Electric power customer service work order abstract generation method and system based on voice conversation

Abstract

The invention discloses an electric power customer service work order abstract generation method based on voice dialogue, and relates to the technical field of voice recognition, and the method comprises the steps: receiving voice dialogue content; key entities related to electric power are extracted; performing deep understanding on the voice text based on the semantic analysis model; based on the extracted key information, automatically generating an abstract of the electric power customer service work order; submitting the automatically generated work order abstract for manual auditing; the invention discloses an electric power customer service work order abstract generation system based on voice conversation. The system comprises a voice receiving module, a voice text processing module, a keyword extraction module, a semantic analysis module, a work order generation module, a work order auditing module and a work order publishing module. Through advanced voice recognition, natural language processing, semantic analysis and emotion recognition technologies, automation of the work order generation process can be achieved, the work order processing quality and efficiency can be optimized, and therefore remarkable benefits and improvement are brought to electric customer service work.

Description

Technical Field

[0001] The present invention relates to the technical field of speech recognition, and in particular to a method and system for generating a summary of an electric customer service work order based on speech dialogue. Background Art

[0002] With the continuous development of speech recognition technology and natural language processing technology, more and more fields have begun to apply speech recognition technology to improve work efficiency, especially in the customer service industry. As a typical service industry, the power customer service field faces a large number of customer consultations and repair requests. Traditional manual processing methods are often inefficient and prone to errors. Therefore, the automated work order generation method based on voice dialogue has become a key technology to improve the work efficiency and service quality of power customer service.

[0003] Existing work order generation methods usually rely on manual operations, which have problems such as long time, low efficiency, and inconsistent processing results. For the generation of power customer service work orders, it is necessary to accurately capture the questions, needs and related information raised by customers, and to be able to classify and prioritize work orders based on multi-dimensional factors such as customer emotions and the urgency of the problem. In addition, the process of manual review and processing of work orders also needs to be optimized and automated on the basis of quality assurance. Therefore, the power customer service work order summary generation method based on voice dialogue can effectively improve the automation level of work order generation by converting speech to text, using natural language processing technology for semantic analysis, key entity recognition and context understanding, and further improve the quality and accuracy of work orders through manual review mechanism. This technology can not only reduce the manual burden, but also improve service response speed and customer satisfaction, and optimize power customer service management. Summary of the invention

[0004] In order to solve the above technical problems, a method and system for generating a summary of an electric power customer service work order based on voice dialogue is provided. This technical solution solves the above problems.

[0005] In order to achieve the above purpose, the technical solution adopted by the present invention is:

[0006] A method for generating a summary of a power customer service work order based on voice dialogue, comprising:

[0007] Receive the voice conversation content between the power customer and the customer service staff based on the speech recognition technology, convert it into text data, and pre-process the received voice text;

[0008] Extract key electricity-related entities from the processed text based on natural language processing technology, including customer name, electricity address, fault type, and repair problem;

[0009] Based on the semantic analysis model, the voice text is deeply understood, and the context information of the conversation is combined to obtain the priority and main issues of customer needs;

[0010] Based on the extracted key information, the summary of the power customer service work order is automatically generated, and preliminary classification is performed based on the work order processing flow;

[0011] The automatically generated work order summary is submitted for manual review, and the reviewed work order summary is published to the power service system.

[0012] Preferably, the voice conversation content between the power customer and the customer service staff is received based on the voice recognition technology and converted into text data, and the pre-processing of the received voice text specifically includes:

[0013] Use speech recognition technology to convert voice conversations between electricity customers and customer service agents into text;

[0014] After obtaining the original text, perform text cleaning operations, remove noise words through regular expressions, fix speech recognition errors, use dictionaries, context relevance and rules for correction, and identify and remove repeated parts;

[0015] Divide the text into complete sentences according to punctuation marks, segment each sentence into words, divide the text into semantic units, and mark the parts of speech;

[0016] Remove stop words, make all text lowercase, remove special symbols from text, check and correct spelling errors.

