Information verification method and device based on machine learning and readable storage medium

Abstract

The invention provides an information verification method based on machine learning, and the method comprises the steps: obtaining customer service communication audio which is communication audio between an artificial customer service and a potential customer; converting the call recording into a text by using an automatic voice recognition technology; performing error correction processing on the text by adopting a text error correction model based on deep learning to obtain a to-be-processed text; identifying the to-be-processed text by using a semantic identification model to obtain a customer service expression group; calculating the similarity between the customer service expression group and the standard expression group; and if the similarity is smaller than a preset value, determining that the customer service communication audio does not meet the verification requirement. According to the scheme, the steps of text error correction, semantic recognition, similarity calculation and the like are all automatically executed by the model, the number of audios capable of being verified every day is greatly increased, the semantic recognition model can more accurately understand customer service expression intentions, fuzziness and uncertainty of manual judgment are reduced, and the credibility of verification results is improved.

Description

Technical Field

[0001] The present application relates to the field of data processing, and in particular to an information verification method, device, and readable storage medium based on machine learning. Background Art

[0002] In the service industry, conversation quality control is becoming increasingly important as market competition intensifies and businesses expand. Effectively monitoring the quality of communication between customer service representatives and their customers is crucial, especially for businesses that handle a high volume of phone calls.

[0003] Currently, traditional quality inspection methods rely primarily on offline manual processes. To standardize the communication process between customer service and property owners, companies often develop standardized dialogue and process specifications. For example, to collect property information (such as area, type, and whether the property has been delivered) and to promote the company's product value proposition, standardized expression groups are summarized into a standard process. However, in actual implementation, offline manual quality inspections have exposed numerous problems.

[0004] From the perspective of quality inspection efficiency, manual quality inspection relies on listening to recordings and reading transcripts, which is extremely inefficient. A successfully collected recording typically lasts 5-7 minutes. A quality inspector working an 8-hour day (480 minutes) can only inspect approximately 68 recordings at most. If the quality inspection department only has four people, they can only inspect a maximum of 272 recordings per day. This low quality inspection coverage can lead some customer service staff to take chances and not adhere to communication standards, further reducing the stability of service quality. Summary of the Invention

[0005] The present invention aims to provide a machine learning-based information verification method, which aims to solve the problems of low efficiency of existing information verification, possible subjective judgment errors, and customer service personnel's failure to comply with communication standards. The machine learning-based information verification method provided in this application includes:

[0006] Acquire communication audio, wherein the communication audio is the audio of the communication between the human customer service and the potential customer;

[0007] Converting the communication audio into text using automatic speech recognition technology;

[0008] Performing error correction processing on the text using a text error correction model based on deep learning to obtain a text to be processed;

[0009] Using a semantic recognition model to recognize the text to be processed, to obtain a customer service expression group;

[0010] Calculating the similarity between the customer service expression group and the standard expression group;

[0011] If the similarity is less than a preset value, it is determined that the customer service communication audio does not meet the verification requirements;

[0012] If the similarity is greater than or equal to the preset value, it is determined that the customer service communication audio meets the verification requirements.

[0013] Based on the information verification method based on machine learning described in the first aspect of the embodiment of the present application, optionally,

[0014] The text error correction model based on deep learning includes: a pinyin similarity rule algorithm.

[0015] Based on the information verification method based on machine learning described in the first aspect of the embodiment of the present application, optionally,

[0016] The text error correction model based on deep learning includes: a keyword replacement algorithm.

[0017] Based on the information verification method based on machine learning described in the first aspect of the embodiment of the present application, optionally,

[0018] The text error correction model based on deep learning includes: a text similarity replacement algorithm.

[0019] Based on the information verification method based on machine learning described in the first aspect of the embodiment of the present application, optionally,

[0020] The semantic recognition model is used to identify the text to be processed to obtain a customer service expression group, including:

[0021] Using a semantic recognition model to recognize the text to be processed, to obtain a plurality of customer service expression groups, each of the customer service expression groups corresponding to a specific semantics;

[0022] Calculating the similarity between the customer service expression group and the standard expression group includes:

[0023] Calculate the similarity between each customer service expression group and its corresponding semantic standard expression group.

