Call content detection method and device, computer equipment and storage medium

Abstract

The invention relates to the technical field of artificial intelligence, can be applied to service system platforms of medical health, financial science and technology and the like, and discloses a call content detection method and device, computer equipment and a storage medium. Call records of a target customer service and a target customer are obtained, and the call records are converted into text data; analyzing the text data, and extracting a reply text of the target customer service; on the basis of the reply text, target content with an abnormal reply in the reply text is retrieved through an abnormal knowledge base; using the target content to establish a cue word of an anomaly detection model, taking the cue word as input, and generating an anomaly detection result of the target customer service through the anomaly detection model; therefore, the problems of unstable service quality and high compliance risk caused by the lack of systematic and standardized call content anomaly detection means in the call communication process between customer service personnel and customers in enterprises in the prior art can be solved.

Description

Technical Field

[0001] The present invention relates to the technical field of artificial intelligence, and particularly relates to a call content detection method, apparatus, computer device, and computer-readable storage medium. Background Art

[0002] Currently, in the insurance industry, the renewal collection business is a key link to ensure the stable income of insurance companies and customer satisfaction. The renewal collection work involves the communication process between customer service personnel and customers by phone to remind customers to pay premiums on time and answer questions that customers may encounter during the renewal process. However, traditional renewal collection work often relies on the personal experience and ability of customer service personnel and lacks systematic and standardized abnormal detection means for call content, resulting in unstable service quality and high compliance risks.

[0003] Certainly, in the fields of healthcare and fintech, during the communication process between customer service personnel and customers (patients in the healthcare field), there are also problems of lack of systematic and standardized abnormal detection means for call content, resulting in unstable service quality and high compliance risks. In the healthcare field, customer service personnel communicate with patients by phone to provide services such as appointment registration, medical condition consultation, and medication guidance. However, due to the lack of standardized communication norms and abnormal detection means, customer service personnel may have problems such as information misleading and inappropriate medical advice during communication, which not only affects the medical experience of patients but may also lead to medical disputes and increase the legal risks of medical institutions. In the fintech field, customer service personnel communicate with customers by phone to provide services such as financial product consultation, transaction operation guidance, and risk reminder. However, due to the complexity and professionalism of financial services, as well as the lack of standardized communication norms and abnormal detection means, if customer service personnel are slightly careless during communication, they may violate relevant laws and regulations or company internal norms, resulting in customers misunderstanding financial product information and making wrong investment decisions, thereby triggering customer complaints and legal disputes and damaging the reputation and interests of financial institutions.

[0004] That is, in the prior art, enterprises often rely on the personal experience and ability of customer service personnel during the communication process between customer service personnel and customers, and lack systematic and standardized abnormal detection means for call content, resulting in unstable service quality and high compliance risks.

[0005] Therefore, the prior art still needs to be improved and developed. Summary of the Invention

[0006] In view of the deficiencies of the above-mentioned prior art, the purpose of the present invention is to provide a call content detection method, device, computer device and computer-readable storage medium, aiming to solve the problems in the prior art that in the process of communication between customer service personnel and customers in enterprises, it often relies on the personal experience and ability of customer service personnel, lacks systematic and standardized abnormal detection means for call content, resulting in unstable service quality and high compliance risks.

[0007] To achieve the above purpose, the present invention adopts the following technical solutions:

[0008] In the first aspect, the present invention provides a call content detection method, which includes:

[0009] Obtain the call recording of the target customer service and the target customer, and convert the call recording into text data;

[0010] Parse the text data to extract the reply text of the target customer service;

[0011] Based on the reply text, retrieve the target content with abnormal replies in the reply text through the abnormal knowledge base;

[0012] Use the target content to form a prompt word of the abnormal detection model, and use the prompt word as an input to generate an abnormal detection result of the target customer service through the abnormal detection model.

[0013] In the second aspect, the present invention provides a call content detection device, which includes:

[0014] An acquisition module, configured to obtain the call recording of the target customer service and the target customer, and convert the call recording into text data;

[0015] A parsing module, configured to parse the text data to extract the reply text of the target customer service;

[0016] A retrieval module, configured to retrieve the target content with abnormal replies in the reply text through the abnormal knowledge base based on the reply text;

[0017] A composition module, configured to use the target content to form a prompt word of the abnormal detection model, and use the prompt word as an input to generate an abnormal detection result of the target customer service through the abnormal detection model.

[0018] In the third aspect, the present invention provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the call content detection method described above is implemented.

[0019] Fourthly, the present invention provides a computer-readable storage medium storing a computer program, wherein the computer program, when executed by a processor, implements the call content detection method as described above.

[0020] Compared with the prior art, the present invention provides a call content detection method, device, computer device and computer-readable storage medium. Among them, by obtaining the call recording of the target customer service and the target customer, and converting the call recording into text data; parsing the text data to extract the reply text of the target customer service; based on the reply text, retrieving the target content with abnormal reply in the reply text through an abnormal knowledge base; using the target content to form a prompt word of an abnormal detection model, and taking the prompt word as an input, generating the abnormal detection result of the target customer service through the abnormal detection model. Thus, the present invention can solve the problems in the prior art that in the process of the call communication between the customer service personnel and the customers in an enterprise, it often relies on the personal experience and ability of the customer service personnel, lacks systematic and standardized abnormal detection means for call content, resulting in unstable service quality and high compliance risks. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the drawings in the following description are only some embodiments recorded in the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0022] Figure 1 It is a schematic diagram of the application environment of a call content detection method provided by an embodiment of the present invention.

