Intelligent customer service assistant processing system for reporting telephone problems of hotels

Abstract

The invention discloses a hotel telephone problem fault reporting intelligent customer service assistant processing system, and relates to the technical field of intelligent operation and maintenance, the customer service assistant processing system comprises a voice-to-text system, a text display system, an AI auxiliary system and a repair reporting management system, and the systems are in electric signal connection; the voice-to-text system is used for collecting voice data and preprocessing the voice data by using a voice processing algorithm. According to the intelligent customer service assistant processing system for reporting the fault of the hotel telephone problem, through the prompt given by the AI large model, the customer service specialist does not need high professional ability, the hotel personnel can be well fed back, the labor cost and the training cost are reduced, through the prompt given by the AI large model, the judgment of the customer service specialist on the problem is more accurate, and the efficiency is improved. Therefore, misjudgment of operation and maintenance engineers in door-to-door processing caused by unnecessary circulation is effectively avoided, downstream labor cost is reduced, hotel personnel are rapidly helped to solve problems, and cost reduction and benefit improvement of enterprises are realized.

Description

Technical Field

[0001] The present invention relates to the field of intelligent operation and maintenance technology, and in particular to an intelligent customer service assistant processing system for hotel telephone problem reporting. Background Art

[0002] The hotel industry uses a large number of networked hardware devices, which require a lot of operation and maintenance work during operation. The operation and maintenance reporting needs are usually handed over to specific operation and maintenance personnel in the form of work order processes. When hotel staff find a problem, they need to call a fixed number and describe the problem to the operation and maintenance customer service specialist. The customer service specialist needs to obtain useful information based on the problem description, judge the problem, and create a work order. The operation and maintenance related personnel then handle these work orders. This process mainly relies on the professional knowledge and experience of the operation and maintenance customer service specialist to locate the problem and describe the work order.

[0003] In the prior art, a large number of work orders do not actually require on-site service by operation and maintenance engineers, and can be solved by telephone communication. However, customer service with insufficient professional ability cannot effectively guide hotel staff to solve the problem by themselves, resulting in unnecessary manpower costs. In addition, operation and maintenance customer service requires high professionalism, and customer service personnel have high mobility and are not easy to recruit, resulting in high manpower training costs. For this reason, an intelligent customer service assistant processing system for hotel telephone problem reporting is proposed to improve the accuracy of problem location, reduce unnecessary manpower costs, and improve operation and maintenance efficiency. Summary of the invention

[0004] To achieve the above objectives, the present invention is implemented through the following technical solutions: an intelligent customer service assistant processing system for hotel telephone problem reporting, the customer service assistant processing system includes a voice-to-text system, a text display system, an AI auxiliary system and a repair management system, wherein the electrical signals between the systems are connected;

[0005] The speech-to-text system collects speech data and uses a speech processing algorithm to pre-process the speech data, converting the speech data into text, ensuring the accuracy of the original data, improving the efficiency and accuracy of subsequent processing, and achieving efficient speech-to-text conversion;

[0006] The text display system displays the conversation text and text operations after voice conversion, assisting customer service specialists to view the conversation content in real time, improving the readability and manageability of the conversation content, and facilitating customer service specialists to quickly understand and respond to customer issues;

[0007] The AI-assisted system assists in problem consultation, locates problems to guide the solution of related problems, provides professional problem analysis and solutions, and improves the accuracy and efficiency of problem solving;

[0008] The repair management system is used to generate repair service work orders for problems that hotel staff cannot solve by themselves, record repair information, and track repair work services.

[0009] Preferably, in the speech-to-text system, the process of converting speech data into text includes:

[0010] The voice signal is collected by a high-sensitivity microphone and converted into a digital signal by an analog-to-digital converter to ensure the quality of the recording equipment is good, so as to reduce noise and distortion and improve the clarity of the voice signal. The collected voice signal is pre-processed to improve the accuracy of subsequent voice recognition. The pre-processing process includes noise reduction and filtering operations. The background noise and interference in the recording are removed by denoising to make the voice signal clearer. The voice signal is filtered to remove unnecessary frequency components to further improve the clarity of the voice signal.

[0011] The pre-processed speech data is further subjected to speech noise reduction using a discrete wavelet transform function. The speech data is converted into a time-frequency diagram, a wavelet basis is selected, the wavelet function is defined using the wavelet basis, and the wavelet function is infinitely scaled and displaced to obtain a set of wavelet functions. The low-frequency approximate coefficients and the high-frequency detail coefficients are analyzed. Discrete wavelet transform is also called wavelet decomposition. For a given original signal, a scaling function is used to perform a convolution operation with the original signal. The obtained value is called an approximate coefficient, also called low-frequency information, which is used to describe the overall trend of the signal. The wavelet function is used to perform a convolution operation with the original signal. The obtained value is called detail coefficient, also known as high-frequency information, which is used to describe the detailed characteristics of the signal. The above is a layer of decomposition. In wavelet decomposition, after multiple decompositions, that is, the approximate coefficients of each layer are further decomposed. Speech denoising ensures the accuracy of the most original data. Discrete wavelet transform function denoising is based on the different properties of the wavelet coefficients of the original signal and noise at different scales. The corresponding rules are constructed to process the wavelet coefficients containing the signal in the wavelet domain. A wavelet basis defines two groups of functions: wavelet function and scale function. Through discrete wavelet decomposition, threshold processing, and wavelet reconstruction, noise reduction is finally achieved.