[0024] Based on the information verification method based on machine learning described in the first aspect of the embodiment of the present application, optionally,

[0025] The multiple customer service expression groups include any one or more of a start expression group, an information confirmation expression group, a room measurement expression group and an end expression group.

[0026] Based on the information verification method based on machine learning described in the first aspect of the embodiment of the present application, optionally,

[0027] Calculating the similarity between the customer service expression group and the standard expression group includes:

[0028] The similarity between the customer service expression group and the standard expression group is calculated using a text comparison algorithm based on cosine similarity.

[0029] A second aspect of the embodiments of the present application provides an information verification device based on machine learning, comprising:

[0030] An acquisition unit is used to acquire customer service communication audio, wherein the communication audio is the audio of the communication between the human customer service and the potential customer;

[0031] a conversion unit, configured to convert the call recording into text using automatic speech recognition technology;

[0032] An error correction unit, configured to perform error correction processing on the text using a text error correction model based on deep learning to obtain a text to be processed;

[0033] A recognition unit, configured to recognize the text to be processed using a semantic recognition model to obtain a customer service expression group;

[0034] a calculation unit, configured to calculate the similarity between the customer service expression group and the standard expression group;

[0035] A non-compliance unit, configured to determine that the customer service communication audio does not meet the verification requirements if the similarity is less than a preset value;

[0036] The compliance unit is used to determine that the customer service communication audio meets the verification requirement if the similarity is greater than or equal to the preset value.

[0037] Based on the information verification device based on machine learning described in the second aspect of the embodiment of the present application, optionally,

[0038] The text error correction model based on deep learning includes: a pinyin similarity rule algorithm.

[0039] Based on the information verification device based on machine learning described in the second aspect of the embodiment of the present application, optionally,

[0040] The text error correction model based on deep learning includes: a keyword replacement algorithm.

[0041] Based on the information verification device based on machine learning described in the second aspect of the embodiment of the present application, optionally,

[0042] The text error correction model based on deep learning includes: a text similarity replacement algorithm.

[0043] Based on the information verification device based on machine learning described in the second aspect of the embodiment of the present application, optionally,

[0044] The recognition unit is specifically configured to: use a semantic recognition model to recognize the text to be processed to obtain a plurality of customer service expression groups, each of which corresponds to a specific semantics;

[0045] The computing unit is specifically configured to:

[0046] Calculate the similarity between each customer service expression group and its corresponding semantic standard expression group.

[0047] Based on the information verification device based on machine learning described in the second aspect of the embodiment of the present application, optionally,

[0048] The multiple customer service expression groups include any one or more of a start expression group, an information confirmation expression group, a room measurement expression group and an end expression group.

[0049] Based on the machine learning-based information verification device described in the second aspect of the embodiment of the present application, optionally, the computing unit is specifically configured to:

[0050] Calculating the similarity between the customer service expression group and the standard expression group includes:

[0051] The similarity between the customer service expression group and the standard expression group is calculated using a text comparison algorithm based on cosine similarity.

[0052] A third aspect of the embodiments of the present application provides an information verification device based on machine learning, comprising:

[0053] CPU, memory, input and output interfaces, wired or wireless network interfaces, and power supply;

[0054] The memory is a transient storage memory or a persistent storage memory;

[0055] The central processing unit is configured to communicate with the memory and execute instruction operations in the memory on the device to perform the method as described in any one of the first aspects of the embodiments of the present application.

[0056] A fourth aspect of the embodiments of the present application provides a computer-readable storage medium, comprising instructions, which, when executed on a computer, enable the computer to execute the method described in any one of the first aspects of the embodiments of the present application.

[0057] A fifth aspect of the embodiments of the present application provides a computer program product comprising instructions, which, when executed on a computer, enables the computer to execute any one of the methods described in the first aspect of the embodiments of the present application.

