Work order generation method and device based on natural language processing, equipment and medium

Abstract

The invention relates to the technical field of artificial intelligence and can be applied to the field of financial science and technology. The invention discloses a work order generation method and device based on natural language processing, equipment and a medium. The method comprises the following steps: acquiring voice packet data of an incoming call of a user, extracting acoustic characteristics in the voice packet data, and determining an emotional state of the user according to the acoustic characteristics; converting the voice packet data into text content through a voice-to-text algorithm; performing named entity recognition on the text content to obtain a user incoming call type, an incoming call transaction type and user basic data; generating a pre-filling work order according to the user emotional state, the user incoming call type, the incoming call transaction type and the user basic data; and according to the pre-filled work order, notifying a corresponding person in charge to perform work order processing. By automatically generating the pre-filling work order, the skill requirement on customer service can be reduced, the service processing efficiency is improved, and the user satisfaction is ensured.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to a work order generation method, device, equipment and medium based on natural language processing. Background Art

[0002] When users encounter problems in transaction processing, they often choose to call the customer service center for business assistance. Currently, when a customer service center of a financial institution receives a call from a user, it usually creates a work order manually based on memory or repeated listening to the voice. Then, the corresponding work order person is found based on the knowledge base to allocate and handle the work order.

[0003] This process is time-consuming and error-prone, and requires high standards from customer service personnel, who need to undergo multiple rounds of professional training. Customer service personnel are frequently rotated, and the efficiency of processing varies greatly due to individual abilities. Human factors can easily lead to problems such as inaccurate information records and incorrect classification of work order records.

[0004] In order to improve the efficiency of customer service work order processing, it is urgent to propose an automated customer service work order generation method. Summary of the invention

[0005] The present invention provides a work order generation method, device, equipment and medium based on natural language processing to improve business processing efficiency and ensure user satisfaction.

[0006] According to one aspect of the present invention, a work order generation method based on natural language processing is provided, the method comprising:

[0007] Acquire voice packet data of an incoming call from a user, extract acoustic features from the voice packet data, and determine the user's emotional state based on the acoustic features;

[0008] Converting the voice packet data into text content through a voice-to-text algorithm;

[0009] Performing named entity recognition on the text content to obtain the user's incoming call type, incoming call transaction type and user basic data;

[0010] Generate a pre-filled work order based on the user's emotional state, the user's incoming call type, the incoming call transaction type, and the user's basic data;

[0011] Notify the corresponding person in charge to process the work order according to the pre-filled work order.

[0012] According to another aspect of the present invention, a work order generation device based on natural language processing is provided, the device comprising:

[0013] A user emotional state determination module, used to obtain voice packet data of a user's incoming call, extract acoustic features from the voice packet data, and determine the user's emotional state based on the acoustic features;

[0014] A text conversion module, used to convert the voice packet data into text content through a voice-to-text algorithm;

[0015] A named entity recognition module is used to perform named entity recognition on the text content to obtain the user's incoming call type, incoming call transaction type and user basic data;

[0016] A pre-filled work order generation module, used to generate a pre-filled work order according to the user's emotional state, the user's incoming call type, the incoming call transaction type and the user's basic data;

[0017] The work order notification module is used to process the work order according to the pre-filled work order notification corresponding to the person in charge.

[0018] According to another aspect of the present invention, an electronic device is provided, the electronic device comprising:

[0019] at least one processor; and

[0020] a memory communicatively connected to the at least one processor; wherein,

[0021] The memory stores a computer program that can be executed by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor can execute the work order generation method based on natural language processing described in any embodiment of the present invention.

[0022] According to another aspect of the present invention, a computer-readable storage medium is provided, wherein the computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a processor to implement the work order generation method based on natural language processing described in any embodiment of the present invention when executed.

[0023] According to another aspect of the present invention, a computer program product is provided, including a computer program, wherein when the computer program is executed by a processor, the method for generating a work order based on natural language processing according to any embodiment of the present invention is implemented.

[0024] The technical solution of the embodiment of the present invention obtains voice package data of user incoming calls, extracts acoustic features in the voice package data, and determines the user's emotional state based on the acoustic features; converts the voice package data into text content through a speech-to-text algorithm; performs named entity recognition on the text content to obtain the user's incoming call type, the incoming call transaction type, and the user's basic data; generates a pre-filled work order based on the user's emotional state, the user's incoming call type, the incoming call transaction type, and the user's basic data; and notifies the corresponding person in charge to process the work order based on the pre-filled work order, which solves the problems of large human factors and low efficiency in the work order processing of the prior art. By automatically generating pre-filled work orders, the skill requirements for customer service can be reduced, the business processing efficiency can be improved, and user satisfaction can be guaranteed.