[0017] Preferably, the extraction of key electricity-related entities from the processed text based on natural language processing technology, including customer name, electricity address, fault type and repair problem, specifically includes:

[0018] Extract features from the preprocessed audio signal, where the feature extraction formula is:

[0019]

[0020] In the formula, MFCC k is the kth Mel frequency cepstrum coefficient, X K (n) represents the frequency component of the signal after Fourier transformation, N is the total number of sample points of frequency analysis, K is the total number of MFCC coefficients, n is the index variable, and k is the index of the cepstral coefficient;

[0021] The extracted features are input into the speech recognition model to predict the phoneme words, where the prediction formula is:

[0022] P t =Softmax(f(x t,θ))

[0023] Where P t is the probability distribution of the model output at time step t, Softmax is the activation function, f(x t ,θ) is the model’s forward function f, which accepts the input feature x t and model parameters θ as input and output an unnormalized score, x t represents the input feature vector at time step t, which represents the parameters of the model;

[0024] Use the classification model to determine which repair problem category "power outage" belongs to, and output the result. The classification model formula is:

[0025]

[0026] In the formula, P(C i |x) means that given the input x, category C i The conditional probability of For a specific category C i , input x and category C i The exponential function of the score between Indicates that for all categories C, i Sum.

[0027] Preferably, the in-depth understanding of the speech text based on the semantic analysis model, combined with the context information of the conversation, to obtain the priority and main issues of customer needs specifically includes:

[0028] Analyze the customer's emotional tendencies to determine whether the customer is satisfied, angry, confused, or seeking help, provide clues for prioritization, and identify the customer's core intent through a semantic understanding model;

[0029] Use knowledge graph technology to combine entity and topic information in the conversation to obtain customer needs;

[0030] Identify the main issues raised by customers and prioritize them based on the results of the classification model;

[0031] Assess the priority of issues based on customer sentiment analysis results, issue severity and urgency factors;

[0032] Generate personalized automatic responses based on customer needs, and assign issues to corresponding customer service personnel and systems for processing based on priority and issue type.

[0033] Preferably, the automatic generation of a summary of the power customer service work order based on the extracted key information and the preliminary classification in combination with the work order processing flow specifically include:

[0034] Use natural language generation technology to generate a work order summary based on the extracted key information, including problem description, customer contact information, problem type, and processing requirements;

[0035] Obtain the work order classification category, assign priorities to the work orders, and automatically assign the work orders to the corresponding staff based on the work order type and priority;

[0036] The system updates the status of the work order based on the progress of the work order, sends notifications of the work order processing progress to the customer through automated tools, provides real-time feedback, and automatically records and updates the work order summary if there are any changes during the manual processing;

[0037] The processing personnel take corresponding measures to solve the problem according to the content of the work order. After the solution is implemented, they confirm with the customer whether the problem has been properly handled and collect customer feedback on the service through questionnaires.

[0038] Preferably, submitting the automatically generated work order summary to manual review and publishing the reviewed work order summary to the power service system specifically includes:

[0039] Submit the automatically generated work order summary to the manual reviewer for review and confirmation. The manual review process includes:

[0040] Auditors check key information in the work order summary, including customer information, problem type, and equipment information;

[0041] Check the format, including titles, content structure and timestamps;

[0042] Correct the errors and check the priority of the issues. The auditor will give feedback on the processing results of the work order summary. If the review is passed, it will proceed to the next step. If there are problems with the work order summary, the auditor will mark it as pending correction and provide specific modification suggestions;

[0043] After approval, the work order summary is published to the work order management module of the power service system. The system assigns the work order to the corresponding processing team based on its classification and priority.

[0044] The system records the log of manual review, including the name of the reviewer, review time and review opinion.

