech-to-text conversion module, a sensitive word recognition module, a compliance assessment module, a communication frequency determination module, a repayment willingness determination module, and a result output module;

[0108] The voice-to-text conversion module is responsible for acquiring the conversation voice of the debt collector in real time and converting the conversation voice into voice-to-text;

[0109] The sensitive word recognition module is responsible for constructing the text record of the debt collector based on the voice text, communication records of the instant messaging software, and SMS records, and performing sensitive word recognition based on the text record;

[0110] The compliance assessment module is responsible for assessing the compliance of the debt collector;

[0111] The communication frequency determination module is responsible for determining the communication frequency between the debt collector and the debtor;

[0112] The repayment willingness determination module is responsible for constructing a prediction model to obtain the repayment willingness of the debtor based on the number of the positive keywords, the number of collection times, and the weight of the positive keywords;

[0113] The result output module is responsible for warning the debt collector or suspending the debt collection process according to the compliance level and determining the debt collection frequency of the debtor according to the debtor's willingness to repay.

[0114] In several embodiments provided in the present application, it should be understood that the disclosed system and method can also be implemented in other ways. The system embodiments described above are merely schematic. For example, the flowcharts and block diagrams in the accompanying drawings show the possible architecture, functions and operations of the systems, methods and computer program products according to multiple embodiments of the present invention. In this regard, each box in the flowchart or block diagram can represent a module, a program segment or a part of the code, and a part of the module, program segment or code contains one or more executable instructions for implementing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order from the order marked in the accompanying drawings. For example, two consecutive boxes can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram and / or flowchart, and the combination of boxes in the block diagram and / or flowchart can be implemented with a dedicated hardware-based system that performs a specified function or action, or can be implemented with a combination of dedicated hardware and computer instructions.

[0115] In addition, the functional modules in the various embodiments of the present invention may be integrated together to form an independent part, or each module may exist independently, or two or more modules may be integrated to form an independent part.

[0116] If the function is implemented in the form of a software function module and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the methods of each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc. Various media that can store program codes.

[0117] Based on the above ideal embodiments of the present invention, the relevant staff can make various changes and modifications without departing from the technical concept of the present invention through the above description. The technical scope of the present invention is not limited to the contents of the specification, and its technical scope must be determined according to the scope of the claims.

Claims

1. A debt collection method based on a smart mobile terminal, characterized in that: Specifically include: Acquire the conversation voice of the debt collector in real time, and convert the conversation voice into speech text through the FBank feature and MFCC feature of the conversation voice; Constructing text records based on the voice text, communication records of instant messaging software, and text information of SMS records, determining sensitive communication according to the recognition results of sensitive words in the text records, and determining the weight of sensitive communication according to the types of sensitive words in the sensitive communication and the number of different types of sensitive words, communication time, and the proportion of communication sentences including sensitive words, and dividing sensitive communication into general sensitive communication and concerned sensitive communication according to the weight; The compliance of the debt collector is evaluated by the number and proportion of general sensitive communications, the number and proportion of sensitive communications, and the proportion and weight of the sensitive communications; The historical communications excluding sensitive communications are regarded as effective communications, and positive keywords are obtained based on the matching results of the text records of the effective communications with the keyword database, and the basic weights of the positive keywords are determined according to the collection time of the text records where the positive keywords are located, and the repayment willingness of the debtor is obtained based on the number of the positive keywords, the number of communications, and the sum of the basic weights of the positive keywords, and the collection frequency of the debtor is determined according to the repayment willingness of the debtor; Extracting keywords based on the text record to obtain text record keywords, and identifying sensitive words according to the matching results of the text record keywords and the sensitive word library, where the types of sensitive words include general sensitive words and special sensitive words; The specific steps of determining the weight of the sensitive communication are: S31: determining the number of general sensitive words and the number of special sensitive words in the sensitive communication according to the number of sensitive words in the sensitive communication, and determining whether the sensitive communication belongs to the sensitive communication of concern according to the number of general sensitive words and the number of special sensitive words in the sensitive communication; if so, determining the weight of the sensitive communication according to the number of general sensitive words and the number of special sensitive words; if not, proceeding to the next step; S32: taking the communication sentences including sensitive words in the sensitive communication as sensitive communication sentences, and determining whether the sensitive communication belongs to the sensitive communication of concern according to the proportion of the sensitive communication sentences, if so, proceeding to step S33, if not, proceeding to step S34; S33 divides the sensitive communication sentences into special sensitive communication sentences and general sensitive communication sentences according to the number of general sensitive words and the number of special sensitive words in the sensitive communication sentences, and determines whether the ratio of the special sensitive communication sentences and the ratio of the general sensitive communication sentences meet the requirements. If so, proceed to step S34; if not, determine the weight of sensitive communication according to the ratio of the special sensitive communication sentences and the ratio of the general sensitive communication sentences; S34 determines the weight of the sensitive communication according to the number of general sensitive words and the number of special sensitive words in the sensitive communication, the ratio of special sensitive communication sentences and the ratio of general sensitive communication sentences, and the communication time of the sensitive communication; The specific steps of evaluating the compliance of the debt collectors are as follows: Obtain the number and proportion of sensitive communications of the collection personnel, and determine whether the compliance of the collection personnel meets the requirements according to the number and proportion of sensitive communications of the collection personnel. If so, determine that the collection personnel is compliant; if not, proceed to the next step; Obtain the number and proportion of general sensitive communications and the number and proportion of sensitive communications of concern by the debt collector, and determine whether the compliance of the debt collector meets the requirements based on the number and proportion of general sensitive communications and the number and proportion of sensitive communications of concern by the debt collector; if so, determine that the debt collector is compliant; if not, proceed to the next step; Determine the compliance of the general sensitive communication of the debt collector based on the number and proportion of the general sensitive communication of the debt collector, combined with the weight of the general sensitive communication of the debt collector and the compliance of the general sensitive communication of the debt collector; Determine the compliance of the collection personnel's sensitive communications based on the number and proportion of the collection personnel's sensitive communications and the weight of the collection personnel's sensitive communications; The compliance of the collection personnel is determined based on the compliance of the collection personnel's sensitive communications and the compliance of general sensitive communications, and in combination with the number and proportion of the collection personnel's sensitive communications and the weight of the collection personnel's sensitive communications; The specific steps for assessing the debtor's willingness to repay are: The historical communications excluding sensitive communications are regarded as effective communications, and the effective communications are divided into positive effective communications, neutral effective communications, and negative effective communications according to the number of positive keywords in the effective communications; Obtain the number of negative effective communications and the ratio of the number of negative effective communications to the number of effective communications, and determine the debtor's willingness to repay the negative effective communications in combination with the communication time of the negative effective communications, and determine whether the debtor's willingness to repay is low through the willingness to repay of the negative effective communications. If so, determine that the debtor's willingness to repay is low and increase the frequency of collection. If not, proceed to the next step; Determine the debtor's willingness to repay the positive and effective communication based on the number of positive and effective communications and the ratio of the number of positive and effective communications to the number of effective communications, combined with the communication time of the positive and effective communications; Determine the debtor's willingness to repay without an inclination for effective communication based on the number of communications without an inclination for effective communication and the ratio of the number of communications without an inclination for effective communication to the number of communications for effective communication, combined with the communication time of the positive effective communication; The debtor's willingness to repay is determined by the debtor's willingness to repay through negative effective communication, positive effective communication, and neutral effective communication, combined with the debtor's communication times and the basic weights of positive keywords.