[0025] It should be understood that the contents described in this section are not intended to identify the key or important features of the embodiments of the present invention, nor are they intended to limit the scope of the present invention. Other features of the present invention will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0027] Figure 1 is a flowchart of a work order generation method based on natural language processing provided according to Embodiment 1 of the present invention;

[0028] Figure 2 This is an application diagram of a work order generation method based on natural language processing provided in Embodiment 1 of the present invention;

[0029] Figure 3 is a flowchart of a method for generating a work order based on natural language processing according to Embodiment 2 of the present invention;

[0030] Figure 4 is a structural schematic diagram of a work order generation device based on natural language processing provided according to Embodiment 3 of the present invention;

[0031] Figure 5 It is a structural schematic diagram of an electronic device for implementing the work order generation method based on natural language processing according to an embodiment of the present invention. DETAILED DESCRIPTION

[0032] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0033] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0034] Embodiment 1

[0035] Figure 1 This is a flowchart of a method for generating a work order based on natural language processing according to the first embodiment of the present invention. This embodiment can be applied to the situation where customer service in financial technology handles user calls. The method can be executed by a work order generating device based on natural language processing. The work order generating device based on natural language processing can be implemented in the form of hardware and / or software. The work order generating device based on natural language processing can be configured in an electronic device such as a computer. Figure 1 As shown, the method includes:

[0036] Step 110: Acquire voice packet data of an incoming call from a user, extract acoustic features from the voice packet data, and determine the user's emotional state based on the acoustic features.

[0037] Among them, voice package data can be the voice data generated when a user calls a customer service center when he needs manual assistance for financial issues. During the telephone communication, the user-related information collected is the information and data authorized by the user or fully authorized by all parties, and the method of acquisition is reasonable and legal.

[0038] When the voice package data is obtained, the acoustic features in the voice package data can be extracted. Acoustic features include but are not limited to frequency, speech rate change, intonation, volume, and the use of specific vocabulary. The user's emotional state can be obtained based on the analysis of the acoustic features. For example, when the speech rate changes from slow to fast, it can be determined that the user's emotion begins to change from calm to angry; for another example, when the volume changes from low to high, it can be determined that the user's emotion begins to change from calm to angry; for another example, when the number of times a specific vocabulary is used exceeds a set number, it can be determined that the user's emotion is very angry. When determining the user's emotional state, multiple acoustic features can be integrated to determine the final emotional state. Among them, specific vocabulary can be vocabulary that can reflect the user's emotions. The user's emotional state includes but is not limited to: very angry, generally angry, calm, and happy.

[0039] In an optional implementation of an embodiment of the present invention, acoustic features are extracted from voice package data, and the user's emotional state is determined based on the acoustic features, including: extracting frequency domain features, time domain features, and specific vocabulary features of the voice from the voice package data, and using the frequency domain features, time domain features, and specific vocabulary features as acoustic features; wherein the frequency domain features include at least one of the following: Mel-frequency cepstral coefficients and fundamental frequency; the time domain features include at least one of the following: speaking rate and volume; the specific vocabulary features include the frequency of use of specific vocabulary; and the acoustic features are input into a pre-trained recurrent neural network model to obtain the user's emotional state.

[0040] Mel-frequency cepstral coefficients (MFCCs) can be coefficients obtained by converting voice packet data into frequency domain signals through Mel filter banks and cepstral transformation. Mel-frequency cepstral coefficients can reflect the spectral profile of speech and are closely related to emotions. Keynote can reflect changes in intonation and is an important clue for emotion recognition. By calculating the time interval between speech frames and analyzing changes in speech rate, the user's emotional state can be determined. By calculating the energy or amplitude of each frame of speech and evaluating the volume of the speech, the user's emotional state can be determined. The frequency of use of specific words can also assist in determining the user's emotional state. By inputting the Mel-frequency cepstral coefficients, fundamental frequency, speech rate, volume, and the frequency of use of specific words into a pre-trained recurrent neural network model, the user's emotional state can be accurately determined.

[0041] The training of the recurrent neural network model (RNN) can be performed using a pre-measured emotion dataset based on the basic RNN model architecture. During the training process, the loss function is used to quantify the model prediction output P. model (y|x) and the actual observed value P trueA key component of the gap between (y|x). For a given sample x, assuming it belongs to category y, ideally the model should give a probability of this category close to 1 and a probability of other categories close to 0. In RNN model training, the loss function can choose cross-entropy loss for multi-classification tasks. When the predicted probability p is close to the true label y, the loss will be close to 0; conversely, if the prediction is wrong, the loss will be larger. The model parameters can be optimized according to the loss function to reduce the loss value of the model and improve the accuracy of the model.

[0042] The cross entropy loss function formula is L is the model loss value, C is the total number of categories, y c is the true label of the sample belonging to the cth class, and P c is the probability that the model predicts that the sample belongs to the cth class. In the embodiment of the present invention, C=4, y c The real labels include very angry, moderately angry, calm, and happy.

[0043] Before performing acoustic feature extraction, the voice packet data can be preprocessed. Preprocessing includes but is not limited to: signal denoising, framing, and voice activity detection. Among them, noise filtering of the voice packet can remove background noise and determine the correctness of the analysis. Framing can be a fixed-length segmentation of the voice packet data into multiple short time frames. Because long audio will cause computational complexity. Specifically, in order to preprocess the voice packet data, a process including standardized sampling rate, framing, windowing and Fourier transform can be performed to convert the time domain signal into a feature vector suitable for machine learning input. Preprocessing can also be voice activity detection (VAD). VAD can distinguish between speech and non-language segments to retain only valid voice data.