[0045] A power customer service work order summary generation system based on voice dialogue, comprising:

[0046] Voice receiving module: used to receive the voice conversation between customers and customer service staff and perform voice recognition;

[0047] Speech and text processing module: The speech and text processing module is electrically connected to the speech receiving module, and the speech and text processing module is used to pre-process the recognized text;

[0048] Keyword extraction module: The keyword extraction module is electrically connected to the speech text processing module, and the keyword extraction module is used to extract key information and power-related entities from the processed text;

[0049] Semantic analysis module: The semantic analysis module is electrically connected to the keyword extraction module, and the semantic analysis module is used to perform semantic understanding in combination with the context and analyze customer needs and problems;

[0050] Work order generation module: The work order generation module is electrically connected to the semantic analysis module, and the work order generation module is used to automatically generate a summary of the power customer service work order and perform preliminary classification;

[0051] Work order review module: The work order review module is electrically connected to the work order generation module, and the work order review module is used to manually review and optimize the generated work order summary;

[0052] Work order publishing module: The work order publishing module is electrically connected to the work order review module, and the work order publishing module is used to publish the reviewed work order to the power service system for subsequent processing.

[0053] Preferably, the speech-to-text processing module specifically includes:

[0054] Text cleaning unit: cleans the recognized text to remove useless characters, spaces, special symbols and repeated words;

[0055] Spelling Correction Unit: Automatically check and correct spelling errors in recognized texts;

[0056] Word segmentation and part-of-speech tagging unit: segment the text and mark the part of speech for each word;

[0057] Entity recognition unit: recognizes entity information in text, including names of people, places, times, dates, and organization names;

[0058] Stop word removal unit: removes stop words from the text.

[0059] Preferably, the keyword extraction module specifically includes:

[0060] Power entity recognition unit: Identify and extract power-related entities in text, extract power-related proper nouns, and identify the relationship between entities based on context;

[0061] Power business related term extraction unit: Identify and extract professional terms, terminology and abbreviations in the power industry;

[0062] Word vector analysis unit: Use deep learning technology to analyze word vectors and extract key information from texts;

[0063] Topic modeling and clustering: cluster similar keywords through cluster analysis method to extract power-related information with similar background;

[0064] Keyword filtering and priority sorting unit: Filter and sort the extracted keywords and give the keywords according to the priority.

[0065] Preferably, the semantic analysis module specifically includes:

[0066] Context analysis unit: Analyze the contextual relationship between words and sentences in the text, and identify the different semantics of the same word in different contexts;

[0067] Sentiment analysis unit: Analyzes the sentiment tendency in the text and identifies the emotional state of the customer;

[0068] Intent Identification Unit: Identify customer intent based on context and keywords.

[0069] Compared with the prior art, the present invention has the following beneficial effects:

[0070] The present invention proposes a method for generating a summary of a power customer service work order based on voice dialogue. Through advanced speech recognition, natural language processing, semantic analysis and emotion recognition technology, it can not only realize the automation of the work order generation process, but also optimize the quality and efficiency of work order processing, thereby bringing significant benefits and improvements to power customer service work. BRIEF DESCRIPTION OF THE DRAWINGS

[0071] Figure 1 It is a step flow framework diagram of the present invention;

[0072] Figure 2 It is a system framework diagram of the present invention. DETAILED DESCRIPTION

[0073] The following description is used to disclose the present invention so that those skilled in the art can implement the present invention. The preferred embodiments described below are only examples, and those skilled in the art may think of other obvious variations.

[0074] Reference Figure 1 As shown, a method for generating a summary of a power customer service work order based on voice dialogue includes:

[0075] Step 1:

[0076] Use speech recognition technology to convert voice conversations between electricity customers and customer service agents into text;

[0077] After obtaining the original text, perform text cleaning operations, remove noise words through regular expressions, fix speech recognition errors, use dictionaries, context relevance and rules for correction, and identify and remove repeated parts;

[0078] Divide the text into complete sentences according to punctuation marks, segment each sentence into words, divide the text into semantic units, and mark the parts of speech;

[0079] Remove stop words, make all text lowercase, remove special symbols from text, check and correct spelling errors.