[0023] Figure 2 It is a schematic flowchart of a call content detection method provided by an embodiment of the present invention.

[0024] Figure 3 It is a schematic diagram of the program modules of a call content detection device provided by an embodiment of the present invention.

[0025] Figure 4 It is a schematic diagram of the structure of a computer device provided by an embodiment of the present invention.

[0026] Figure 5 It is another schematic diagram of the structure of a computer device provided by an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0027] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0028] It should be understood that when used in the specification of the present invention and the appended claims, the term "comprising" indicates the presence of the described features, wholes, steps, operations, elements, and / or components, but does not exclude the presence or addition of one or more other features, wholes, steps, operations, elements, components, and / or their combinations.

[0029] It should also be understood that the term "and / or" as used in the specification of the present invention and the appended claims refers to any combination and all possible combinations of one or more of the associated listed items, and includes these combinations.

[0030] As used in the specification of the present invention and the appended claims, the term "if" can be interpreted as "when", "once", "in response to determining", or "in response to detecting" depending on the context. Similarly, the phrase "if determined" or "if detecting [the described condition or event]" can be interpreted as meaning "once determined", "in response to determining", "once detecting [the described condition or event]", or "in response to detecting [the described condition or event]" depending on the context.

[0031] In addition, in the description of the specification of the present invention and the appended claims, the terms "first", "second", "third", etc. are only used for distinguishing descriptions and cannot be understood as indicating or implying relative importance.

[0032] The reference to "one embodiment" or "some embodiments" or the like described in the specification of the present invention means that a specific feature, structure, or characteristic described in conjunction with the embodiment is included in one or more embodiments of the present invention. Thus, the statements "in one embodiment", "in some embodiments", "in other some embodiments", "in still other embodiments", etc. that appear in different places in this specification do not necessarily refer to the same embodiment, but mean "one or more but not all embodiments", unless otherwise specifically emphasized in other ways. The terms "comprising", "including", "having" and their variants all mean "including but not limited to", unless otherwise specifically emphasized in other ways.

[0033] It should be understood that the magnitudes of the sequence numbers of the steps in the following embodiments do not mean the order of execution is prior or posterior. The order of execution of each process should be determined according to its function and internal logic, and should not constitute any limitation to the implementation process of the embodiments of the present invention.

[0034] To illustrate the technical solution of the present invention, the following will be described through specific embodiments.

[0035] A call content detection method provided by an embodiment of the present invention can be applied in an application environment as Figure 1 shown, in which the client communicates with the server through a network. Among them, the client includes but is not limited to computer devices such as a palm computer, a desktop computer, a notebook computer, an ultra-mobile personal computer (UMPC), a netbook, a cloud computer device, and a personal digital assistant (PDA). The server can be an independent server or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, Content Delivery Network (CDN), and big data and artificial intelligence platforms.

[0036] Please refer to Figure 2 , an embodiment of the present invention provides a call content detection method, wherein the method includes the following steps:

[0037] S100. Obtain the call recording of the target customer service and the target customer, and convert the call recording into text data;

[0038] S200. Parse the text data and extract the reply text of the target customer service;

[0039] S300. Based on the reply text, retrieve the target content with abnormal replies in the reply text through an abnormal knowledge base;

[0040] S400. Use the target content to form a prompt word of an anomaly detection model, and use the prompt word as an input to generate an anomaly detection result of the target customer service through the anomaly detection model.

[0041] Specifically, when implemented, the call content detection method of this embodiment effectively solves the problems existing in the communication process between enterprise customer service personnel and customers in the prior art through a series of systematic steps. The specific systematic and standardized detection process is as follows:

[0042] 1. Data acquisition and conversion: This method first obtains the call recordings between the customer service and the customer and converts them into text data. This process digitizes and texturizes the call content through technical means, providing a basis for subsequent analysis and detection. This avoids the instability of simply relying on the personal records and memories of customer service personnel, ensuring the integrity and accuracy of the data.

[0043] 2. Text parsing and extraction: By parsing the text data, the reply text of the customer service is extracted. This step uses natural language processing technology to accurately identify and extract the words of the customer service personnel, avoiding the subjectivity and inconsistency of manual screening. This automated extraction method ensures the objectivity and standardized processing of the data.

[0044] 3. Abnormal content retrieval: Based on the extracted reply text of the customer service, the target content with abnormal replies (referring to content suspected of violating regulations, but it still needs to be further determined whether it actually violates the regulations) is retrieved through the abnormal knowledge base. This process uses the pre-constructed abnormal knowledge base and combines semantic matching technology to quickly and accurately identify potential abnormal behaviors in the customer service's words. This method not only improves the detection efficiency but also reduces omissions and misjudgments caused by manual review, enhancing the stability of service quality.

[0045] 4. Application of the abnormal detection model: The abnormal detection model is used to generate the final abnormal detection results. Through pre-trained large language models and fine-tuning training, this abnormal detection model can reason based on the input prompt words (generated based on the target content with abnormal replies) to determine whether the customer service's words actually violate the regulations and provide a detailed detection report. This AI-based detection method not only improves the accuracy and efficiency of detection but also can provide timely feedback on the results, helping customer service personnel correct mistakes and enhance compliance.