[0012] The pre-processed speech signal is framed and the long speech is divided into multiple short segments. The length of each frame is 10-30 milliseconds, and the frame shift is about 10 milliseconds. According to the physiological characteristics of the human ear, the sound waveform is standardized and digitized, and the Mel frequency cepstral coefficient features are extracted for subsequent speech recognition;

[0013] The extracted features are matched and recognized using a trained speech recognition model to generate corresponding text content, wherein the speech recognition model is built based on a hidden Markov model. A large amount of annotated speech data is combined with the hidden Markov model to train the speech recognition model to recognize different speech units, thereby matching audio signals and speech and identifying candidate word sequences.

[0014] Preferably, in the text display system, the display process of the dialogue text and the text operation includes:

[0015] The text display system receives the converted text data from the speech recognition system and further processes the recognized text, including correcting recognition errors, formatting text, and adding punctuation marks. It automatically corrects recognition errors in the text data to improve the accuracy of the text, further formats the corrected text, adjusts the font, font size, and color to meet the requirements of the customer service software interface, and automatically adds appropriate punctuation marks according to the context to make the text more fluent and easy to read;

[0016] The processed text is displayed in real time on the customer service software work interface of the customer service specialist, and a conversation management function is provided to track the conversation history and conversation status, so as to help the customer service specialist understand the context of the conversation. By providing a complete conversation history function to record all the communication content between the customer and the customer service specialist, it is helpful for the customer service specialist to review the conversation process when necessary and understand the context of the conversation. By tracking the status of the conversation, it is helpful for the customer service specialist to better manage his or her own workflow and ensure that no important information is missed;

[0017] Automatically identify and highlight customer speech to help customer service specialists quickly distinguish between the two parties in a conversation and associate the current conversation text with historical records so that customer service specialists can quickly understand the customer's historical problems and solutions.

[0018] Preferably, the AI-assisted system includes an AI large model module, a vector knowledge base module, a problem location module, and an intelligent guidance module;

[0019] The AI ​​big model module is used to analyze and match problems using the AI ​​big model;

[0020] The vector knowledge base module is used to store common solutions, equipment information, and troubleshooting steps for hotel phone problems, and professionally supplement the knowledge content of the AI ​​big model, provide domain-specific knowledge support, enhance the domain expertise of the AI ​​big model, and improve the professionalism and accuracy of problem handling;

[0021] The problem location module automatically locates the type of problem and the corresponding solution based on the user's problem and the vector knowledge base information, thereby improving the accuracy of problem location and reducing manual intervention and unnecessary door-to-door service.

[0022] The intelligent guidance module, based on the problem location result, cooperates with the customer service specialist to guide the hotel staff to solve the problem by themselves through dynamically generated guidance words, thereby improving the user's ability to solve problems by themselves and reducing labor costs.

[0023] Preferably, the problem location and solution generation process specifically includes:

[0024] Based on the preprocessed customer speech text, the question text related to the hotel telephone question is extracted from it, and the question text is converted into a semantic vector using the semantic model pre-trained by the question location module. The semantic vector can capture the semantic information of the text so as to be compared with the information stored in the vector database;

[0025] Using vector similarity search technology, the cosine similarity value between the semantic vector of the user's question and each vector in the vector database is calculated. Based on the similarity value, the most relevant information to the user's question is retrieved and input into the AI ​​big model module. The vector database is a pre-built database containing common problems and their solutions, and each problem and its solution is converted into a semantic vector.

[0026] Use the AI ​​big model (Transformer model) to understand the content of the question, capture contextual information, conduct in-depth analysis of the hotel phone questions, and match them with the information in the vector knowledge base. Based on the retrieved vectors and corresponding solutions, further match the problem types, including equipment failure, service problems, and operation problems.

[0027] Generate corresponding solutions based on the matched problem type and the analysis of the AI ​​big model. The solutions include specific operation steps, troubleshooting methods, and contacting maintenance personnel.

[0028] The proposed solutions are optimized through grammar checking and semantic verification to ensure their accuracy, fluency and readability, and the optimized proposed solutions are output to customer service specialists in text form so that they can provide assistance to hotel staff based on the solutions.

[0029] Preferably, the calculation expression of the cosine similarity value is:

[0030]

[0031] Among them, CS is the cosine similarity value, A is the semantic vector of the user's question, B is the semantic vector of the question in the vector database, and A iis the value of vector A in the i-th dimension, representing the semantic weight of the user's question in this dimension, B i It is the value of vector B in the i-th dimension, representing the semantic weight of the question in the database in this dimension. n is the number of dimensions of the vector. The value of cosine similarity ranges from -1 to 1, where -1 means completely opposite, 0 means no correlation, and 1 means exactly the same.

[0032] Preferably, in the intelligent guidance module, the process of generating the guide words includes:

[0033] Based on the results of problem location, the intelligent guidance module further analyzes the specific context of the problem, such as the urgency of the problem, the skill level of hotel staff, etc., and dynamically generates customized guidance for the proposed solution based on the results of the context analysis and the optimized proposed solution output to the customer service specialist, including simple step-by-step instructions, troubleshooting suggestions, and operation guides;

[0034] Integrate the introductory language with the contextual information of the conversation, so that the introductory language is closely related to the customer's question, provide targeted guidance, and then optimize the introductory language. The optimized introductory language is provided to the customer service specialist as a reference for communicating with hotel staff;

[0035] Based on the content of the introductory statement, the customer service specialist guides hotel personnel to troubleshoot and resolve issues. During the communication between the customer service specialist and the customer, the customer service specialist collects customer feedback, records whether the problem has been resolved, and any new problems that may be encountered during the resolution process.