[0080] Step 2:

[0081] Extract features from the preprocessed audio signal, where the feature extraction formula is:

[0082]

[0083] In the formula, MFCC k is the kth Mel frequency cepstrum coefficient, X K (n) represents the frequency component of the signal after Fourier transformation, N is the total number of sample points of frequency analysis, K is the total number of MFCC coefficients, n is the index variable, and k is the index of the cepstral coefficient;

[0084] The extracted features are input into the speech recognition model to predict the phoneme words, where the prediction formula is:

[0085] P t =Softmax(f(x t ,θ))

[0086] Where P t is the probability distribution of the model output at time step t, Softmax is the activation function, f(x t ,θ) is the model’s forward function f, which accepts the input feature x t and model parameters θ as input and output an unnormalized score, x t represents the input feature vector at time step t, which represents the parameters of the model;

[0087] Use the classification model to determine which repair problem category "power outage" belongs to, and output the result. The classification model formula is:

[0088]

[0089] In the formula, P(C i |x) means that given the input x, category C i The conditional probability of For a specific category C i , input x and category C i The exponential function of the score between Indicates that for all categories C,i Sum.

[0090] Step 3:

[0091] Analyze the customer's emotional tendencies to determine whether the customer is satisfied, angry, confused, or seeking help, provide clues for prioritization, and identify the customer's core intent through a semantic understanding model;

[0092] Use knowledge graph technology to combine entity and topic information in the conversation to obtain customer needs;

[0093] Identify the main issues raised by customers and prioritize them based on the results of the classification model;

[0094] Assess the priority of issues based on customer sentiment analysis results, issue severity and urgency factors;

[0095] Generate personalized automatic responses based on customer needs, and assign issues to corresponding customer service personnel and systems for processing based on priority and issue type.

[0096] Step 4:

[0097] Use natural language generation technology to generate a work order summary based on the extracted key information, including problem description, customer contact information, problem type, and processing requirements;

[0098] Obtain the work order classification category, assign priorities to the work orders, and automatically assign the work orders to the corresponding staff based on the work order type and priority;

[0099] The system updates the status of the work order based on the progress of the work order, sends notifications of the work order processing progress to the customer through automated tools, provides real-time feedback, and automatically records and updates the work order summary if there are any changes during the manual processing;

[0100] The processing personnel take corresponding measures to solve the problem according to the content of the work order. After the solution is implemented, they confirm with the customer whether the problem has been properly handled and collect customer feedback on the service through questionnaires.

[0101] Step 5:

[0102] Submit the automatically generated work order summary to the manual reviewer for review and confirmation. The manual review process includes:

[0103] Auditors check key information in the work order summary, including customer information, problem type, and equipment information;

[0104] Check the format, including titles, content structure and timestamps;

[0105] Correct the errors and check the priority of the issues. The auditor will give feedback on the processing results of the work order summary. If the review is passed, it will proceed to the next step. If there are problems with the work order summary, the auditor will mark it as pending correction and provide specific modification suggestions;

[0106] After approval, the work order summary is published to the work order management module of the power service system. The system assigns the work order to the corresponding processing team based on its classification and priority.

[0107] The system records the log of manual review, including the name of the reviewer, review time and review opinion.

[0108] A power customer service work order summary generation system based on voice dialogue, comprising:

[0109] Voice receiving module: used to receive the voice conversation between customers and customer service staff and perform voice recognition;

[0110] Speech and text processing module: The speech and text processing module is electrically connected to the speech receiving module, and the speech and text processing module is used to pre-process the recognized text;

[0111] Keyword extraction module: The keyword extraction module is electrically connected to the speech text processing module, and the keyword extraction module is used to extract key information and power-related entities from the processed text;

[0112] Semantic analysis module: The semantic analysis module is electrically connected to the keyword extraction module, and the semantic analysis module is used to perform semantic understanding in combination with the context and analyze customer needs and problems;

[0113] Work order generation module: The work order generation module is electrically connected to the semantic analysis module, and the work order generation module is used to automatically generate a summary of the power customer service work order and perform preliminary classification;

[0114] Work order review module: The work order review module is electrically connected to the work order generation module, and the work order review module is used to manually review and optimize the generated work order summary;

[0115] Work order publishing module: The work order publishing module is electrically connected to the work order review module, and the work order publishing module is used to publish the reviewed work order to the power service system for subsequent processing.