[0058] As can be seen from the above technical solutions, the embodiments of the present application have the following advantages: a machine learning-based information verification method, characterized by comprising: obtaining customer service communication audio, the communication audio being the audio of the communication between a human customer service representative and a potential customer; converting the call recording into text using automatic speech recognition technology; correcting the text using a deep learning-based text correction model to obtain a text to be processed; recognizing the text to be processed using a semantic recognition model to obtain a customer service expression group; calculating the similarity between the customer service expression group and a standard expression group; if the similarity is less than a preset value, determining that the customer service communication audio does not meet the verification requirements; if the similarity is greater than or equal to the preset value, determining that the customer service communication audio meets the verification requirements. This solution, with the help of automatic speech recognition technology, can quickly convert customer service communication audio into text, saving a lot of manual transcription time. In addition, the subsequent steps of text correction, semantic recognition, and similarity calculation are all automatically performed by the model, greatly increasing the amount of audio that can be verified daily, solving the problem of low manual verification coverage, and enabling enterprises to conduct timely verification of large amounts of customer service communication audio. The semantic recognition model can more accurately understand the customer service's intentions and determine in a quantitative manner whether the audio meets the verification requirements, reducing the ambiguity and uncertainty of human judgment and improving the credibility of the verification results. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] To more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are merely embodiments of the present application. A person of ordinary skill in the art can also derive other drawings based on the provided drawings without inventive effort. It should be understood that the drawings provided in this section are only used to better understand the present solution and do not constitute a limitation of the present application.

[0060] Figure 1 A flow chart of an embodiment of the information verification method based on machine learning provided in this application.

[0061] Figure 2 This is another flowchart of an embodiment of the information verification method based on machine learning provided in this application.

[0062] Figure 3 A structural diagram of an embodiment of the information verification device based on machine learning provided in this application.

[0063] Figure 4 This is another structural diagram of an embodiment of the information verification device based on machine learning provided in this application. DETAILED DESCRIPTION

[0064] In order to enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application are clearly and completely described below. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative work should fall within the scope of protection of this application. At the same time, for the sake of clarity and conciseness, the description of well-known functions and structures is omitted in the following description.

[0065] The terms "first," "second," "third," "fourth," and the like (if any) in the specification and claims of this application and in the accompanying drawings are used to distinguish similar objects and are not necessarily used to describe a particular order or precedence. It should be understood that the terms used in this manner are interchangeable where appropriate so that the embodiments described herein can be implemented in an order other than that illustrated or described herein. In addition, the terms "including" and "having," and any variations thereof, are intended to cover non-exclusive inclusions, e.g., a process, method, product, or apparatus comprising a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such process, method, product, or apparatus.

[0066] In the service industry, conversation quality control is becoming increasingly important as market competition intensifies and businesses expand. Effectively monitoring the quality of communication between customer service representatives and their customers is crucial, especially for businesses that handle a high volume of phone calls.

[0067] Currently, traditional quality inspection methods rely primarily on offline manual processes. To standardize the communication process between customer service and property owners, companies often develop standardized dialogue and process specifications. For example, to collect property information (such as area, type, and whether the property has been delivered) and to promote the company's product value proposition, standardized expression groups are summarized into a standard process. However, in actual implementation, offline manual quality inspections have exposed numerous problems.

[0068] From the perspective of quality inspection efficiency, manual quality inspection relies on listening to recordings and reading transcripts, which is extremely inefficient. A successfully collected recording typically lasts 5-7 minutes. A quality inspector working an 8-hour day (480 minutes) can only inspect approximately 68 recordings at most. If the quality inspection department only has four people, they can only inspect a maximum of 272 recordings per day. This low quality inspection coverage can lead some customer service staff to take chances and not adhere to communication standards, further reducing the stability of service quality.

[0069] To solve the above problems, the present invention provides an information verification method based on machine learning. Figure 1, an embodiment of the information verification method based on machine learning provided in this application includes: steps 101 to 105.

[0070] 101. Obtain customer service communication audio.

[0071] Acquire customer service communication audio, which is the audio of exchanges between human customer service and potential customers.

[0072] In a company's customer service system, there is usually a dedicated recording module to record the communication between human customer service representatives and potential customers in real time. This audio data will be stored in the company's server or storage device, and these audio files can be obtained from the storage location through the corresponding interface. The company's customer service system usually provides an API interface, and developers can write code to call these interfaces to obtain the specified customer service communication audio according to certain rules (such as time range, customer service ID, etc.). The source of the audio can be local storage, cloud storage, or real-time streaming audio data. The audio format can be common MP3, WAV, etc., and the system should have the ability to process audio in different formats.

[0073] 102. Use automatic speech recognition technology to convert call recordings into text.