[0036] Preferably, the repair management system includes a work order generation and management module and a repair record and analysis module. The work order generation and management module automatically generates a work order for problems that require on-site service by an operation and maintenance engineer, tracks the processing progress of the work order, optimizes the work order processing process, and improves the operation and maintenance efficiency;

[0037] The repair record and analysis module records and analyzes the detailed information of each repair request, including the repair time, problem description, and processing progress, to help the hotel management team track and analyze the repair work;

[0038] Among them, in the work order generation and management module, the process of tracking the generation and processing progress of the work order includes:

[0039] When the customer service specialist finds that the hotel staff cannot solve the problem by themselves, he / she shall submit a repair request through the repair management system. The submitted content shall include the fault description, faulty equipment information, the information of the person reporting the repair, and the expected repair time.

[0040] The repair management system reviews the submitted repair request and automatically generates a repair service work order after passing the review;

[0041] The repair management system automatically assigns the work order to the operation and maintenance engineer, and after the work order is assigned, the system notifies the relevant operation and maintenance engineer through SMS and system notification;

[0042] The operation and maintenance engineer conducts on-site maintenance according to the work order information and solves the problems reported by the hotel staff. During the process, the engineer records the processing steps and progress and updates the work order status. The recorded processing steps include arriving at the site, diagnosing the fault, repairing or unable to complete, and repairing completed steps;

[0043] Based on the updated work order status, if the operation and maintenance engineer determines that the maintenance task cannot be completed alone, the maintenance task status is analyzed, the task difficulty assessment coefficient is calculated, and the task difficulty is determined to determine whether to transfer the order or request assistance;

[0044] The progress of work order processing is tracked in real time. Customer service specialists track the status of work orders and keep in touch with hotel staff to update the progress. After the maintenance is completed, the operation and maintenance engineer updates the work order status to "Completed". The hotel staff confirms that the problem has been resolved and the system records the confirmation information.

[0045] Preferably, the task difficulty determination process includes:

[0046] The operation and maintenance engineer updates the work order status in the repair management system and marks it as "needs further support". Based on the on-site conditions and maintenance needs, the engineer enters the evaluation indicators that affect the difficulty of the task, including the estimated maintenance time, fault complexity, and expected workload. According to historical data and experience, the estimated maintenance time is divided into different intervals, and each interval is assigned a corresponding score. 0-30 minutes is set as 1 point, 30-60 minutes is set as 2 points, and so on. According to the type of fault, the number of systems or components involved, and the diagnostic difficulty factor, the fault complexity is divided into different levels, and each level is assigned a score. Simple faults are 1 point, medium-complexity faults are 2 points, and complex faults are 3 points. The number of personnel, type and quantity of tools, and spare parts demand factors required in the maintenance process are comprehensively analyzed to divide the expected workload into different levels, and each level is assigned a score. Low-workload tasks are 1 point, medium-workload tasks are 2 points, and high-workload tasks are 3 points.

[0047] The system extracts data of similar tasks from historical maintenance records, including past maintenance time, resource consumption and success rate, analyzes the maintenance difficulty and determines different task difficulty levels, namely, level 1 difficulty level, level 2 difficulty level, level 3 difficulty level and level 4 difficulty level. The difficulty level increases from level 1 to level 4.

[0048] Combined with the evaluation indicator-related data input by the engineer, each evaluation indicator is quantitatively scored, the task difficulty evaluation coefficient is calculated, and combined with the preset different task difficulty levels, the corresponding difficulty threshold is matched for each difficulty level;

[0049] Based on the set difficulty level, the corresponding decision support is matched and suggestions are provided to engineers. Engineers are advised to choose to transfer the order or request assistance. For tasks with difficulty levels 1 and 2, engineers are advised to request assistance. For tasks with difficulty levels 3 and 4, engineers are advised to transfer the order. If the order is transferred, the work order will be transferred to other professional operation and maintenance teams. If assistance is requested, reinforcements will be requested and other engineers will be called in to jointly handle the maintenance task.

[0050] The calculation expression of the task difficulty evaluation coefficient is:

[0051]

[0052] Among them, D is the task difficulty assessment coefficient, and the value range of D is between 0 and 1. T is the expected maintenance time score, which reflects the time required to complete the task. C is the fault complexity score, which is a dimensionless score value indicating the complexity of the fault. W is the expected workload score, which is a dimensionless score value indicating the workload required to complete the task. a, b, and c are weight coefficients used to adjust the influence of each evaluation indicator on the task difficulty assessment coefficient. ω is a normalization factor used to ensure that the value of D is between 0 and 1.

[0053] The plurality of difficulty levels correspond one-to-one to the plurality of difficulty thresholds, specifically:

[0054] The difficulty threshold for level 1 difficulty is 0 <D≤D 1 ;

[0055] The difficulty threshold of the secondary difficulty level is D 1 <D≤D 2 ;

[0056] The difficulty threshold of the third level difficulty level is D 2 <D≤D 3 ;

[0057] The difficulty threshold of the fourth level difficulty level is D 3 <D<1;

[0058] Among them, D is the task difficulty assessment coefficient, D 1 is the upper threshold of the first difficulty level and the lower threshold of the second difficulty level, D 2 is the upper threshold of the second level difficulty level and the lower threshold of the third level difficulty level, D 3 is the upper threshold of the third level of difficulty and the lower threshold of the fourth level of difficulty, D 1=0.25, D 2 =0.5, D 3 =0.75.

[0059] Preferably, in the repair recording and analysis module, the process of recording and analyzing the repair request information includes:

[0060] The repair management system stores all repair records and processing progress information in the database to ensure data security and accessibility, and analyzes the collected data to evaluate the work efficiency of operation and maintenance engineers, including response time and repair time;

[0061] Generate maintenance work reports based on analysis results, including maintenance completion rate, average response time, maintenance success rate, and propose improvement suggestions;

[0062] Based on the maintenance work report, analyze the fault data, identify the recurring fault modes, provide a basis for preventive maintenance, and put forward suggestions for improving the maintenance process and improving maintenance efficiency;

[0063] Use historical data to predict potential failures, schedule maintenance work in advance, reduce unexpected downtime, and collect feedback from hotel staff on maintenance services and integrate them into analysis reports for continuous improvement of service quality.