2. The debt collection method based on the intelligent mobile terminal according to claim 1, characterized in that: The specific steps of converting the call voice into voice text are: The call speech is pre-emphasized by a first-order high-pass filter, and the pre-emphasized call speech is converted into a spectrum signal, the spectrum signal is modulo squared to obtain a speech signal power spectrum, and FBank features are obtained by a Mel filter group and a logarithmic operation, based on the FBank features, an acoustic model based on the BIRNN algorithm is used to predict an output sequence, and based on the output sequence, a CTC decoder is used to perform decoding by a beam search to obtain a speech and text recognition result; Extracting MFCC features based on the call speech to obtain MFCC features, using an acoustic model based on a BIGRU algorithm to predict an output sequence, and based on the output sequence, using a CTC decoder and a beam search to perform decoding to obtain a speech and text prediction result; Based on the speech-to-text recognition result and the speech-to-text prediction result, fusion and de-duplication processing are performed to obtain speech-to-text.

3. The debt collection method based on the intelligent mobile terminal according to claim 1, characterized in that: The compliance degree of the collection personnel ranges from 0 to 1, wherein the smaller the compliance degree of the collection personnel, the worse the evaluation result of the compliance degree of the collection personnel is, and a compliance reminder is output to the collection personnel or the collection process is suspended based on the compliance degree of the collection personnel.

4. The debt collection method based on the intelligent mobile terminal according to claim 1, characterized in that: The specific steps of constructing the basic weights of the keywords are: Based on the collection time of the text record where the positive keyword is located, determining the time difference between the collection time and the current time, and obtaining the basic weight value of the positive keyword based on the time difference; Based on the sum of the basic weights of all positive keywords, the sum of the basic weights of the positive keywords is obtained.

5. The debt collection method based on a smart mobile terminal according to claim 1, characterized in that: When the debtor's repayment willingness is less than a preset willingness value, the debtor's collection frequency is increased. The preset willingness value is determined based on the debtor's overdue amount and overdue time. The larger the debtor's overdue amount and the longer the overdue time, the larger the preset willingness value.

6. A debt collection device based on a smart mobile terminal, using a debt collection method based on a smart mobile terminal as described in any one of claims 1 to 5, comprising a speech-to-text conversion module, a sensitive word recognition module, a compliance assessment module, a communication frequency determination module, a repayment willingness determination module, and a result output module; The voice-to-text conversion module is responsible for acquiring the conversation voice of the debt collector in real time and converting the conversation voice into voice-to-text; The sensitive word recognition module is responsible for constructing text records based on the voice text, communication records of instant messaging software, and SMS records, and performing sensitive word recognition based on the text records; The compliance assessment module is responsible for assessing the compliance of the debt collector; The communication frequency determination module is responsible for determining the communication frequency between the debt collector and the debtor; The repayment willingness determination module is responsible for obtaining the repayment willingness of the debtor based on the number of the positive keywords, the number of communications, and the sum of the basic weights of the positive keywords; The result output module is responsible for warning the debt collector or suspending the debt collection process according to the compliance level and determining the debt collection frequency of the debtor according to the debtor's willingness to repay.

Images