[0044] By determining the user's emotional state based on the voice packet data of the user's incoming call, the user's emotional issues can be taken into account in the subsequent work order processing, so as to determine the specific processing mechanism based on the user's emotion and improve service satisfaction.

[0045] Step 120: Convert the voice packet data into text content through a speech-to-text algorithm.

[0046] The speech-to-text algorithm may be an automatic speech recognition (ASR) algorithm.

[0047] Step 130: Perform named entity recognition on the text content to obtain the user's incoming call type, incoming call transaction type and user basic data.

[0048] Named entity recognition can be used to identify entities related to financial services in text content. There are many ways to perform named entity recognition, such as based on rules, statistical methods, and neural network models. Among them, entities related to financial services can include user call types, call transaction types, and user basic data.

[0049] The types of user calls may include but are not limited to business consultation, complaints, emergency assistance, special services, technical support, service requests, and others. The types of call transactions may include but are not limited to credit cards, personal finance, online finance, settlement and cash management, operation management, inclusive finance, and others. User basic data includes but is not limited to user name, user contact information, user star rating (1 star to 6 stars, or special identification users), user location, urgency (urgent or general), specific event content, work order time, and acceptance channel.

[0050] In an optional implementation of an embodiment of the present invention, named entity recognition is performed on text content to obtain user call types, call transaction types and user basic data, including: classifying and labeling content entities in a customer service corpus through a remote supervision algorithm to obtain preset user call types, call transaction types, and description information of each type; training a named entity recognition model based on a dual affine classifier according to the preset user call types, call transaction types, and description information of each type; performing named entity recognition on the text content through the trained named entity recognition model based on a dual affine classifier to obtain user call types, call transaction types and user basic data.

[0051] Among them, the remote supervision algorithm can be a machine learning algorithm for relationship extraction. In the embodiment of the present invention, the remote supervision algorithm can be used to classify and annotate the content entities in the customer service corpus. Specifically, the annotation information of each label in the user call type can be: Business consultation refers to the user's need to understand the basic information of the financial institution's products, such as deposit interest rates, loan conditions, financial product details, credit card discount activities, foreign exchange rates, etc., or to check the balance, transaction records, points and other account information. Complaints refer to the user's dissatisfaction with the services or products of the financial institution, involving service quality, fee disputes, transaction disputes, etc. Emergency help refers to the user encountering an emergency, such as fraud, card loss, stolen card, etc., contacting customer service for help. Special services refer to special groups such as the elderly, visually impaired people, foreigners, etc. seeking special service help, such as foreign language translation, voice services, etc. Technical support refers to the user's need for technical support to solve problems, which are generally online channel problems, such as login failure, transaction error, application (APP) failure, security authentication failure, etc. Service request refers to the user's need to perform some non-face-to-face online service operations, such as resource account activation, password modification, report loss, unblock, transfer remittance, application for increase, etc. Other user calls that cannot be categorized in detail can be put into the "Other" category.

[0052] The label information of each label in the incoming call transaction type can be: Credit card refers to credit card related, including credit card collection, credit limit, installment payment, card application, card replacement, automobile special installment, report loss, cancellation, merchant acquisition, electronic toll collection system (ETC), points, value-added services, repayment, deduction of arrears, statement, etc. Personal finance refers to personal related, such as personal housing loan, consumer loan, credit consumer loan products, debit card, personal comprehensive points, savings deposit, personal financial management, personal foreign exchange, personal star, ICBC messenger, rights and interests service, etc. Internet finance refers to online channels and marketing activities such as mobile banking, corporate online, e-commerce platform products, public accounts, mini programs, online life APP, etc. Settlement and cash management refers to public related, such as public accounts, bank-to-merchant transfers, self-service receipt printing charges, etc. Operation management refers to branch related, such as branch personnel services, branch window opening, calling, business hours, inter-bank remittances, self-service equipment, etc. Inclusive finance refers to small and micro related, such as small and micro enterprise loans, personal business loans, etc.

[0053] According to the preset user call type, call transaction type, and the description information of each type, the named entity recognition model based on the dual affine classifier is trained as follows: building a named entity recognition model based on the dual affine classifier, Chinese character recognition and annotation, obtaining text sentence level features, decoding text sentence level features, and selecting the entity recognition type with the highest probability. Specifically, based on the BERT (a method for pre-training language representation) context vector and the GloVe (an unsupervised learning algorithm for obtaining word vectors) word vector model, the representation of Chinese characters in the text is obtained, and then the bidirectional long short-term memory network (BiLSTM) encoding is used to obtain the text sentence level features, and finally the dual affine classifier is used to decode the sentence features, generate a classification matrix, and finally select the category with the highest probability. By continuously training and adjusting the model parameters, a named entity recognition model based on the dual affine classifier is finally generated, and the text content is recognized by named entities to obtain the user call type, call transaction type and user basic data.

[0054] Among them, when the user's call type is a complaint, it can be further identified whether the problem has been resolved, whether it is a repeated complaint, and the user's deadline requirements, etc.