[0116] Preferably, the speech-to-text processing module specifically includes:

[0117] Text cleaning unit: cleans the recognized text to remove useless characters, spaces, special symbols and repeated words;

[0118] Spelling Correction Unit: Automatically check and correct spelling errors in recognized texts;

[0119] Word segmentation and part-of-speech tagging unit: segment the text and mark the part of speech for each word;

[0120] Entity recognition unit: recognizes entity information in text, including names of people, places, times, dates, and organization names;

[0121] Stop word removal unit: removes stop words from the text.

[0122] Preferably, the keyword extraction module specifically includes:

[0123] Power entity recognition unit: Identify and extract power-related entities in text, extract power-related proper nouns, and identify the relationship between entities based on context;

[0124] Power business related term extraction unit: Identify and extract professional terms, terminology and abbreviations in the power industry;

[0125] Word vector analysis unit: Use deep learning technology to analyze word vectors and extract key information from texts;

[0126] Topic modeling and clustering: cluster similar keywords through cluster analysis method to extract power-related information with similar background;

[0127] Keyword filtering and priority sorting unit: Filter and sort the extracted keywords and give the keywords according to the priority.

[0128] Preferably, the semantic analysis module specifically includes:

[0129] Context analysis unit: Analyze the contextual relationship between words and sentences in the text, and identify the different semantics of the same word in different contexts;

[0130] Sentiment analysis unit: Analyzes the sentiment tendency in the text and identifies the emotional state of the customer;

[0131] Intent Identification Unit: Identify customer intent based on context and keywords.

[0132] In summary, the advantages of the present invention are:

[0133] Using speech recognition technology to automatically convert conversations between customers and customer service staff into text greatly reduces the time and workload of manual recording. Through automated text processing and keyword extraction, key entity information related to electricity can be quickly extracted to improve the efficiency of work order generation;

[0134] Based on natural language processing technology, especially through the deep understanding of semantic analysis models, the system can accurately identify customer needs, question types and key information from voice conversation texts to ensure the accuracy of work order content;

[0135] After the work order summary is generated, the system will automatically classify and prioritize the work order content, and conduct a comprehensive assessment based on the customer sentiment analysis results and the urgency of the problem. This automated prioritization and problem classification helps improve the timeliness and accuracy of problem handling.

[0136] Combined with customer needs and sentiment analysis, the system can automatically generate personalized responses based on customer intent and question type, and assign them to the appropriate handler based on the priority of the work order, thus ensuring that customer issues are responded to in a timely manner and reducing processing delays;

[0137] The system automatically generates work order summaries and submits them for manual review, reducing the burden of manual work order processing. Manual reviewers only need to check and correct key information and formats to ensure the quality and accuracy of the work order, thereby improving overall work efficiency.

[0138] The above shows and describes the basic principles, main features and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The above embodiments and descriptions only describe the principles of the present invention. The present invention may be subject to various changes and improvements without departing from the spirit and scope of the present invention. These changes and improvements fall within the scope of the present invention. The scope of protection claimed by the present invention is defined by the attached claims and their equivalents.