[0046] The call content detection method of this embodiment realizes the standardized and systematic detection of customer service call content through systematic data processing, automated text parsing, accurate abnormal content retrieval, and large model-based abnormal detection. This method not only improves the stability of service quality but also significantly reduces compliance risks, providing an efficient and reliable solution for enterprises to detect customer service call content.

[0047] It can be understood that the call content detection method provided by the embodiments of the present invention can be applied to the call content detection scenarios related to the medical and health field. The following is a specific example:

[0048] Background: In the medical and health field, customer service personnel provide services such as appointment registration, disease consultation, and medication guidance to patients through phone calls. Due to the complexity and professionalism of medical information, if customer service personnel are slightly careless during communication, they may provide incorrect medical advice or mislead patients, thus triggering medical disputes.

[0049] Specific scenario: Imagine a hospital customer service representative receives a call from a patient asking about the usage and side effects of a medication. The representative might provide incorrect medication recommendations due to inaccurate memory or insufficient understanding of medication information.

[0050] The process of applying the method of the present invention:

[0051] 1. The system obtains the call recordings between customer service and patients in real time and converts them into text data through speech recognition technology.

[0052] 2. The system parses the text data and extracts the customer service representative’s response text.

[0053] 3. The system searches for abnormal reply content in the reply text based on the abnormal knowledge base.

[0054] 4. The system uses the anomaly detection model to generate anomaly detection results, prompting the customer service staff that there are errors in their answers and generating a detailed anomaly detection report.

[0055] Effect: Through the call content detection method of the present invention, hospitals can promptly discover and correct errors made by customer service personnel in communication, avoid medical disputes caused by incorrect medical advice, improve service quality, and enhance patient trust.

[0056] It is understandable that the call content detection method provided in the embodiment of the present invention can also be applied to call content detection scenarios related to the financial technology field. The following is a specific example:

[0057] Background: In the fintech sector, customer service representatives provide clients with financial product consultation, trading guidance, and risk warnings over the phone. Due to the complexity and specialized nature of financial services, even the slightest inadvertent communication could violate relevant laws and regulations or internal company regulations, leading to misunderstandings and incorrect investment decisions.

[0058] Specific scenario: Imagine a customer service representative at a financial institution receives a call from a customer inquiring about the yield and risks of a certain financial product. To facilitate the transaction, the representative might deliberately exaggerate the product's yield without fully explaining the potential risks.

[0059] The process of applying the method of the present invention:

[0060] 1. The system obtains the call recordings between customer service and customers in real time and converts them into text data through speech recognition technology.

[0061] 2. The system parses the text data and extracts the customer service representative’s response text.

[0062] 3. The system searches for abnormal reply content in the reply text based on the abnormal knowledge base.

[0063] 4. The system uses the anomaly detection model to generate anomaly detection results, prompting the customer service staff that there are errors in their answers and generating a detailed anomaly detection report.

[0064] Effect: Through the call content detection method of the present invention, financial institutions can promptly discover and correct violations by customer service personnel in communication, avoid customer complaints and legal disputes caused by misleading publicity, improve service quality, and enhance customer trust.

[0065] As demonstrated by the aforementioned examples in healthcare and fintech, the call content detection method of the present invention has significant application value in both fields. Through a systematic detection process, it can effectively identify and correct customer service staff errors and violations during communication, thereby improving service quality, reducing compliance risks, and providing strong support for digital transformation and risk management across various industries.

[0066] Furthermore, in one embodiment, the call content detection method, wherein obtaining a call recording between a target customer service representative and a target customer and converting the call recording into text data, specifically comprises the steps of:

[0067] Use recording equipment to capture the call audio stream between the target customer service representative and the target customer in real time, and perform noise reduction on the call audio stream to obtain the corresponding call recording;

[0068] Saving the call recording as an audio file and marking the audio file with a unique identifier;

[0069] The audio file is subjected to speech recognition by using speech recognition technology in combination with a context-aware algorithm, and is converted into corresponding text data.

[0070] During specific implementation, the specific implementation process of the steps in this embodiment is roughly as follows:

[0071] 1. Recording collection and noise reduction processing

[0072] Real-time audio capture: Use professional recording equipment (such as a high-quality microphone or telephone recording system) to capture the audio stream of the call between the target agent and the target customer in real time. Ensure that the recording equipment can capture the voice signals of both the agent and the customer to avoid signal interference.

[0073] Noise Reduction: During the recording process, real-time noise reduction is performed on the call audio stream. Advanced noise reduction algorithms (such as deep learning noise reduction models) are used to remove background noise and intermittent noise, improving voice quality. This step can significantly improve the accuracy of subsequent speech recognition.

[0074] 2. Call Recording Preservation and Identification

[0075] Save as an audio file: Save the denoised call audio stream as a high-fidelity audio file, usually in a standard audio format (such as WAV or MP3). Ensure the uniformity of the audio file format for subsequent processing.

[0076] Mark a unique identifier: Generate and mark a unique identifier (such as UUID or timestamp) for each audio file. The unique identifier should contain information such as the call time, customer service staff number, customer number, etc., for quick retrieval and management of audio files in the future.