[0055] By identifying the type of user calls and the type of call transactions, we can accurately match the corresponding person in charge for the work order, improving the accuracy and efficiency of transaction processing allocation. In addition, named entity recognition based on the biaffine classifier can automatically read, understand and classify the content of the work order submitted by the user without manual screening, reducing the workload of customer service staff. Automated processing reduces labor costs and reduces the burden on customer service staff and increases efficiency. The automated entry process ensures the consistency, standardization and accuracy of the work order records, facilitating subsequent data analysis and management decisions.

[0056] Step 140: Generate a pre-filled work order based on the user's emotional state, the user's incoming call type, the incoming call transaction type, and the user's basic data.

[0057] The work order may be structured data. When generating a pre-filled work order, a work order template may be obtained, and the user's emotional state, the user's incoming call type, the incoming call transaction type, and the user's basic data may be filled into the work order template.

[0058] After the pre-filled work order is generated, the work order number can be further generated. For example, the work order number can be determined based on information such as the work order generation sequence. The pre-filled work order can also be checked for integrity.

[0059] Step 150: Notify the corresponding person in charge to process the work order according to the pre-filled work order.

[0060] When notifying the responsible person, the responsible person can be determined based on the user call type and call transaction type in the pre-filled work order. For example, a responsible person contact table can be pre-built, and the responsible person can be determined based on the mapping relationship in the responsible person contact table, the user call type and the call transaction type.

[0061] In an optional implementation of an embodiment of the present invention, a corresponding person in charge is notified to process the work order according to a pre-filled work order, including: performing an integrity check on the pre-filled work order, and when the check passes, matching the person in charge according to the pre-filled work order; when the match is successful, using a first-level notification method to notify the corresponding person in charge to process the work order; when the match fails, using a second-level notification method to notify the corresponding person in charge to process the work order; when the check fails, using a second-level notification method to notify the corresponding person in charge to process the work order; wherein the first-level notification method has a lower priority than the second-level notification method.

[0062] Among them, the integrity check can be to determine whether the pre-filled work order contains all the preset key information to ensure that the transaction can be processed normally according to the pre-filled work order. When the integrity check is passed, the corresponding responsible outlets and responsible persons can be retrieved and matched according to the key elements such as the responsible outlets mentioned in the user's call matters and the user's affiliated outlets in the system. If the match is successful, the first level notification method can be used to notify the person in charge to process the work order as soon as possible. The first level notification method can be a common notification method. For example, the first level notification method can be a text message and / or email notification method. If the match fails, the second level notification method can be used to notify the corresponding person in charge to process the work order. The second level notification method can be an efficient notification method. For example, the second level notification method can be to notify the on-duty personnel through intelligent outbound calling technology, and the on-duty personnel determine the person in charge according to the pre-filled work order, and then notify the person in charge to process the work order as soon as possible. When the check fails, the corresponding person in charge can also be notified to process the work order through the second level notification method.

[0063] By checking the integrity of the pre-filled work orders and determining whether they can be matched to the corresponding person in charge, and selecting different notification methods, it can be ensured that the pre-filled work orders can be reasonably and promptly matched to the corresponding person in charge for processing, avoiding the omission of work orders due to incomplete pre-filled work orders or failure to match the person in charge, thereby reducing service quality.

[0064] Figure 2 Schematic diagram of an application of a method for generating a work order based on natural language processing according to the first embodiment of the present invention. Figure 2As shown, an exemplary application process of an embodiment of the present invention is: obtaining the voice packet data of the user's incoming call, identifying the user's emotional state based on the RNN model; converting the voice packet data into text content through automatic speech recognition technology. The named entity recognition model based on the dual affine classifier extracts the user's incoming call type and the incoming call transaction type in the text content. Identify the user's basic data through text recognition or named entity recognition. Generate a pre-filled work order based on the user's emotional state, the user's incoming call type, the incoming call transaction type, and the user's basic data. Verify the integrity of the pre-filled work order. When the integrity check passes, the pre-filled work order can be entered into the business pending system and a work order number is generated. In the business pending system, the person in charge can be matched according to the pre-filled work order. If the match is successful, the person in charge is notified to process the work order. If the match fails, or the integrity check fails, the on-duty personnel are notified, and the on-duty personnel determine and notify the person in charge based on the pre-filled work order. Finally, the person in charge processes the pre-filled work order.

[0065] The technical solution of this embodiment obtains voice package data of user calls, extracts acoustic features in the voice package data, and determines the user's emotional state based on the acoustic features; converts the voice package data into text content through a speech-to-text algorithm; performs named entity recognition on the text content to obtain the user's call type, call transaction type, and user basic data; generates a pre-filled work order based on the user's emotional state, user call type, call transaction type, and user basic data; and notifies the corresponding person in charge to process the work order based on the pre-filled work order, which solves the problems of large human factors and low efficiency in work order processing in the prior art. By automatically generating pre-filled work orders, the skill requirements for customer service can be reduced, the business processing efficiency can be improved, and user satisfaction can be guaranteed.