[0064] The present invention provides a hotel telephone problem reporting intelligent customer service assistant processing system. It has the following beneficial effects:

[0065] 1. The hotel’s intelligent customer service assistant processing system for reporting telephone problems uses the prompts given by the AI ​​big model. Customer service specialists do not need high professional capabilities and can provide good feedback to hotel staff, reducing labor costs and training costs. In addition, through the prompts given by the AI ​​big model, customer service specialists can judge problems more accurately, effectively avoiding unnecessary transfers and misjudgments that lead to on-site handling by operation and maintenance engineers, reducing downstream labor costs, and quickly helping hotel staff solve problems, thereby achieving cost reduction and efficiency improvement for the company.

[0066] 2. The hotel's telephone problem reporting intelligent customer service assistant processing system, through the integrated AI large model and vector knowledge base, can quickly and accurately understand customer problems and provide solutions. Through natural language processing technology, it converts customer voice into text, and uses pre-trained semantic models to convert text into semantic vectors, which are matched with the information in the vector database to quickly locate the type of problem. The intelligent processing method greatly reduces the time for manual understanding and recording of problems in traditional customer service and improves the response speed. BRIEF DESCRIPTION OF THE DRAWINGS

[0067] Figure 1 This is a module structure diagram of an intelligent customer service assistant processing system for reporting hotel telephone problems according to the present invention;

[0068] Figure 2 A flow chart for problem location and solution generation of the present invention;

[0069] Figure 3 This is a flow chart for tracking the generation and processing progress of a work order of the present invention. DETAILED DESCRIPTION

[0070] The present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments. The embodiments of the present invention are provided for the purpose of illustration and description, and are not intended to be exhaustive or to limit the present invention to the disclosed forms. Many modifications and variations will be apparent to those of ordinary skill in the art. The embodiments are selected and described in order to better illustrate the principles and practical applications of the present invention, and to enable those of ordinary skill in the art to understand the present invention and thereby design various embodiments with various modifications suitable for specific uses.

[0071] The first embodiment, as Figure 1 , Figure 2 As shown, the present invention provides a technical solution: an intelligent customer service assistant processing system for hotel telephone problem reporting, the customer service assistant processing system includes a speech-to-text system, a text display system, an AI auxiliary system and a repair management system, wherein the electrical signals between the systems are connected;

[0072] The speech-to-text system collects speech data and uses speech processing algorithms for preprocessing to convert speech data into text, ensure the accuracy of the original data, improve the efficiency and accuracy of subsequent processing, and achieve efficient speech-to-text conversion. The speech signal is collected through a high-sensitivity microphone and converted into a digital signal through an analog-to-digital converter to ensure the good quality of the recording equipment to reduce noise and distortion, improve the clarity of the speech signal, and preprocess the collected speech signal to improve the accuracy of subsequent speech recognition. The preprocessing process includes noise reduction and filtering operations. Denoising removes background noise and interference in the recording to make the speech signal clearer. The speech signal is filtered to remove unnecessary frequency components to further improve the clarity of the speech signal. The preprocessed speech data is further subjected to speech noise reduction using a discrete wavelet transform function. The speech data is converted into a time-frequency diagram, a wavelet basis is selected, a wavelet function is defined using the wavelet basis, and the wavelet function is infinitely scaled and shifted to obtain a set of wavelet functions. The low-frequency approximate coefficients and high-frequency detail coefficients are obtained by analysis. The discrete wavelet transform is also called wavelet division. Solution: For a given original signal, the scaling function is used to perform convolution operation with the original signal. The obtained value is called the approximate coefficient, also known as low-frequency information, which is used to describe the overall trend of the signal. The wavelet function is used to perform convolution operation with the original signal. The obtained value is called the detail coefficient, also known as high-frequency information, which is used to describe the detailed characteristics of the signal. The above is a layer of decomposition. In wavelet decomposition, after multiple decompositions, that is, the approximate coefficients of each layer are further decomposed. Speech denoising ensures the accuracy of the most original data. Discrete wavelet transform function denoising is based on the different properties of the wavelet coefficients of the original signal and noise at different scales. The corresponding rules are constructed to process the wavelet coefficients containing the signal in the wavelet domain. A wavelet basis defines two groups of functions: wavelet function and scaling function. Through discrete wavelet decomposition, threshold processing, and wavelet reconstruction, noise reduction is finally achieved. The pre-processed speech signal is framed and the long speech is divided into multiple short segments. The length of each frame is 10-30 milliseconds, and the frame shift is about 10 milliseconds. According to the physiological characteristics of the human ear, the sound waveform is standardized and digitized, and the Mel frequency cepstral coefficient features are extracted for subsequent speech recognition.

[0073] The expression formula for wavelet function scaling and displacement is:

[0074]

[0075] Among them, a is the scaling factor (scale factor), which controls the scale of the wavelet function, that is, the reflection frequency, and b is the translation factor, which controls the translation of the wavelet function on the time axis, that is, the reaction time. Represents the wavelet function, uses the trained speech recognition model to match and recognize the extracted features, and generates the corresponding text content. The speech recognition model is built based on the hidden Markov model. A large amount of annotated speech data is combined with the hidden Markov model to train the speech recognition model to recognize different speech units, thereby matching the audio signal and speech and identifying the candidate word sequence.