[0077] 3. Speech Recognition and Context Awareness

[0078] Speech recognition: Use advanced speech recognition technology (such as a deep learning-based ASR model) to convert the saved audio file into text data. The speech recognition system should support multi-language and dialect recognition to adapt to the speech characteristics of different regions and customers.

[0079] Context awareness optimization: During the speech recognition process, combine context awareness algorithms (such as a Transformer-based context model) to dynamically adjust the recognition results. The context awareness algorithm can optimize the recognition results based on the context information of the call content, improving the accuracy and semantic coherence of the recognition. For example, for professional terms or vocabulary in a specific field, the system can automatically recognize and correct common pronunciation errors.

[0080] 4. Text Data Verification and Optimization

[0081] Text verification: Verify the generated text data to ensure the integrity and accuracy of the text. The verification process can include removing duplicate text, fixing sentence breaks, correcting spelling mistakes, etc.

[0082] Semantic integrity check: Use natural language processing technology (such as language model evaluation) to check the semantic integrity of the text data. Ensure the semantic coherence of each sentence or paragraph, and avoid incomplete or semantically chaotic text fragments caused by speech recognition errors.

[0083] Through the above process, this embodiment can achieve high-quality call recording acquisition, noise reduction processing, speech recognition, and text optimization, providing an accurate and complete data basis for subsequent call content detection.

[0084] Further, in one embodiment, for the call content detection method, where parsing the text data to extract the reply text of the target customer service specifically includes the steps:

[0085] Preprocess the text data, including removing irrelevant characters, unifying the text format, and performing text segmentation;

[0086] Use voiceprint recognition technology and combine it with the pre-labeled voice characteristics of the target customer service and the target customer to perform speaker role recognition on the preprocessed text data;

[0087] Extract the reply text of the target customer service from the preprocessed text data according to the speaker role recognition result.

[0088] Specifically in implementation, the specific implementation process of the steps in this embodiment is roughly as follows:

[0089] 1. Text data preprocessing

[0090] Remove irrelevant characters: Clean the text data, remove irrelevant content such as punctuation marks, special characters, and stop words to ensure the neatness of the text.

[0091] Unify the text format: Unify the text data into a standard format, such as unifying it to lowercase or uppercase, removing redundant spaces and line breaks to ensure the consistency of the text.

[0092] Text segmentation: Use natural language processing technology to split the text data into independent phrases, sentences, or paragraphs for subsequent analysis.

[0093] 2. Voiceprint recognition and speaker role annotation

[0094] Voiceprint feature extraction: Extract the voice characteristics of the target customer service and the target customer from the call recording. Use voiceprint recognition technology (such as a voiceprint recognition model based on deep learning) to extract the unique voiceprint features of each speaker.

[0095] Speaker role annotation: Combine the pre-labeled voice characteristics of the target customer service and the target customer to perform speaker role recognition on the preprocessed text data. Label the identity of the speaker ("customer service" or "customer") for each sentence or paragraph.

[0096] 3. Customer service reply text extraction

[0097] Filter customer service text: According to the speaker role recognition result, extract all text segments labeled as "customer service" from the preprocessed text data. These segments are the reply text of the target customer service.

[0098] Semantic integrity verification: Perform semantic integrity verification on the extracted reply text to ensure that the extracted text segments are semantically complete and coherent. For example, by checking the subject-predicate-object structure of the sentence or using a language model to evaluate the semantic integrity of the text segment.

[0099] Through the above process, this embodiment can achieve high-quality preprocessing of text data, speaker recognition, speaker role annotation, and extraction of customer service response text, providing an accurate and complete data basis for subsequent call content detection.

[0100] Further, in one embodiment, the call content detection method, wherein, based on the response text, the target content with abnormal responses in the response text is retrieved through an abnormal knowledge base, specifically including the steps of:

[0101] Construct an abnormal knowledge base;

[0102] Using a pre-trained semantic model, vectorize the abnormal knowledge base and the response text respectively to obtain the corresponding abnormal vector knowledge base and response text vectors;

[0103] Based on the abnormal vector knowledge base and the response text vectors, retrieve the target content with abnormal responses in the response text through the cosine similarity algorithm.

[0104] Further, the call content detection method, wherein, based on the abnormal vector knowledge base and the response text vectors, retrieving the target content with abnormal responses in the response text through the cosine similarity algorithm specifically includes the steps of:

[0105] Extract the vector representations of each discourse content in the response text vector;

[0106] Calculate the cosine similarity between the vector representation of each discourse content and each abnormal discourse vector in the abnormal vector knowledge base;

[0107] According to the calculation results, screen out all target discourse contents in the response text whose similarity scores are greater than the similarity threshold, and integrate all the target discourse contents to obtain the target content with abnormal responses.

[0108] Wherein, the response text consists of multiple independent discourse contents, and the discourse content can be an independent phrase, sentence or paragraph.

[0109] Specifically in implementation, the specific implementation process of the steps of this embodiment is roughly as follows:

[0110] 1. Construct an abnormal knowledge base

[0111] Sort out abnormal discourses: Collect and organize abnormal discourses, training manuals, and historical abnormal cases related to the industry (such as the insurance industry) to form a comprehensive abnormal knowledge base. Each abnormal discourse should be clearly marked with information such as abnormal type, keyword, semantic pattern, etc.