[0074] Specifically, automatic speech recognition (ASR) technology is based on deep learning models, such as end-to-end convolutional neural networks (CNNs), recurrent neural networks (RNNs), and their variants (such as LSTMs and GRUs). These models are trained on large amounts of speech data, learning the mapping between speech signals and text, thereby converting input speech signals into corresponding text.

[0075] During the speech-to-text conversion process, the system can pre-define tags for customer service and customer voices, such as using specific symbols (such as "Customer Service:" and "Customer:") in the text to distinguish between the two speakers. Before audio-to-text conversion, audio processing technology is used to extract the timbre characteristics of the speech. Different people have unique timbre, so a timbre classification model can be trained to classify the speech in the audio into two categories: customer service and customer voice. For example, a Gaussian mixture model (GMM) can be used to model timbre characteristics and train the model to distinguish between customer service and customer voices. In actual applications, audio is input into the trained model, and the audio clips are classified based on the output. Subsequent speech-to-text conversion and semantic recognition are then performed only on the customer service audio clips. It is understandable that different ASR technologies may vary in recognition accuracy, processing speed, and applicable scenarios. In actual applications, appropriate technologies and services can be selected based on needs and actual circumstances. Furthermore, for speech used in certain domains, targeted model training may be required to improve recognition accuracy.

[0076] 103. Use a text error correction model based on deep learning to perform error correction processing on the text to obtain a text to be processed.

[0077] Specifically, the model learns the rules and patterns of text correction by learning from a large number of correct and incorrect text pairs. When given an incorrect text input, the model outputs the corrected text. Understandably, the performance of text correction models is affected by factors such as the quality and quantity of training data and the choice of model architecture. In practical applications, continuous model optimization and updated training data are necessary to improve correction accuracy. Furthermore, special vocabulary and technical terminology may require additional processing.

[0078] 104. Use a semantic recognition model to recognize the text to be processed to obtain a customer service expression group.

[0079] Specifically, during the text conversion step, if the system can pre-define tags for customer service and customer utterances, such as using specific symbols (e.g., "Customer Service:," "Customer:") to distinguish between the two sides of the text, then semantic recognition only requires extracting the text content tagged with "Customer Service:." Regular expressions can be used to match text lines with specific tags, and these lines can then be grouped as customer service utterances for subsequent processing.

[0080] 105. Calculate the similarity between the customer service expression group and the standard expression group.

[0081] Convert the customer service expression group and the standard expression group into vector representations (such as word vectors, sentence vectors, etc.), and then calculate the cosine similarity between the two vectors. Different similarity calculation methods are suitable for different scenarios and data types. In actual applications, it is necessary to select an appropriate method based on the specific situation and make reasonable adjustments and processing to the calculation results. The standard expression group is a pre-set standard expression method that covers polite language and accurate information inquiry that should be used during customer service. The specific content of the standard expression group can be adjusted and set according to the actual situation.

[0082] 106. It is determined that the customer service communication audio does not meet the verification requirements.

[0083] Specifically, the calculated similarity value is compared with the preset threshold. If the similarity is less than the threshold, the customer service communication audio is deemed to not meet the verification requirements. The preset threshold can be set according to the business needs and actual situation of the enterprise. The setting of the preset threshold will affect the strictness of the verification. If the threshold is set too high, a large number of audios that meet the requirements may be mistakenly judged as not meeting the requirements; if the threshold is set too low, some audios that do not meet the requirements may pass the verification. Therefore, adjustments and optimizations are required based on actual conditions. In the actual setting process, strong verification associations can also be set for certain words or words. For example, if a specific sentence does not appear, the customer service communication audio is directly deemed to not meet the verification requirements. The specific rules can be determined according to the actual situation and are not limited here.

[0084] 107. Determine that the customer service communication audio meets the verification requirements.

[0085] Specifically, the calculated similarity value is compared with a preset threshold. If the similarity is greater than or equal to the threshold, the customer service communication audio is deemed to meet the verification requirements, and the customer service communication process corresponding to the audio complies with the specifications. For audio that meets the verification requirements, further analysis and processing can be performed as needed, such as displaying it as an excellent case, using it as material for training new customer service staff, or serving as a basis for model training used in the aforementioned steps, etc. The specific details can be determined based on actual conditions and are not limited here.