[0066] Embodiment 2

[0067] Figure 3 1 is a flowchart of a method for generating a work order based on natural language processing according to Embodiment 2 of the present invention. This embodiment is a further refinement of the above technical solution. The technical solution in this embodiment can be combined with each optional solution in one or more of the above embodiments. Figure 3 As shown, the method includes:

[0068] Step 310: Acquire voice packet data of an incoming call from a user, extract frequency domain features, time domain features, and specific vocabulary features of the voice from the voice packet data, and use the frequency domain features, time domain features, and specific vocabulary features as acoustic features.

[0069] Among them, the frequency domain features include at least one of the following: Mel frequency cepstral coefficients and fundamental frequency; the time domain features include at least one of the following: speaking rate and volume; and the specific vocabulary features include the frequency of use of specific vocabulary.

[0070] Step 320: Input the acoustic features into a pre-trained recurrent neural network model to obtain the user's emotional state.

[0071] Step 330: Convert the voice packet data into text content through a voice-to-text algorithm.

[0072] Step 340: Perform named entity recognition on the text content through the trained named entity recognition model based on the dual affine classifier to obtain the user's incoming call type, incoming call transaction type and user basic data.

[0073] Optionally, before the named entity recognition model based on the dual affine classifier that has been trained performs named entity recognition on the text content, it also includes: classifying and labeling the content entities in the customer service corpus through a remote supervision algorithm to obtain preset user call types, call transaction types, and description information of each type; training the named entity recognition model based on the dual affine classifier according to the preset user call types, call transaction types, and description information of each type.

[0074] Step 350: Generate a pre-filled work order based on the user's emotional state, the user's incoming call type, the incoming call transaction type, and the user's basic data.

[0075] In an optional implementation of an embodiment of the present invention, a pre-filled work order is generated based on the user's emotional state, the user's call type, the call transaction type, and the user's basic data, including: generating a work order processing level based on the user's emotional state, the user's call type, the call transaction type, the user's basic data, and the preset weights corresponding to each content; generating a work order processing category based on the user's call type and the call transaction type; generating a work order number based on the work order processing level and the work order processing category.

[0076] Among them, when the user's emotional state is more anxious, the work order processing level is relatively high; when the user's emotional state is more moderate, the work order processing level can be relatively low. For example, the user's emotional state is very angry, generally angry, calm, happy, and other conditions are the same. The corresponding work order processing level can be gradually reduced. When the user's call type and call transaction type are emergency categories, the work order processing level is relatively high. For example, a higher work order processing level can be set for user call types such as complaints, emergency assistance, and special services. When the user level or importance in the user basic data is high, the work order processing level is relatively high.

[0077] When generating the work order processing level, different weights can be set according to the user's emotional state, the user's call type, the call transaction type, and the importance of the user's basic data. For example, the preset weight of the user's call type is greater than the preset weight of the user's emotional state, the preset weight of the user's emotional state is greater than the preset weight of the call transaction type, and the preset weight of the call transaction type is greater than the preset weight of the user's basic data. The user's emotional state, the user's call type, the call transaction type, and the specific parameters in the user's basic data can be mapped to different urgency values ​​according to the emergency situation. The work order processing level can be obtained by performing weighted calculation based on the urgency value and the corresponding preset weight.

[0078] When generating a work order number, in addition to considering the work order processing level, the work order processing category can also be considered. The work order processing category can be determined by the user call type and the call transaction type. Exemplarily, the work order processing category can be generated by splicing the user call type and the call transaction type.

[0079] When actually generating the work order number, the work order processing level and the work order processing category can be digitized. For example, the first 3 digits of the work order number are the digitized result of the work order processing category, the middle 2 digits are the digitized result of the work order processing level, and the last 4 digits are the work order sequence number. The work order sequence number can be generated based on the user's call time and / or sequence.

[0080] By generating a work order number that includes a work order processing level and a work order processing category, the practicality of the work order number can be improved. When pre-filling work orders according to the work order number, the rationality of the work order processing sequence can be guaranteed, thereby improving the service quality.

[0081] Step 360: Perform integrity check on the pre-filled work order. When the check passes, match the responsible person based on the pre-filled work order.

[0082] Step 370: When the match is successful, the first level notification method is used to notify the corresponding person in charge to process the work order.

[0083] Step 380: When the matching fails, the second level notification method is used to notify the corresponding person in charge to process the work order.

[0084] Step 390: When the verification fails, the second level notification method is used to notify the corresponding person in charge to process the work order.

[0085] In an optional implementation of an embodiment of the present invention, when notifying a corresponding person in charge to process a work order, it includes: retrieving the associated historical work order according to the work order number, and determining the similarity between the pre-filled work order and the historical work order; based on the similarity, screening similar work orders in the historical work orders; and notifying the corresponding person in charge to process the work order based on the pre-filled work order and the similar work orders.