[0112] Annotation and Classification: Conduct detailed annotation on abnormal utterances, including abnormal types (such as misleading publicity, improper commitments, etc.), keywords, semantic patterns, etc., for subsequent semantic matching.

[0113] Storage and Management: Store the abnormal knowledge base in a database to ensure data security and accessibility. A relational database (such as MySQL) or a non-relational database (such as MongoDB) can be used for storage.

[0114] 2. Vectorization Processing

[0115] Select a Semantic Model: Select a pre-trained semantic model (such as BERT, RoBERTa, or other pre-trained language models) to convert text information into vector representations.

[0116] Vectorization of the Abnormal Knowledge Base: Perform vectorization processing on each abnormal utterance in the abnormal knowledge base to generate corresponding vector representations, and store these vectors in the abnormal vector knowledge base.

[0117] Vectorization of the Reply Text: Perform vectorization processing on the extracted customer service reply text to generate vector representations for each utterance content. Ensure that the dimensions of the reply text vectors are consistent with those of the vectors in the abnormal vector knowledge base for similarity calculation.

[0118] 3. Cosine Similarity Calculation

[0119] Extract Vector Representations of Utterance Content: Extract vector representations of each utterance content from the reply text vectors. For example, split the reply text into independent phrases, sentences, or paragraphs, and extract the vector representations of each phrase, sentence, or paragraph.

[0120] Calculate Similarity: Use the cosine similarity algorithm to calculate the similarity between the vector representations of each utterance content and each abnormal utterance vector in the abnormal vector knowledge base.

[0121] 4. Screen Content with Abnormal Replies

[0122] Set Similarity Threshold: Based on business requirements and historical data, set a similarity threshold (e.g., 0.7). Only when the cosine similarity between the vector representation of the utterance content and the abnormal utterance vector is higher than this similarity threshold is it considered that the utterance content has an abnormal reply.

[0123] Screen High-Similarity Content: According to the calculation results, screen out all utterance contents in the reply text with similarity scores greater than the similarity threshold, and mark these utterance contents as the target utterance contents with abnormal replies.

[0124] Integrated abnormal response content: Summarize all the target discourse contents with abnormal responses into a list to generate the target content with abnormal responses.

[0125] Through the above process, this embodiment can achieve high-quality abnormal knowledge base construction, vectorization processing, similarity calculation, and screening of abnormal response content, providing an accurate and complete data basis for subsequent call content detection.

[0126] Further, in one embodiment, the call content detection method, wherein, using the target content to form a prompt word of an abnormal detection model, and using the prompt word as an input, generating an abnormal detection result of the target customer service through the abnormal detection model, specifically includes the steps:

[0127] Based on a pre-trained large language model, construct an abnormal detection model through fine-tuning training;

[0128] Use the target content as background information, and combine with abnormal content descriptions to form the prompt word of the abnormal detection model;

[0129] Based on the prompt word, use the abnormal detection model to generate an abnormal detection result of the target customer service.

[0130] Further, the call content detection method, wherein, after generating the abnormal detection result of the target customer service based on the prompt word using the abnormal detection model, it specifically further includes the steps:

[0131] Analyze the abnormal detection result to determine whether the target customer service has abnormal discourse behavior;

[0132] If the target customer service has abnormal discourse behavior, generate an abnormal detection report including abnormal discourse, abnormal type, and improvement suggestions, and return the abnormal detection report to the target customer service terminal;

[0133] If the target customer service does not have abnormal discourse behavior, generate a standard compliance report and return the standard compliance report to the target customer service terminal.

[0134] Specifically in implementation, the specific implementation process of the steps of this embodiment is roughly as follows:

[0135] 1. Construct an abnormal detection model

[0136] Select a pre-trained model: Select a pre-trained large language model (such as GPT, BERT, or RoBERTa), which perform well in natural language processing tasks and can understand the semantics and context information of text.

[0137] Fine-tuning Training: Use the labeled abnormal and normal customer service discourse data to fine-tune the pre-trained large language model. The labeled data should include various types of abnormal behaviors (such as misleading publicity, improper commitments, etc.) and normal (i.e., compliant) customer service discourses. Through fine-tuning training, the generated anomaly detection model can accurately identify and judge abnormal discourse behaviors in the customer service reply content.

[0138] 2. Compose Prompt

[0139] Extract Target Content: Extract the target content with abnormal replies from the reply text, which is screened through previous steps (such as cosine similarity calculation).

[0140] Combine Abnormal Descriptions: Use the target content as background information and combine it with the abnormal content description to generate a structured prompt.

[0141] For example:

[0142] Suppose you are an expert in the insurance field, responsible for examining whether there are abnormal discourses in the communication between the customer service and the customer.

[0143] Customer service reply content: "……";

[0144] Related abnormal discourses in the abnormal knowledge base: "1. ……; 2. ……; 3. ……";

[0145] Abnormal content description: If the central idea and view of the customer service reply content are the same as a certain abnormal discourse in the current abnormal knowledge base, it is considered that the reply content is abnormal;

[0146] Output format: Answer [Yes or No], if abnormal, return the abnormal discourse closest to the content of the abnormal knowledge base.