[0086] As can be seen from the above technical solutions, the embodiments of the present application have the following advantages: a machine learning-based information verification method, characterized by comprising: obtaining customer service communication audio, the communication audio being the audio of the communication between a human customer service representative and a potential customer; converting the call recording into text using automatic speech recognition technology; correcting the text using a deep learning-based text correction model to obtain a text to be processed; recognizing the text to be processed using a semantic recognition model to obtain a customer service expression group; calculating the similarity between the customer service expression group and a standard expression group; if the similarity is less than a preset value, determining that the customer service communication audio does not meet the verification requirements; if the similarity is greater than or equal to the preset value, determining that the customer service communication audio meets the verification requirements. This solution, with the help of automatic speech recognition technology, can quickly convert customer service communication audio into text, saving a lot of manual transcription time. In addition, the subsequent steps of text correction, semantic recognition, and similarity calculation are all automatically performed by the model, greatly increasing the amount of audio that can be verified daily, solving the problem of low manual verification coverage, and enabling enterprises to conduct timely verification of large amounts of customer service communication audio. The semantic recognition model can more accurately understand the customer service's intentions and determine in a quantitative manner whether the audio meets the verification requirements, reducing the ambiguity and uncertainty of human judgment and improving the credibility of the verification results.

[0087] The above content provides an overall description of this solution. Further, this solution can be further refined during the actual implementation process. For details, please refer to Figure 2 An embodiment of the information verification algorithm based on machine learning provided in this application includes steps 201 to 207.

[0088] 201. Obtain customer service communication audio.

[0089] 202. Use automatic speech recognition technology to convert call recordings into text.

[0090] This solution is applied to the decoration platform system as an example, where the customer service staff is the decoration platform system customer service, and the potential customers are customers who have logged into the decoration platform system or expressed their willingness to decorate. The above steps 201 to 202 are the same as the above steps. Figure 1 Steps 101 to 102 in the corresponding embodiment are similar, and details can be referred to the above content, which will not be repeated here.

[0091] 203. Use a deep learning-based text error correction model to perform error correction processing on the text to obtain a text to be processed.

[0092] Specifically, since the system is applied to a decoration platform system, the text error correction model based on deep learning needs to include a variety of characteristic settings, including:

[0093] (1) Pinyin similarity rule algorithm.

[0094] The algorithm's core principle is Chinese pinyin. In Chinese, some characters have identical or similar pronunciations, which can easily lead to errors during speech recognition. By calculating pinyin similarity, incorrect text with similar pronunciations is corrected. Because this solution is applied to a renovation platform, certain terms related to the renovation field will be adjusted. For example, "old house renovation" and "just put renovation" are very similar in pinyin and should be recognized as "old house renovation."

[0095] (2) Keyword replacement algorithm.

[0096] When customer service representatives communicate with customers, certain keywords or phrases are strictly regulated and must be accurately conveyed. However, these keywords may be incorrect due to errors in speech recognition or slips of the tongue. By analyzing large amounts of text, we identify common errors and then forcibly replace them. For example, the standard expression "one-stop decoration guarantee platform" is often recognized as "one-stop decoration high-rise platform." Using a keyword replacement algorithm, when "one-stop decoration high-rise platform" is detected in text, it is forcibly replaced with "one-stop decoration guarantee platform."

[0097] (3) Text similarity replacement algorithm.

[0098] In Chinese, some words have similar semantics although their spellings are slightly different. By calculating the semantic similarity between texts, incorrect or similar expressions are replaced with standard expressions. For example, for "毛坏房" (literally "raw house"), calculate its semantic similarity with "毛坯房" (literally "blank housing") in the standard vocabulary library. If the similarity exceeds the threshold, then replace "毛坏房" with "毛坯房".

[0099] 204. Use a semantic recognition model to recognize the text to be processed, and obtain multiple customer service expression groups.

[0100] Specifically, use a semantic recognition model to process the text to be processed and recognize multiple customer service expression groups. These customer service expression groups respectively correspond to specific semantics, such as a start expression group, an information confirmation expression group, a house measurement expression group, and an end expression group, etc., and may include any one or more of them. The key to this step is to accurately distinguish the speech segments with different semantics, especially to handle the situation of repeated expressions well, providing a basis for subsequent comparison with the standard expression groups.