[0086] When the work order number includes the work order processing level and the work order processing category information, the associated historical work orders can be retrieved according to the work order processing level and the work order processing category information in the work order number. For example, the work order with the same first 5 digits as the work order number of the pre-filled work order is obtained as the historical work order. Similarity calculation can be performed based on the content in the pre-filled work order and the historical work order. For example, the similarity between the pre-filled work order and the historical work order is determined by the cosine similarity algorithm. The historical work orders with similarity greater than the preset similarity threshold are regarded as similar work orders. The preset similarity threshold is an adjustable parameter. For example, the preset similarity threshold is 85%. When notifying the person in charge to process the work order, similar work orders can be provided simultaneously when providing the pre-filled work order.

[0087] By retrieving similar work orders corresponding to pre-filled work orders and providing them simultaneously to the person in charge for work order processing, a reference solution can be provided to the person in charge, thereby reducing the processing capability requirements for the person in charge and improving the timeliness, reliability and satisfaction of business processing.

[0088] Based on the above implementation method, optionally, after notifying the corresponding person in charge to process the work order, it also includes: retrieving the target work order corresponding to the same user based on the user basic data, and generating transaction processing tracking results based on the user emotional state, user call type, and call transaction type in each target work order; and based on the transaction processing tracking results, providing transaction processing service improvement suggestions.

[0089] Generating transaction tracking results can help managers centrally display and track recent customer service quality, summarize experiences, learn lessons, and do a good job in service management and control of similar incidents and problems, thereby further improving service levels.

[0090] The technical solution of the embodiment of the present invention obtains the voice packet data of the user's incoming call, extracts the frequency domain features, time domain features, and specific vocabulary features of the voice in the voice packet data, and uses the frequency domain features, time domain features, and specific vocabulary features as acoustic features; inputs the acoustic features into a pre-trained recurrent neural network model to obtain the user's emotional state; converts the voice packet data into text content through a speech-to-text algorithm; performs named entity recognition on the text content through a trained named entity recognition model based on a dual affine classifier to obtain the user's incoming call type, incoming call transaction type, and user basic data; generates a pre-filled Work order; perform integrity check on the pre-filled work order, and when the check passes, match the person in charge according to the pre-filled work order; when the match is successful, use the first level notification method to notify the corresponding person in charge to process the work order; when the match fails, use the second level notification method to notify the corresponding person in charge to process the work order; when the check fails, use the second level notification method to notify the corresponding person in charge to process the work order, which solves the problem of large human factors and low efficiency in the work order processing of the existing technology. By automatically generating pre-filled work orders, the skill requirements for customer service can be reduced, the business processing efficiency can be improved, and user satisfaction can be guaranteed; by using different levels of notification methods in different situations, the timeliness of work order processing can be guaranteed to avoid missing work orders.

[0091] Embodiment 3

[0092] Figure 4 Schematic diagram of a work order generation device based on natural language processing according to the third embodiment of the present invention. Figure 4 As shown, the device includes: a user emotional state determination module 410, a text conversion module 420, a named entity recognition module 430, a pre-filled work order generation module 440 and a work order notification module 450. Among them:

[0093] The user emotional state determination module 410 is used to obtain the voice packet data of the user's incoming call, extract the acoustic features in the voice packet data, and determine the user's emotional state according to the acoustic features;

[0094] The text conversion module 420 is used to convert the voice packet data into text content through a voice-to-text algorithm;

[0095] Named entity recognition module 430, used to perform named entity recognition on the text content to obtain the user's incoming call type, incoming call transaction type and user basic data;

[0096] A pre-filled work order generation module 440 is used to generate a pre-filled work order according to the user's emotional state, the user's incoming call type, the incoming call transaction type and the user's basic data;

[0097] The work order notification module 450 is used to notify the corresponding person in charge to process the work order according to the pre-filled work order.

[0098] Optionally, the user emotional state determination module 410 includes:

[0099] An acoustic feature extraction unit, used to extract frequency domain features, time domain features, and specific vocabulary features of speech from the speech packet data, and use the frequency domain features, time domain features, and specific vocabulary features as acoustic features;

[0100] The frequency domain features include at least one of the following: Mel frequency cepstral coefficient and fundamental frequency; the time domain features include at least one of the following: speech rate and volume; the specific vocabulary features include the frequency of use of specific vocabulary;

[0101] The user emotional state determination unit is used to input the acoustic features into a pre-trained recurrent neural network model to obtain the user emotional state.

[0102] Optionally, the named entity recognition module 430 includes:

[0103] A classification and annotation unit is used to classify and annotate content entities in the customer service corpus through a remote supervision algorithm to obtain preset user call types, call transaction types, and description information of each type;

[0104] A named entity recognition model training unit, used for training a named entity recognition model based on a dual affine classifier according to preset user incoming call types, incoming call transaction types, and description information of each type;

[0105] The named entity recognition unit is used to perform named entity recognition on the text content through a trained named entity recognition model based on a dual affine classifier to obtain the user's incoming call type, incoming call transaction type and user basic data.

[0106] Optionally, the pre-filled work order generation module 440 includes:

[0107] A work order processing level generating unit, used to generate a work order processing level according to the user's emotional state, the user's incoming call type, the incoming call transaction type, the user's basic data, and the preset weights corresponding to each content;

[0108] A work order processing category generating unit, used to generate a work order processing category according to a user's incoming call type and an incoming call transaction type;

[0109] The work order number generating unit is used to generate a work order number according to the work order processing level and the work order processing category.