[0147] Optimize Prompt: Use natural language generation technology (such as GPT or other generative models) to optimize the prompt, ensure that the prompt has clear semantics, natural expression, and can guide the anomaly detection model to accurately make anomaly judgments.

[0148] 3. Generate Anomaly Detection Results

[0149] Input Prompt: Convert the optimized prompt into an input format acceptable to the anomaly detection model and pass it to the anomaly detection model.

[0150] Model Inference: The anomaly detection model makes inferences based on the input prompt, analyzes the semantic association between the customer service reply text and the abnormal discourses, and generates anomaly detection results.

[0151] 4. Analyze Anomaly Detection Results

[0152] Analysis result: Analyze the output result of the anomaly detection model to determine whether there are abnormal speech behaviors of the target customer service.

[0153] Generate a report:

[0154] Anomaly report: If abnormal speech behaviors are detected, generate a detailed anomaly detection report. The report content includes the specific content of the abnormal speech, the corresponding abnormal type, improvement suggestions, etc.

[0155] Compliance report: If no abnormal speech behaviors are detected, generate a standard compliance report to confirm that the communication behaviors of the customer service staff comply with the company's norms and laws and regulations.

[0156] 5. Return the detection report

[0157] Real-time feedback: Push the anomaly detection report or the standard compliance report to the target customer service terminal in real time through the customer service management system. The report should be presented in an easy-to-understand format to help customer service staff promptly understand the problems in communication and make improvements.

[0158] Record and analysis: Store the detection report in the database for subsequent analysis and management. Record the abnormal content and frequency to provide data support for the training and management of customer service staff.

[0159] Through the above process, this embodiment can achieve the construction of a high-quality anomaly detection model, the generation of prompt words, the generation of anomaly detection results, result analysis, and report return, providing comprehensive and accurate technical support for call content detection.

[0160] As can be seen from the above method embodiments, the call content detection method provided by the present invention includes: obtaining the call recording of the target customer service and the target customer, and converting the call recording into text data; parsing the text data to extract the reply text of the target customer service; based on the reply text, retrieving the target content with abnormal replies in the reply text through an anomaly knowledge base; using the target content to form the prompt words of the anomaly detection model, and taking the prompt words as input to generate the anomaly detection result of the target customer service through the anomaly detection model. In this way, through the method of the present invention, it can solve the problems in the prior art that in the process of the call communication between the customer service staff and the customer in an enterprise, it often relies on the personal experience and ability of the customer service staff, lacks systematic and standardized anomaly detection means for call content, resulting in unstable service quality and high compliance risks.

[0161] It should be understood that although the present application provides method operation steps as described in the embodiments or flowcharts, more or fewer operation steps may be included based on routine or non-creative labor, and these operation steps do not necessarily need to be executed in sequence according to the order of the embodiments or flowcharts. The step order listed in the embodiments or flowcharts is only one way among many step execution orders and does not represent the only execution order. It should be noted that there is not necessarily a certain sequence between the above steps. Those of ordinary skill in the art can understand according to the description of the embodiments of the present invention that in different embodiments, the above steps can have different execution orders, that is, they can be executed in parallel, or can be exchanged, etc. Moreover, at least a part of the steps in the embodiments or flowcharts may include multiple sub-steps or multiple stages. These sub-steps or stages do not necessarily need to be executed at the same time, but can be executed at different times, and the execution order of these sub-steps or stages is not necessarily sequential, but can be executed alternately, alternately or synchronously with at least a part of other steps or sub-steps or stages of other steps.

[0162] Based on the above method embodiments, please refer to Figure 3 , another embodiment of the present invention further provides a call content detection device, wherein the device includes:

[0163] An acquisition module 11, configured to acquire a call recording of a target customer service and a target customer, and convert the call recording into text data;

[0164] An analysis module 12, configured to analyze the text data and extract the reply text of the target customer service;

[0165] A retrieval module 13, configured to retrieve, based on the reply text, target content with abnormal replies in the reply text through an abnormal knowledge base;

[0166] An assembly module 14, configured to use the target content to assemble a prompt word of an anomaly detection model, and use the prompt word as an input to generate an anomaly detection result of the target customer service through the anomaly detection model.

[0167] Further, in one embodiment, for the call content detection device, wherein the acquisition module 11 is specifically configured to:

[0168] Use a recording acquisition device to capture the call audio stream of the target customer service and the target customer in real time, and perform noise reduction processing on the call audio stream to obtain the corresponding call recording;

[0169] Save the call recording as an audio file, and mark a unique identifier for the audio file;

[0170] The audio file is subjected to speech recognition by using speech recognition technology in combination with a context-aware algorithm, and is converted into corresponding text data.

[0171] Furthermore, in one embodiment, in the call content detection device, the parsing module 12 is specifically configured to:

[0172] Preprocessing the text data, including removing irrelevant characters, unifying the text format, and segmenting the text;

[0173] Using voiceprint recognition technology and combining pre-labeled voice features of the target customer service representative and the target customer, the pre-processed text data is used to identify the speaker role;

[0174] According to the speaker role recognition result, the reply text of the target customer service is extracted from the pre-processed text data.