[0101] The model first makes a preliminary analysis of the text according to preset general rules. For example, the start expression group usually begins with a greeting and a statement of identity and purpose, such as "Hello, I am [name of the customer service] from [company name]"; the end expression group generally includes a closing statement and a notification of subsequent arrangements, such as "Thank you for answering the call. The designer will contact you later". Through these obvious features, the model can initially identify the start and end speech segments.

[0102] It should be noted that for words like "half an hour" that may appear repeatedly in different speech segments, the model needs to combine context information for accurate discrimination. The specific approach is as follows:

[0103] Context semantic analysis: Analyze the text content before and after "half an hour". If information related to "house measurement" is mentioned before, such as "We measure the house", and then "half an hour" is mentioned later, then it is determined that this "half an hour" belongs to the house measurement speech segment; if the expression related to the closing statement is before, such as "After ending the call", and then "half an hour" and "the designer will contact" are mentioned later, then it is determined as the closing statement speech segment.

[0104] In addition to "half an hour", other relevant keywords can also be focused on. For example, keywords such as "house size" and "measurement" may frequently appear in the house measurement speech segment; words such as "goodbye" and "thank you" may appear in the closing statement speech segment. Through the association of these keywords, assist in judging the speech segment where "half an hour" is located.[[ID=第十九]] [[ID=第二十]]

[0105] Finally, after identification and segmentation, the text is divided into different expression groups, which are classified into corresponding customer service expression groups such as the start expression group, information confirmation expression group, room measurement expression group and end expression group. Each expression group contains speech content with specific semantics. It is understandable that in the actual implementation process, the category of expression groups can be adjusted according to actual conditions, such as "product introduction expression group", "activity expression group" and so on. Specifically, the semantic recognition model can be adjusted and trained accordingly, which is not limited here. In order to improve the accuracy of expression group division, different models can be used to process the same text, and the processing results can be compared and verified, which is not limited here.

[0106] 205. Calculate the similarity between each customer service expression group and its corresponding semantic standard expression group.

[0107] Specifically, each customer service expression group is compared with its corresponding semantic standard expression group, and the similarity between the two is calculated to obtain a similarity result for each group. Each semantic standard expression group is pre-set and its specific content is not limited. In actual implementation, the similarity results of each group can be directly output and compared with a subsequent preset value. If any group does not exceed the preset value, step 206 is executed to determine that the customer service communication audio does not meet the verification requirements. If all groups exceed the preset value, step 207 is executed to determine that the customer service communication audio meets the verification requirements. The comparison results can also be assigned different weights, and the overall result is compared with the preset value. Specifically, a higher weight is assigned to expression groups related to the core business process. In the renovation business, the room measurement expression group is directly related to the accuracy and cost accounting of the subsequent renovation plan. If the room measurement information is miscommunicated, it can lead to significant deviations in the renovation budget, construction delays, and other serious consequences. Therefore, the weight of the room measurement expression group can be set to 0.4. The initial expression group, while important, mainly establishes initial contact, but has a relatively smaller direct impact on the business, and thus a weight of 0.1 can be set. The information confirmation expression group involves collecting key information about the owner's home, which affects the targeted nature of subsequent services. Its weight can be set to 0.3. The closing expression group is used to inform the customer of subsequent arrangements and affects the customer's overall experience and expectations of the service. Its weight is set to 0.2.

[0108] During implementation, a text comparison algorithm based on cosine similarity can be used to calculate the similarity between each customer service expression group and its corresponding semantic standard expression group. Cosine similarity is a commonly used method for calculating text similarity. It measures similarity by calculating the cosine value of the angle between two vectors. During text processing, the text is first converted into a vector representation, and then the cosine similarity between these vectors is calculated. The closer the cosine value is to 1, the more similar the two texts are; the closer it is to 0, the lower the similarity.

[0109] 206. It is determined that the customer service communication audio does not meet the verification requirements.

[0110] If the similarity is less than the threshold, the customer service communication audio is deemed to have failed the verification requirements. This result can be reported to superiors or returned to the corresponding customer service representative for further learning and enhancement of service capabilities.

[0111] 207 Determines that the customer service communication audio meets the verification requirements

[0112] Specifically, the calculated similarity value is compared with a preset threshold. If the similarity is greater than the threshold, it is determined that the customer service communication audio meets the verification requirements.