[0076] The text display system displays the conversation text and text operations after voice conversion, assists customer service specialists to view the conversation content in real time, improves the readability and manageability of the conversation content, and facilitates customer service specialists to quickly understand and respond to customer issues. The text display system receives the converted text data from the speech recognition system and further processes the recognized text, including correcting recognition errors, formatting text, and adding punctuation marks. It automatically corrects recognition errors in the text data to improve the accuracy of the text, further formats the corrected text, adjusts the font, font size, and color to meet the work interface requirements of the customer service software, and automatically adds appropriate punctuation marks according to the context to make the text more fluent and easy to read. The processed text is displayed in real time on the customer service software work interface of the customer service specialist, and provides conversation management functions such as conversation history and conversation status tracking to help customer service specialists understand the context of the conversation. By providing a complete conversation history function to record all communication content between customers and customer service specialists, it helps customer service specialists review the conversation process when necessary and understand the context of the conversation. By tracking the status of the conversation, it helps customer service specialists better manage their own workflows to ensure that no important information is missed. It automatically identifies and highlights customer speeches to help customer service specialists quickly distinguish between the two parties in the conversation and associate the current conversation text with the history so that customer service specialists can quickly understand the customer's historical problems and solutions.

[0077] AI-assisted system, which assists in problem consultation and locates problems to guide the solution of related problems, provides professional problem analysis and solutions, and improves the accuracy and efficiency of problem solving;

[0078] Furthermore, the AI-assisted system includes an AI big model module, a vector knowledge base module, a problem location module, and an intelligent guidance module, wherein the AI ​​big model module is used to analyze and match problems using the AI ​​big model, the vector knowledge base module is used to store common solutions, equipment information, and troubleshooting steps for hotel phone problems, and professionally supplement the knowledge content of the AI ​​big model, provide domain-specific knowledge support, enhance the domain expertise of the AI ​​big model, and improve the expertise and accuracy of problem handling, and the problem location module automatically locates the type of problem and the corresponding solution based on the user's problem and the vector knowledge base information, improves the accuracy of problem location, and reduces manual intervention and unnecessary door-to-door services;

[0079] Furthermore, the problem location and solution generation process specifically includes:

[0080] Based on the preprocessed customer speech text, the question text related to the hotel telephone problem is extracted from it, and the semantic model pre-trained by the problem location module is used to convert the question text into a semantic vector. The semantic vector can capture the semantic information of the text so that it can be compared with the information stored in the vector database. The vector similarity search technology is used to calculate the cosine similarity value between the semantic vector of the user's question and each vector in the vector database. According to the similarity value, the information most relevant to the user's question is retrieved and input into the AI ​​big model module. The vector database is a pre-built database containing common problems and their solutions, and each problem and its solution is converted into a semantic vector. The AI ​​big model (Trans The former model) understands the content of the problem, captures contextual information, conducts in-depth analysis of the hotel phone problems, and matches them with the information in the vector knowledge base. According to the retrieved vectors and corresponding solutions, the problem types are further matched. The problem types include equipment failure, service problems, and operation problems. According to the matched problem types, they are integrated with the analysis of the AI ​​big model to generate corresponding solutions. The solutions include specific operation steps, troubleshooting methods, and contacting maintenance personnel. The recommended solutions are optimized by syntax checking and semantic verification to ensure their accuracy, fluency, and readability. The optimized recommended solutions are output to the customer service specialists in text form so that they can provide assistance to hotel staff according to the solutions.

[0081] Furthermore, the calculation expression of the cosine similarity value is:

[0082]

[0083] Among them, CS is the cosine similarity value, A is the semantic vector of the user's question, B is the semantic vector of the question in the vector database, and A i is the value of vector A in the i-th dimension, representing the semantic weight of the user's question in this dimension, B i is the value of vector B in the i-th dimension, representing the semantic weight of the problem in the database in this dimension. n is the number of dimensions of the vector. The value range of cosine similarity is from -1 to 1, where -1 means completely opposite, 0 means no correlation, and 1 means completely identical. In the system, we look for vectors with cosine similarity values ​​close to 1, indicating that the user's problem is very similar to a problem in the database in semantics, thereby providing a more accurate solution. In the process of hotel telephone problem handling, we calculate the similarity between the user's problem vector and the common problem vectors and their solution vectors stored in the vector database, compare the similarity scores, quickly identify the information most relevant to the user's problem, and input it into the AI ​​big model module for further analysis and solution generation;

[0084] The intelligent guidance module, based on the problem location results, uses dynamically generated guidance words to cooperate with customer service specialists to guide hotel staff to solve the problem by themselves, improve users' ability to solve problems by themselves, and reduce labor costs. Based on the results of problem location, the intelligent guidance module further analyzes the specific context of the problem, such as the urgency of the problem, the skill level of hotel staff, etc., and dynamically generates customized guidance words for the recommended solution based on the situational analysis results and the optimized recommended solution output to the customer service specialist, including simple step-by-step instructions, troubleshooting suggestions, and operation guides, aiming to help hotel staff understand the problem and solve it by themselves. The guidance words include simple step-by-step instructions (such as "Please follow the steps below to restart the device"), troubleshooting suggestions (such as "Check the power supply The content of the introductory language will be appropriately adjusted according to the skill level of the hotel staff and the urgency of the problem to ensure that it is easy to understand and execute. The introductory language will be integrated with the contextual information of the conversation to make it closely related to the customer's problem, provide targeted guidance, and then optimize the introductory language. The optimized introductory language will be provided to the customer service specialist as a reference for communication with the hotel staff. The customer service specialist will guide the hotel staff to troubleshoot and solve the problem based on the content of the introductory language. In the process of communication between the customer service specialist and the customer, the customer service specialist will collect customer feedback, record whether the problem has been solved, and any new problems that may be encountered during the problem solving process;

[0085] The repair management system is used to generate repair service work orders for problems that hotel staff cannot solve by themselves, record repair information, and track repair work services.