[0175] Furthermore, in one embodiment, in the call content detection device, the retrieval module 13 is specifically configured to:

[0176] Build an abnormal knowledge base;

[0177] Using a pre-trained semantic model, the abnormal knowledge base and the reply text are respectively vectorized to obtain corresponding abnormal vector knowledge base and reply text vector;

[0178] Based on the abnormal vector knowledge base and the reply text vector, the target content with abnormal replies in the reply text is retrieved by using a cosine similarity algorithm.

[0179] Furthermore, the call content detection device, wherein the method of retrieving target content containing abnormal replies in the reply text by using a cosine similarity algorithm based on the abnormal vector knowledge base and the reply text vector, specifically includes:

[0180] Extracting a vector representation of each utterance in the reply text vector;

[0181] Calculating the cosine similarity between the vector representation of each of the utterance contents and each abnormal utterance vector in the abnormal vector knowledge base;

[0182] According to the calculation result, all target speech contents with similarity scores greater than a similarity threshold are screened out from the reply text, and all the target speech contents are integrated to obtain target contents with abnormal replies.

[0183] Furthermore, in one embodiment, in the call content detection device, the component module 14 is specifically configured to:

[0184] Based on a pre-trained large language model, an anomaly detection model is constructed through fine-tuning training;

[0185] Taking the target content as background information and combining it with the description of abnormal content to form the prompt words of the anomaly detection model;

[0186] Based on the prompt words, using the anomaly detection model to generate the anomaly detection result of the target customer service.

[0187] Further, for the call content detection device, after generating the anomaly detection result of the target customer service based on the prompt words using the anomaly detection model, it specifically further includes:

[0188] Analyzing the anomaly detection result to determine whether the target customer service has abnormal discourse behavior;

[0189] If the target customer service has abnormal discourse behavior, generate an anomaly detection report including abnormal discourse, anomaly type and improvement suggestions, and return the anomaly detection report to the target customer service terminal;

[0190] If the target customer service does not have abnormal discourse behavior, generate a standard compliance report and return the standard compliance report to the target customer service terminal.

[0191] It should be noted that in the device embodiment of the present invention, for the information interaction, execution process, etc. between the above modules, since it is based on the same concept as the method embodiment of the present invention, its specific functions and the technical effects brought, specifically, reference can be made to the method embodiment part described above, and details will not be repeated here.

[0192] Based on the above method embodiment, another embodiment of the present invention further provides a computer device, which can be a server, and its internal structure diagram can be as Figure 4 shown. The computer device includes a processor, a memory, a network interface and a database connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The network interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, it realizes the functions or steps of the call content detection method on the server side in any of the above method embodiments.

[0193] Based on the above method embodiment, another embodiment of the present invention further provides a computer device, which can be a client, and its internal structure diagram can be as Figure 5As shown. The computer device includes a processor, a memory, a network interface, a display screen, and an input device connected via a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The network interface of the computer device is used to communicate with an external terminal via a network connection. When the computer program is executed by the processor, it implements the functions or steps on the client side of the call content detection method in any of the above method embodiments.

[0194] Those skilled in the art can understand that Figure 4 and Figure 5 the structural schematic diagram shown in is only a schematic diagram of a part of the structure related to the solution of the present invention, and does not constitute a limitation on the computer device to which the solution of the present invention is applied. The specific computer device may include more components than those shown in the figure, or combine certain components, or have different component arrangements.

[0195] Among them, the so-called processor may be a CPU, and the processor may also be other general-purpose processors, digital signal processors (DSPs), application specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.

[0196] Among them, the memory includes a readable storage medium, an internal memory, etc. The internal memory can be the memory of a computer device, and the internal memory provides an environment for the operation of the operating system and computer-readable instructions in the readable storage medium. The readable storage medium can be the hard disk of a computer device, and in some other embodiments, it can also be an external storage device of the computer device. For example, a plug-in hard disk, a Smart Media Card (SMC), a Secure Digital (SD) card, a Flash Card, etc. equipped on the computer device. Further, the memory can also include both the internal storage unit of the computer device and the external storage device. The memory is used to store the operating system, application programs, a BootLoader, data, and other programs, such as the program code of a computer program. The memory can also be used to temporarily store the data that has been output or will be output.

[0197] Based on the above method embodiments, another embodiment of the present invention further provides a computer-readable storage medium storing a computer program, wherein the computer program, when executed by a processor, implements the call content detection method in any one of the above method embodiments. The computer-readable storage medium can be non-volatile or volatile.

[0198] It should be noted that for the functions or steps that the above computer-readable storage medium or computer device can achieve, and the technical effects brought by the functions / steps, reference can be made to the relevant descriptions in the foregoing method embodiments. To avoid repetition, they will not be described one by one here.

[0199] Those of ordinary skill in the art can understand that all or part of the processes in the methods of the above embodiments can be completed by instructing relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above methods. Among them, any reference to a memory, storage, database, or other medium used in the embodiments provided in the present application can include non-volatile and / or volatile memories. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and Rambus dynamic RAM (RDRAM), etc. The memory components or memories of the operating environment disclosed herein are intended to include one or more of these and / or any other suitable types of memories.