[0113] As can be seen from the above technical solutions, the embodiments of the present application have the following advantages: a machine learning-based information verification method, characterized by comprising: obtaining customer service communication audio, the communication audio being the audio of the communication between a human customer service representative and a potential customer; converting the call recording into text using automatic speech recognition technology; correcting the text using a deep learning-based text correction model to obtain a text to be processed; recognizing the text to be processed using a semantic recognition model to obtain a customer service expression group; calculating the similarity between the customer service expression group and a standard expression group; if the similarity is less than a preset value, determining that the customer service communication audio does not meet the verification requirements; if the similarity is greater than or equal to the preset value, determining that the customer service communication audio meets the verification requirements. This solution, with the help of automatic speech recognition technology, can quickly convert customer service communication audio into text, saving a lot of manual transcription time. In addition, the subsequent steps of text correction, semantic recognition, and similarity calculation are all automatically performed by the model, greatly increasing the amount of audio that can be verified daily, solving the problem of low manual verification coverage, and enabling enterprises to conduct timely verification of large amounts of customer service communication audio. The semantic recognition model can more accurately understand the customer service's intentions and determine in a quantitative manner whether the audio meets the verification requirements, reducing the ambiguity and uncertainty of human judgment and improving the credibility of the verification results.

[0114] The above content describes the information verification method based on machine learning provided by this application. To support the implementation of the above embodiment, this application also provides an information verification device based on machine learning. Figure 3 , an embodiment of the information verification device based on machine learning provided by the present application includes:

[0115] An acquisition unit 301 is configured to acquire customer service communication audio, wherein the communication audio is audio of communication between a human customer service representative and a potential customer;

[0116] A conversion unit 302 is used to convert the call recording into text using automatic speech recognition technology;

[0117] An error correction unit 303 is configured to perform error correction processing on the text using a text error correction model based on deep learning to obtain a text to be processed;

[0118] The recognition unit 304 is configured to recognize the text to be processed using a semantic recognition model to obtain a customer service expression group;

[0119] A calculation unit 305 is used to calculate the similarity between the customer service expression group and the standard expression group;

[0120] The compliance unit 306 is configured to determine that the customer service communication audio meets the verification requirement if the similarity is greater than or equal to the preset value.

[0121] A non-compliant unit 307 is configured to determine that the customer service communication audio does not meet the verification requirement if the similarity is less than a preset value;

[0122] Optionally, the deep learning-based text error correction model includes: a pinyin similarity rule algorithm.

[0123] Optionally, the deep learning-based text error correction model includes: a keyword replacement algorithm.

[0124] Optionally, the deep learning-based text error correction model includes: a text similarity replacement algorithm.

[0125] Optionally, the recognition unit is specifically configured to: use a semantic recognition model to recognize the text to be processed to obtain a plurality of customer service expression groups, each of the customer service expression groups corresponding to a specific semantics;

[0126] The computing unit is specifically configured to:

[0127] Calculate the similarity between each customer service expression group and its corresponding semantic standard expression group.

[0128] Optionally, the multiple customer service expression groups include: any one or more of a start expression group, an information confirmation expression group, a room measurement expression group and an end expression group.

[0129] Optionally, the computing unit is specifically configured to:

[0130] Calculating the similarity between the customer service expression group and the standard expression group includes:

[0131] The similarity between the customer service expression group and the standard expression group is calculated using a text comparison algorithm based on cosine similarity.

[0132] In this embodiment, the process performed by each unit in the information verification device based on machine learning is the same as the above Figure 1 、 2The method processes described in the corresponding embodiments are similar and will not be repeated here.

[0133] Figure 4 It is a structural diagram of an information verification device based on machine learning provided in an embodiment of the present application. The information verification device 400 may include one or more central processing units (CPU) 401 and a memory 405, and the memory 405 stores one or more applications or data.

[0134] In this embodiment, the specific functional module division in the central processing unit 401 can be the same as the above Figure 3 The functional module division method of each unit described in is similar and will not be repeated here.

[0135] Memory 405 may be volatile or persistent storage. The program stored in memory 405 may include one or more modules, each of which may include a series of instructions for operating on the server. Furthermore, CPU 401 may be configured to communicate with memory 405 and execute the series of instructions in memory 405 on server 400.

[0136] The information verification device 400 may further include one or more power supplies 402 , one or more wired or wireless network interfaces 403 , and one or more input / output interfaces 404 .