[0086] The second embodiment is based on the first embodiment. Figure 3 As shown, the repair management system includes a work order generation and management module and a repair record and analysis module;

[0087] Among them, the work order generation and management module automatically generates work orders for problems that require on-site service by operation and maintenance engineers, tracks the progress of work order processing, optimizes the work order processing process, and improves operation and maintenance efficiency. When the customer service specialist finds that the hotel staff cannot solve the problem by themselves, he / she submits a repair request through the repair management system. The submitted content includes fault description, faulty equipment information, repair person information, and expected repair time. The repair management system reviews the submitted repair request. After the review is passed, it automatically generates a maintenance service work order. The repair management system automatically assigns the work order to the operation and maintenance engineer. After the work order is assigned, the system notifies the relevant operation and maintenance engineer through text messages and system notifications, and the operation and maintenance engineer performs on-site maintenance according to the work order information. Repair, solve the problems reported by hotel staff. During the processing, the engineer records the processing steps and progress and updates the work order status. The recorded processing steps include arriving at the site, diagnosing the fault, repairing or unable to complete, and repair completion steps. According to the updated work order status, if the operation and maintenance engineer determines that the maintenance task cannot be completed alone, the maintenance task status is analyzed, the task difficulty assessment coefficient is calculated, and the task difficulty is determined to determine whether to transfer the order or apply for assistance. The progress of the work order is tracked in real time. The customer service specialist tracks the work order status and keeps in touch with the hotel staff to update the progress. After the maintenance is completed, the operation and maintenance engineer updates the work order status to "completed", the hotel staff confirms that the problem has been solved, and the system records the confirmation information;

[0088] Furthermore, the task difficulty determination process includes:

[0089] The operation and maintenance engineer updates the work order status in the repair management system and marks it as "further support required". Based on the on-site conditions and maintenance needs, the engineer enters evaluation indicators that affect the difficulty of the task, including the estimated maintenance time, fault complexity, and expected workload. Among them, based on historical data and experience, the estimated maintenance time is divided into different intervals, and each interval is assigned a corresponding score. 0-30 minutes is set to 1 point, 30-60 minutes is set to 2 points, and so on. According to the type of fault, the number of systems or components involved, and the diagnostic difficulty factor, the fault complexity is divided into different levels, and each level is assigned a score. Simple faults are 1 point, medium-complexity faults are 2 points, and complex faults are 3 points. The number of personnel required in the maintenance process, the type and quantity of tools, and the spare parts demand factors are comprehensively analyzed, and the expected workload is divided into different levels, and each level is assigned a score. Low-workload tasks are 1 point, medium-workload tasks are 2 points, and high-workload tasks are 3 points. The system extracts data of similar tasks from historical maintenance records, including past maintenance time, resource consumption and success rate, analyzes maintenance difficulty and determines different task difficulty levels, namely, level 1 difficulty level, level 2 difficulty level, level 3 difficulty level and level 4 difficulty level. The difficulty level increases from level 1 to level 4. Combined with the evaluation indicator-related data input by the engineer, each evaluation indicator is quantitatively scored, and the task difficulty evaluation coefficient is calculated. Combined with the preset different task difficulty levels, the corresponding difficulty threshold is matched for each difficulty level. Based on the set difficulty level, the corresponding decision support is matched, and suggestions are provided to engineers. Engineers are advised to choose to transfer the order or apply for assistance. For tasks with difficulty levels 1 and 2, engineers are advised to apply for assistance. For tasks with difficulty levels 3 and 4, engineers are advised to transfer the order. If the order is transferred, the work order will be transferred to other professional operation and maintenance teams. If assistance is requested, reinforcements will be requested and other engineers will be called to jointly handle the maintenance task.

[0090] Furthermore, the calculation expression of the task difficulty evaluation coefficient is:

[0091]

[0092] Among them, D is the task difficulty assessment coefficient, and the value range of D is between 0 and 1. T is the expected maintenance time score, which reflects the time required to complete the task. C is the fault complexity score, which is a dimensionless score value indicating the complexity of the fault. W is the expected workload score, which is a dimensionless score value indicating the workload required to complete the task. a, b, and c are weight coefficients used to adjust the influence of each evaluation indicator on the task difficulty assessment coefficient. ω is a normalization factor used to ensure that the value of D is between 0 and 1. e a·T Used to amplify the impact of the estimated maintenance time. As T increases, e a·TIt grows rapidly, indicating a significant impact of time on task difficulty. ln(1+b·C) is used to smooth the impact of fault complexity. The logarithmic function grows slowly, indicating a gradual impact of complexity on task difficulty. (1+c·W) is used to indicate the direct contribution of expected workload to task difficulty. As W increases, task difficulty increases linearly. ω is used to ensure that the value of the task difficulty assessment coefficient D is between 0 and 1. When T, C or W increases, D also increases, indicating an increase in task difficulty. When T, C or W decreases, D decreases, indicating a decrease in task difficulty.