[0200] Those skilled in the art can clearly understand that, for the convenience and brevity of description, in the device embodiments of the present invention, only the above-mentioned division of each functional unit and module is used for illustration. In actual applications, the above functions can be allocated to different functional units and modules according to needs, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. Each functional unit and module in the embodiments can be integrated in a processing unit, or each unit can exist physically alone, or two or more units can be integrated in one unit. The above integrated units can be implemented in the form of hardware or in the form of software functional units. In addition, the specific names of each functional unit and module are only for the convenience of mutual distinction and do not limit the protection scope of the present invention. The specific working processes of the units and modules in the above device can refer to the corresponding processes in the foregoing method embodiments and will not be repeated here. If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium.

[0201] Those of ordinary skill in the art can realize that the units and algorithm steps of each example described in combination with the embodiments disclosed herein can be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are executed in a hardware or software manner depends on the specific application and design constraints of the technical solution. A professional technician can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of the present invention.

[0202] In the embodiments provided by the present invention, it should be understood that the disclosed device / computer equipment and method can be implemented in other ways. For example, the device / computer equipment embodiments described above are only illustrative. For example, the division of modules or units is only a logical function division. In actual implementation, there may be other division methods. 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 displayed or discussed couplings or direct couplings or communication connections between each other can be through some interfaces. The indirect couplings or communication connections of devices or units can be in electrical, mechanical or other forms.

[0203] The units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they can be located in one place, or can be distributed to multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0204] It should be noted that if non-company software tools or components appear in the embodiments of this application, they are only for example introduction and do not represent actual use. The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements for some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included in the protection scope of the present invention.

Claims

1. A call content detection method, characterized in that, Including: Obtain the call recording of the target customer service and the target customer, and convert the call recording into text data; Parse the text data to extract the reply text of the target customer service; Based on the reply text, retrieve the target content with abnormal replies in the reply text through the abnormal knowledge base; Use the target content to form a prompt for the abnormal detection model, and use the prompt as input to generate the abnormal detection result of the target customer service through the abnormal detection model.

2. The call content detection method according to claim 1, wherein The obtaining of the call recording of the target customer service and the conversion of the call recording into text data include: Use a recording acquisition device to capture the call audio stream of the target customer service and the target customer in real time, and perform noise reduction processing on the call audio stream to obtain the corresponding call recording; Save the call recording as an audio file and mark a unique identifier for the audio file; Use speech recognition technology and combine it with a context-aware algorithm to perform speech recognition on the audio file and convert it into corresponding text data.

3. The call content detection method according to claim 1, characterized in that, The parsing of the text data to extract the reply text of the target customer service includes: Perform preprocessing on the text data, including removing irrelevant characters, unifying the text format, and text segmentation; Use voiceprint recognition technology and combine it with the pre-labeled voice features of the target customer service and the target customer to perform speaker role recognition on the preprocessed text data; According to the speaker role recognition result, extract the reply text of the target customer service from the preprocessed text data.

4. The call content detection method according to claim 1, wherein The retrieving of the target content with abnormal replies in the reply text based on the reply text through the abnormal knowledge base includes: Construct an abnormal knowledge base; Use a pre-trained semantic model to perform vectorization processing on the abnormal knowledge base and the reply text respectively to obtain the corresponding abnormal vector knowledge base and reply text vector; Based on the abnormal vector knowledge base and the reply text vector, retrieve the target content with abnormal replies in the reply text through the cosine similarity algorithm.

5. The call content detection method according to claim 4, wherein, The retrieving of the target content with abnormal replies in the reply text based on the abnormal vector knowledge base and the reply text vector through the cosine similarity algorithm includes: Extract the vector representation of each discourse content in the reply text vector; Calculate the cosine similarity between the vector representation of each discourse content and each abnormal discourse vector in the abnormal vector knowledge base; According to the calculation result, screen out all target discourse contents in the reply text whose similarity score is greater than the similarity threshold, and integrate all the target discourse contents to obtain the target content with abnormal replies.

6. The call content detection method according to claim 1, wherein The using of the target content to form a prompt for the abnormal detection model and using the prompt as input to generate the abnormal detection result of the target customer service through the abnormal detection model includes: Construct an abnormal detection model through fine-tuning training based on a pre-trained large language model; Use the target content as background information and combine it with the abnormal content description to form the prompt of the abnormal detection model. Based on the prompt word, use the anomaly detection model to generate the anomaly detection result of the target customer service agent.

7. The call content detection method according to claim 6, wherein After generating the anomaly detection result of the target customer service agent based on the prompt word, it further includes: Analyze the anomaly detection result to determine whether the target customer service agent has abnormal discourse behavior; If the target customer service agent has abnormal discourse behavior, generate an anomaly detection report including abnormal discourse, anomaly type, and improvement suggestions, and return the anomaly detection report to the target customer service agent terminal; If the target customer service agent does not have abnormal discourse behavior, generate a standard compliance report and return the standard compliance report to the target customer service agent terminal.

8. A call content detection device, characterized in that, It includes: An acquisition module for acquiring the call recording between the target customer service agent and the target customer, and converting the call recording into text data; An analysis module for analyzing the text data to extract the response text of the target customer service agent; A retrieval module for retrieving, based on the response text, the target content with abnormal responses in the response text through an anomaly knowledge base; A composition module for using the target content to compose the prompt word of the anomaly detection model, and using the prompt word as the input to generate the anomaly detection result of the target customer service agent through the anomaly detection model.

9. A computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the call content detection method according to any one of claims 1-7.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the call content detection method according to any one of claims 1-7.

Images