[0110] Optionally, the work order notification module 450 includes:

[0111] A work order similarity determination unit, used to retrieve the associated historical work orders according to the work order number, and determine the similarity between the pre-filled work order and the historical work order;

[0112] The work order notification unit is used to filter similar work orders from historical work orders based on similarity; and to notify the corresponding person in charge to process the work order based on the pre-filled work orders and similar work orders.

[0113] Optionally, the work order notification module 450 includes:

[0114] The responsible person matching unit is used to perform integrity verification on the pre-filled work order. If the verification passes, the responsible person is matched according to the pre-filled work order.

[0115] The first notification unit is used to notify the corresponding person in charge to process the work order in a first-level notification manner when the match is successful;

[0116] The second notification unit is used to notify the corresponding person in charge to process the work order by using the second level notification method when the matching fails;

[0117] The third notification unit is used to notify the corresponding person in charge to process the work order by using the second level notification method when the verification fails;

[0118] Among them, the first level notification method has a lower priority than the second level notification method.

[0119] The work order generation device based on natural language processing provided in an embodiment of the present invention can execute the work order generation method based on natural language processing provided in any embodiment of the present invention, and has the corresponding functional modules and beneficial effects of the execution method.

[0120] In the technical solution of the embodiment of the present invention, the user personal information involved (such as voice package data, user basic data, etc.) is information and data authorized by the user or fully authorized by all parties, and the collection, storage, use, processing, transmission, provision, disclosure and application of the relevant data comply with the relevant laws, regulations and standards of the relevant countries and regions, take necessary confidentiality measures, do not violate public order and good customs, and provide corresponding operation entrances for users to choose to authorize or refuse.

[0121] Embodiment 4

[0122] Figure 5A schematic diagram of the structure of an electronic device 10 that can be used to implement an embodiment of the present invention is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices (such as helmets, glasses, watches, etc.) and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present invention described and / or required herein.

[0123] like Figure 5 As shown, the electronic device 10 includes at least one processor 11, and a memory connected to the at least one processor 11, such as a read-only memory (ROM) 12, a random access memory (RAM) 13, etc., wherein the memory stores a computer program that can be executed by at least one processor, and the processor 11 can perform various appropriate actions and processes according to the computer program stored in the read-only memory (ROM) 12 or the computer program loaded from the storage unit 18 to the random access memory (RAM) 13. In the RAM 13, various programs and data required for the operation of the electronic device 10 can also be stored. The processor 11, the ROM 12, and the RAM 13 are connected to each other through a bus 14. An input / output (I / O) interface 15 is also connected to the bus 14.

[0124] A number of components in the electronic device 10 are connected to the I / O interface 15, including: an input unit 16, such as a keyboard, a mouse, etc.; an output unit 17, such as various types of displays, speakers, etc.; a storage unit 18, such as a disk, an optical disk, etc.; and a communication unit 19, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 19 allows the electronic device 10 to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks.

[0125] The processor 11 may be a variety of general and / or special processing components with processing and computing capabilities. Some examples of the processor 11 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, a digital signal processor (DSP), and any appropriate processor, controller, microcontroller, etc. The processor 11 executes the various methods and processes described above, such as a work order generation method based on natural language processing.

[0126] In some embodiments, the work order generation method based on natural language processing may be implemented as a computer program, which is tangibly contained in a computer-readable storage medium, such as a storage unit 18. In some embodiments, part or all of the computer program may be loaded and / or installed on the electronic device 10 via the ROM 12 and / or the communication unit 19. When the computer program is loaded into the RAM 13 and executed by the processor 11, one or more steps of the work order generation method based on natural language processing described above may be performed. Alternatively, in other embodiments, the processor 11 may be configured to execute the work order generation method based on natural language processing in any other appropriate manner (e.g., by means of firmware).

[0127] Various implementations of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), load programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various implementations can include: being implemented in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0128] Computer programs for implementing the methods of the present invention may be written in any combination of one or more programming languages. These computer programs may be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, so that when the computer program is executed by the processor, the functions / operations specified in the flow chart and / or block diagram are implemented. The computer program may be executed entirely on the machine, partially on the machine, partially on the machine and partially on a remote machine as a stand-alone software package, or entirely on a remote machine or server.

[0129] In the context of the present invention, a computer-readable storage medium may be a tangible medium that may contain or store a computer program for use by or in combination with an instruction execution system, device or equipment. A computer-readable storage medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices or equipment, or any suitable combination of the foregoing. Alternatively, a computer-readable storage medium may be a machine-readable signal medium. A more specific example of a machine-readable storage medium may include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0130] To provide interaction with a user, the systems and techniques described herein may be implemented on an electronic device having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and a pointing device (e.g., a mouse or trackball) through which the user can provide input to the electronic device. Other types of devices may also be used to provide interaction with the user; for example, the feedback provided to the user may be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user may be received in any form (including acoustic input, voice input, or tactile input).