[0093] The plurality of difficulty levels correspond one-to-one to the plurality of difficulty thresholds, specifically:

[0094] The difficulty threshold for level 1 difficulty is 0 <D≤D 1 ;

[0095] The difficulty threshold of the secondary difficulty level is D 1 <D≤D 2 ;

[0096] The difficulty threshold of the third level difficulty level is D 2 <D≤D 3 ;

[0097] The difficulty threshold of the fourth level difficulty level is D 3 <D<1;

[0098] Among them, D is the task difficulty assessment coefficient, D 1 is the upper threshold of the first difficulty level and the lower threshold of the second difficulty level, D 2 is the upper threshold of the second level difficulty level and the lower threshold of the third level difficulty level, D 3 is the upper threshold of the third level of difficulty and the lower threshold of the fourth level of difficulty, D 1 =0.25, D 2 =0.5, D 3 =0.75;

[0099] The repair record and analysis module records and analyzes the detailed information of each repair request, including the repair time, problem description, and processing progress, to help the hotel management team track and analyze the maintenance work. The repair management system stores all repair records and processing progress information in the database to ensure the security and accessibility of the data, and analyzes the collected data to evaluate the work efficiency of the operation and maintenance engineers, including response time and repair time. According to the analysis results, a maintenance work report is generated, including the maintenance completion rate, average response time, and maintenance success rate, and improvement suggestions are put forward. Based on the maintenance work report, fault data is analyzed to identify recurring failure modes, provide a basis for preventive maintenance, and put forward suggestions for improving the maintenance process and improving maintenance efficiency. Historical data is used to predict potential failures, and maintenance work is scheduled in advance to reduce unexpected downtime. Feedback from hotel personnel on maintenance services is collected and integrated into the analysis report for continuous improvement of service quality.

[0100] Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field and related fields without creative work should fall within the scope of protection of the present invention. The structures, devices and operating methods not specifically described and explained in the present invention are implemented according to the conventional means in the field unless otherwise specified and limited.

Claims

1. A hotel telephone problem reporting intelligent customer service assistant processing system, characterized in that: The customer service assistant processing system includes a speech-to-text system, a text display system, an AI-assisted system, and a repair management system, wherein the electrical signals between the systems are connected; The speech-to-text system collects speech data and uses a speech processing algorithm to pre-process the speech data and convert the speech data into text; The text display system displays the conversation text and text operations after voice conversion, assisting customer service specialists to view the conversation content in real time; The AI-assisted system assists in problem consultation, locates problems to guide the solution of related problems, and provides professional problem analysis and solutions; The repair management system is used to generate repair service work orders for problems that hotel staff cannot solve by themselves, record repair information, and track repair work services.

2. According to claim 1, a hotel telephone problem reporting intelligent customer service assistant processing system is characterized by: In the speech-to-text system, the process of converting speech data into text includes: The voice signal is collected through a microphone, converted into a digital signal through an analog-to-digital converter, and the collected voice signal is preprocessed, wherein the preprocessing process includes noise reduction and filtering operations; The pre-processed speech data is further subjected to speech noise reduction using a discrete wavelet transform function. The speech data is converted into a time-frequency diagram, a wavelet basis is selected, a wavelet function is defined using the wavelet basis, and the wavelet function is scaled and shifted to obtain the low-frequency approximate coefficients and the high-frequency detail coefficients. The pre-processed speech signal is framed to divide the long speech into multiple short segments. According to the physiological characteristics of the human ear, the sound waveform is standardized and digitized to extract the Mel frequency cepstral coefficient features. The extracted features are matched and recognized using a trained speech recognition model to generate corresponding text content, wherein the speech recognition model is constructed based on a hidden Markov model.

3. The hotel telephone problem reporting intelligent customer service assistant processing system according to claim 2 is characterized by: In the text display system, the display process of the dialogue text and the text operation includes: The text presentation system receives the converted text data from the speech recognition system and further processes the recognized text, including correcting recognition errors, formatting the text, and adding punctuation marks; The processed text is displayed in real time on the customer service software work interface of the customer service specialist, and a conversation management function is provided to record the conversation history and track the conversation status, so as to help the customer service specialist understand the context of the conversation; Automatically identify and highlight customer comments, helping customer service specialists quickly distinguish between the two parties in a conversation and associate the current conversation text with historical records.

4. The hotel telephone problem reporting intelligent customer service assistant processing system according to claim 3 is characterized by: The AI-assisted system includes an AI large model module, a vector knowledge base module, a problem location module, and an intelligent guidance module; The AI ​​big model module is used to analyze and match problems using the AI ​​big model; The vector knowledge base module is used to store common solutions, equipment information, troubleshooting steps for hotel phone problems, and to professionally supplement the knowledge content of the AI ​​big model; The problem location module automatically locates the type of problem and the corresponding solution based on the user's problem and the vector knowledge base information; The intelligent guidance module cooperates with the customer service specialist to guide the hotel staff to solve the problem by themselves according to the problem location result and through dynamically generated guidance words.

5. The hotel telephone problem reporting intelligent customer service assistant processing system according to claim 4 is characterized by: The problem location and solution generation process specifically includes: Based on the pre-processed customer speech text, the question text related to the hotel telephone question is extracted from it, and the question text is converted into a semantic vector using the semantic model pre-trained by the question location module; Using vector similarity search technology, the cosine similarity value between the semantic vector of the user's question and each vector in the vector database is calculated. Based on the similarity value, the most relevant information to the user's question is retrieved and input into the AI ​​big model module. Use AI big models to understand the content of the question, capture contextual information, conduct in-depth analysis of hotel phone questions, and match them with the information in the vector knowledge base. Based on the retrieved vectors and corresponding solutions, further match the problem types, including equipment failure, service problems, and operation problems. Generate corresponding solutions based on the matched problem type and the analysis of the AI ​​big model. The solutions include specific operation steps, troubleshooting methods, and contacting maintenance personnel. The proposed solution is optimized through grammar checking and semantic verification, and the optimized proposed solution is output to the customer service specialist in text form.

6. The hotel telephone problem reporting intelligent customer service assistant processing system according to claim 5 is characterized by: The calculation expression of the cosine similarity value is: Among them, CS is the cosine similarity value, A is the semantic vector of the user's question, B is the semantic vector of the question in the vector database, and A i is the value of vector A in the i-th dimension, B i is the value of vector B in the i-th dimension, and n is the number of dimensions of the vector.