[0137] The CPU 401 can execute the aforementioned Figure 1 、 2 The operations performed by the information verification method in the illustrated embodiment will not be described in detail here.

[0138] An embodiment of the present application also provides a computer storage medium for storing computer software instructions used for the above-mentioned information verification method, which includes a program designed for executing the information verification method.

[0139] The information verification method can be as described above Figure 1 、 2 The information verification method described in .

[0140] The present application also provides a computer program product, which includes computer software instructions that can be loaded by a processor to implement the above Figure 1 Figure 2 The process of the information verification method for any one of the items.

[0141] In the several embodiments provided in this application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are merely schematic. For example, the equivalent transformation of circuits and the division of units are only a kind of logical function division. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interfaces, devices or units, which can be electrical, mechanical or other forms.

[0142] The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple units. Some or all of the units may be selected according to actual needs to achieve the purpose of the present embodiment.

[0143] In addition, the functional units in the various embodiments of the present application may be integrated into a single processing unit, or each unit may exist physically separately, or two or more units may be integrated into a single unit. The aforementioned integrated units may be implemented in the form of hardware or software functional units.

[0144] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions or improvements made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for verifying information based on machine learning, characterized in that: include: Acquire communication audio, wherein the communication audio is the audio of the communication between the human customer service and the potential customer; Converting the communication audio into text using automatic speech recognition technology; Performing error correction processing on the text using a text error correction model based on deep learning to obtain a text to be processed; Using a semantic recognition model to recognize the text to be processed, to obtain a customer service expression group; Calculating the similarity between the customer service expression group and the standard expression group; If the similarity is less than a preset value, it is determined that the customer service communication audio does not meet the verification requirements; If the similarity is greater than or equal to the preset value, it is determined that the customer service communication audio meets the verification requirements.

2. The information verification method based on machine learning according to claim 1, characterized in that: The text error correction model based on deep learning includes: a pinyin similarity rule algorithm.

3. The information verification method based on machine learning according to claim 1, characterized in that: The text error correction model based on deep learning includes: a keyword replacement algorithm.

4. The information verification method based on machine learning according to claim 1, characterized in that: The text error correction model based on deep learning includes: a text similarity replacement algorithm.

5. The information verification method based on machine learning according to claim 1, characterized in that: The semantic recognition model is used to identify the text to be processed to obtain a customer service expression group, including: Using a semantic recognition model to recognize the text to be processed, to obtain a plurality of customer service expression groups, each of the customer service expression groups corresponding to a specific semantics; Calculating the similarity between the customer service expression group and the standard expression group includes: Calculate the similarity between each customer service expression group and its corresponding semantic standard expression group.

6. The information verification method based on machine learning according to claim 5, characterized in that: The multiple customer service expression groups include any one or more of a start expression group, an information confirmation expression group, a room measurement expression group and an end expression group.

7. The information verification method based on machine learning according to claim 5, characterized in that: Calculating the similarity between the customer service expression group and the standard expression group includes: The similarity between the customer service expression group and the standard expression group is calculated using a text comparison algorithm based on cosine similarity.

8. An information verification device based on machine learning, characterized in that: include: An acquisition unit, configured to acquire communication audio, wherein the communication audio is audio of communication between a human customer service representative and a potential customer; A conversion unit, configured to convert the communication audio into text using automatic speech recognition technology; An error correction unit, configured to perform error correction processing on the text using a text error correction model based on deep learning to obtain a text to be processed; A recognition unit, configured to recognize the text to be processed using a semantic recognition model to obtain a customer service expression group; a calculation unit, configured to calculate the similarity between the customer service expression group and the standard expression group; A non-compliance unit, configured to determine that the customer service communication audio does not meet the verification requirements if the similarity is less than a preset value; The compliance unit is used to determine that the customer service communication audio meets the verification requirement if the similarity is greater than or equal to the preset value.

9. An information verification device based on machine learning, characterized in that: include: CPU, memory, input and output interfaces, wired or wireless network interfaces, and power supply; The memory is a transient storage memory or a persistent storage memory; The central processing unit is configured to communicate with the memory and execute instructions in the memory on the device to perform the method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that The method comprises instructions, which, when executed on a computer, cause the computer to perform the method according to any one of claims 1 to 7.

Images