[0131] The systems and techniques described herein may be implemented in a computing system that includes backend components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes frontend components (e.g., a user computer with a graphical user interface or a web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such backend components, middleware components, or frontend components. The components of the system may be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: a local area network (LAN), a wide area network (WAN), a blockchain network, and the Internet.

[0132] A computing system may include a client and a server. The client and the server are generally remote from each other and usually interact through a communication network. The client and server relationship is generated by computer programs running on the corresponding computers and having a client-server relationship with each other. The server may be a cloud server, also known as a cloud computing server or cloud host, which is a host product in the cloud computing service system to solve the defects of difficult management and weak business scalability in traditional physical hosts and VPS services.

[0133] It should be understood that the various forms of processes shown above can be used to reorder, add or delete steps. For example, the steps described in the present invention can be executed in parallel, sequentially or in different orders, as long as the desired results of the technical solution of the present invention can be achieved, and this document does not limit this.

[0134] The above specific implementations do not constitute a limitation on the protection scope of the present invention. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modification, equivalent substitution and improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A work order generation method based on natural language processing, characterized in that: include: Acquire voice packet data of an incoming call from a user, extract acoustic features from the voice packet data, and determine the user's emotional state based on the acoustic features; Converting the voice packet data into text content through a voice-to-text algorithm; Performing named entity recognition on the text content to obtain the user's incoming call type, incoming call transaction type and user basic data; Generate a pre-filled work order based on the user's emotional state, the user's incoming call type, the incoming call transaction type, and the user's basic data; Notify the corresponding person in charge to process the work order according to the pre-filled work order.

2. The method according to claim 1, characterized in that Extracting acoustic features from the voice packet data and determining the user's emotional state according to the acoustic features, including: Extracting frequency domain features, time domain features, and specific vocabulary features of speech from the speech packet data, and using the frequency domain features, time domain features, and specific vocabulary features as acoustic features; The frequency domain features include at least one of the following: Mel frequency cepstral coefficient and fundamental frequency; the time domain features include at least one of the following: speech rate and volume; the specific vocabulary features include the frequency of use of specific vocabulary; The acoustic features are input into a pre-trained recurrent neural network model to obtain the user's emotional state.

3. The method according to claim 1, characterized in that Perform named entity recognition on the text content to obtain the user's incoming call type, incoming call transaction type and user basic data, including: The content entities in the customer service corpus are classified and labeled through a remote supervision algorithm to obtain the preset user call type, call transaction type, and description information of each type; The named entity recognition model based on the dual affine classifier is trained according to the preset user call type, call transaction type, and description information of each type; By using a trained named entity recognition model based on a dual affine classifier, named entity recognition is performed on the text content to obtain the user's incoming call type, incoming call transaction type and user basic data.

4. The method according to claim 1, characterized in that: Generate a pre-filled work order according to the user's emotional state, the user's incoming call type, the incoming call transaction type, and the user's basic data, including: Generate a work order processing level according to the user's emotional state, the user's incoming call type, the incoming call transaction type, the user's basic data, and the preset weights corresponding to each content; Generate work order processing categories based on user call types and call transaction types; Generate a work order number according to the work order processing level and the work order processing category.

5. The method according to claim 4, characterized in that According to the pre-filled work order, the corresponding person in charge is notified to process the work order, including: Retrieving the associated historical work order according to the work order number, and determining the similarity between the pre-filled work order and the historical work order; Based on the similarity, similar work orders are screened from the historical work orders; and the corresponding person in charge is notified to process the work order based on the pre-filled work order and the similar work order.

6. The method according to claim 1, characterized in that According to the pre-filled work order, the corresponding person in charge is notified to process the work order, including: Performing integrity check on the pre-filled work order, and matching the responsible person according to the pre-filled work order when the check passes; When the match is successful, the first level notification method is used to notify the corresponding person in charge to process the work order; When the match fails, the second level notification method is used to notify the corresponding person in charge to handle the work order; If the verification fails, the second level notification method is used to notify the corresponding person in charge to process the work order; The priority of the first level notification method is lower than that of the second level notification method.

7. A work order generation device based on natural language processing, characterized in that: include: A user emotional state determination module, used to obtain voice packet data of a user's incoming call, extract acoustic features from the voice packet data, and determine the user's emotional state based on the acoustic features; A text conversion module, used to convert the voice packet data into text content through a voice-to-text algorithm; A named entity recognition module is used to perform named entity recognition on the text content to obtain the user's incoming call type, incoming call transaction type and user basic data; A pre-filled work order generation module, used to generate a pre-filled work order according to the user's emotional state, the user's incoming call type, the incoming call transaction type and the user's basic data; The work order notification module is used to process the work order according to the pre-filled work order notification corresponding to the person in charge.

8. An electronic device, characterized in that: The electronic device comprises: at least one processor; and a memory communicatively connected to the at least one processor; wherein, The memory stores a computer program executable by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor can execute the work order generation method based on natural language processing according to any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a processor to implement the work order generation method based on natural language processing according to any one of claims 1 to 6 when executed.

10. A computer program product, comprising a computer program, wherein when the computer program is executed by a processor, the method for generating a work order based on natural language processing according to any one of claims 1 to 6 is implemented.

Images