7. The hotel telephone problem reporting intelligent customer service assistant processing system according to claim 6 is characterized by: In the intelligent guidance module, the process of generating the guide words includes: Based on the results of problem location, the intelligent guidance module further analyzes the specific context of the problem and dynamically generates customized guidance for the proposed solution based on the context analysis results and the optimized proposed solution output to the customer service specialist, including simple step-by-step instructions, troubleshooting suggestions, and operation guides. Integrate the introductory words with the contextual information of the conversation, so that the introductory words are closely related to the customer's questions, and then optimize the introductory words. Provide the optimized introductory words to the customer service specialists as a reference for communicating with hotel staff; The customer service specialist guides hotel staff to troubleshoot and resolve issues based on the content of the introductory statement. During the communication process between the customer service specialist and the customer, the specialist collects customer feedback and records whether the problem has been resolved.

8. The hotel telephone problem reporting intelligent customer service assistant processing system according to claim 7, characterized in that: The repair management system includes a work order generation and management module and a repair record and analysis module. The work order generation and management module automatically generates a work order for problems that require on-site service by an operation and maintenance engineer, and tracks the processing progress of the work order; The repair record and analysis module records and analyzes the detailed information of each repair request, including the repair time, problem description, and processing progress; Among them, in the work order generation and management module, the process of tracking the generation and processing progress of the work order includes: When the customer service specialist finds that the hotel staff cannot solve the problem by themselves, he / she shall submit a repair request through the repair management system. The submitted content shall include the fault description, faulty equipment information, the information of the person reporting the repair, and the expected repair time. The repair management system reviews the submitted repair request and automatically generates a repair service work order after passing the review; The repair management system automatically assigns the work order to the operation and maintenance engineer, and after the work order is assigned, the system notifies the relevant operation and maintenance engineer through SMS and system notification; The operation and maintenance engineer conducts on-site maintenance according to the work order information and solves the problems reported by the hotel staff. During the process, the engineer records the processing steps and progress and updates the work order status. The recorded processing steps include arriving at the site, diagnosing the fault, repairing or unable to complete, and repairing completed steps; Based on the updated work order status, if the operation and maintenance engineer determines that the maintenance task cannot be completed alone, the maintenance task status is analyzed, the task difficulty assessment coefficient is calculated, and the task difficulty is determined to determine whether to transfer the order or request assistance; The progress of work order processing is tracked in real time. Customer service specialists track the status of work orders and keep in touch with hotel staff to update the progress. After the maintenance is completed, the operation and maintenance engineer updates the work order status to "Completed". The hotel staff confirms that the problem has been resolved and the system records the confirmation information.

9. The hotel telephone problem reporting intelligent customer service assistant processing system according to claim 8, characterized in that: The determination process of the task difficulty includes: The operation and maintenance engineer updates the work order status in the repair management system and marks it as "needs further support". Based on the on-site situation and maintenance requirements, the engineer enters the evaluation indicators that affect the difficulty of the task, including the estimated maintenance time, fault complexity, and expected workload. The estimated maintenance time is divided into different intervals, and each interval is assigned a corresponding score. 0-30 minutes is set as 1 point, 30-60 minutes is set as 2 points, and so on. The fault complexity is divided into different levels, and each level is assigned a score. Simple faults are 1 point, medium-complexity faults are 2 points, and complex faults are 3 points. The expected workload is divided into different levels, and each level is assigned a score. Low-workload tasks are 1 point, medium-complexity tasks are 2 points, and high-workload tasks are 3 points. The system extracts data of similar tasks from historical maintenance records, including past maintenance time, resource consumption and success rate, analyzes the maintenance difficulty and determines different task difficulty levels, namely, level 1 difficulty level, level 2 difficulty level, level 3 difficulty level and level 4 difficulty level. The difficulty level increases from level 1 to level 4. Combined with the evaluation indicator-related data input by the engineer, each evaluation indicator is quantitatively scored, the task difficulty evaluation coefficient is calculated, and combined with the preset different task difficulty levels, the corresponding difficulty threshold is matched for each difficulty level; Based on the set difficulty level, the corresponding decision support is matched and suggestions are provided to engineers. Engineers are advised to choose to transfer the order or apply for assistance. For tasks with difficulty levels 1 and 2, engineers are advised to apply for assistance. For tasks with difficulty levels 3 and 4, engineers are advised to transfer the order. If they choose to transfer the order, the work order will be transferred to other professional operation and maintenance teams. If they apply for assistance, they will request reinforcements and call on other engineers to jointly handle the maintenance task.

10. The hotel telephone problem reporting intelligent customer service assistant processing system according to claim 9, characterized in that: The calculation expression of the task difficulty evaluation coefficient is: Among them, D is the task difficulty assessment coefficient, the value range of D is between 0 and 1, T is the expected maintenance time score, C is the fault complexity score, W is the expected workload score, a, b, c are weight coefficients used to adjust the influence of each evaluation indicator on the task difficulty assessment coefficient, ω is a normalization factor used to ensure that the value of D is between 0 and 1; The plurality of difficulty levels correspond one-to-one to the plurality of difficulty thresholds, specifically: The difficulty threshold for level 1 difficulty is 0 <D≤D1; The difficulty threshold of the secondary difficulty level is D1 <D≤D2; The difficulty threshold of the third level difficulty level is D2 <D≤D3; The difficulty threshold of the fourth level difficulty level is D3 <D<1; Among them, D is the task difficulty assessment coefficient, D1 is the upper threshold of the first difficulty level and the lower threshold of the second difficulty level, D2 is the upper threshold of the second difficulty level and the lower threshold of the third difficulty level, D3 is the upper threshold of the third difficulty level and the lower threshold of the fourth difficulty level, D1=0.25, D2=0.5, D3=0.75.

Images