Claims

1. A method for generating a summary of a power customer service work order based on voice dialogue, characterized in that: include: Receive the voice conversation content between the power customer and the customer service staff based on the speech recognition technology, convert it into text data, and pre-process the received voice text; Extract key electricity-related entities from the processed text based on natural language processing technology, including customer name, electricity address, fault type, and repair problem; Based on the semantic analysis model, the voice text is deeply understood, and the priority and main issues of customer needs are obtained by combining the context information of the conversation; Based on the extracted key information, the summary of the power customer service work order is automatically generated, and preliminary classification is performed based on the work order processing flow; The automatically generated work order summary is submitted for manual review, and the reviewed work order summary is published to the power service system.

2. According to the method for generating a summary of a power customer service work order based on voice dialogue according to claim 1, it is characterized in that: The method of receiving the voice conversation content between the power customer and the customer service staff based on the voice recognition technology and converting it into text data, and preprocessing the received voice text specifically includes: Use speech recognition technology to convert voice conversations between electricity customers and customer service agents into text; After obtaining the original text, perform text cleaning operations, remove noise words through regular expressions, fix speech recognition errors, use dictionaries, context relevance and rules for correction, and identify and remove repeated parts; Divide the text into complete sentences according to punctuation marks, segment each sentence into words, divide the text into semantic units, and mark the parts of speech; Remove stop words, make all text lowercase, remove special symbols from text, check and correct spelling errors.

3. The method for generating a summary of a power customer service work order based on voice dialogue according to claim 2, characterized in that: The extraction of key electricity-related entities from the processed text based on natural language processing technology, including customer name, electricity address, fault type and repair problem, specifically includes: Extract features from the preprocessed audio signal, where the feature extraction formula is: In the formula, MFCC k is the kth Mel frequency cepstrum coefficient, X K (n) represents the frequency component of the signal after Fourier transformation, N is the total number of sample points of frequency analysis, K is the total number of MFCC coefficients, n is the index variable, and k is the index of the cepstral coefficient; The extracted features are input into the speech recognition model to predict the phoneme words, where the prediction formula is: P t =Softmax(f(x t ,θ)) Where P t is the probability distribution of the model output at time step t, Softmax is the activation function, f(x t ,θ) is the model’s forward function f, which accepts the input feature x t and model parameters θ as input and output an unnormalized score, x t represents the input feature vector at time step t, which represents the parameters of the model; Use the classification model to determine which repair problem category "power outage" belongs to, and output the result. The classification model formula is: In the formula, P(C i |x) means that given the input x, category C i The conditional probability of For a specific category C i , input x and category C i The exponential function of the score between Indicates that for all categories C, i Sum.

4. The method for generating a summary of a power customer service work order based on voice dialogue according to claim 3, characterized in that: The semantic analysis model is used to deeply understand the voice text and combine the context information of the conversation to obtain the priority and main issues of customer needs, including: Analyze the customer's emotional tendencies to determine whether the customer is satisfied, angry, confused, or seeking help, provide clues for prioritization, and identify the customer's core intent through a semantic understanding model; Use knowledge graph technology to combine entity and topic information in the conversation to obtain customer needs; Identify the main issues raised by customers and prioritize them based on the results of the classification model; Assess the priority of issues based on customer sentiment analysis results, issue severity and urgency factors; Generate personalized automatic responses based on customer needs, and assign issues to corresponding customer service personnel and systems for processing based on priority and issue type.

5. The method for generating a summary of a power customer service work order based on voice dialogue according to claim 4, characterized in that: Based on the extracted key information, the summary of the power customer service work order is automatically generated, and the preliminary classification is carried out in combination with the work order processing flow, specifically including: Use natural language generation technology to generate a work order summary based on the extracted key information, including problem description, customer contact information, problem type, and processing requirements; Obtain the work order classification category, assign priorities to the work orders, and automatically assign the work orders to the corresponding staff based on the work order type and priority; The system updates the status of the work order based on the progress of the work order, sends notifications of the work order processing progress to the customer through automated tools, provides real-time feedback, and automatically records and updates the work order summary if there are any changes during the manual processing; The processing personnel take corresponding measures to solve the problem according to the content of the work order. After the solution is implemented, they confirm with the customer whether the problem has been properly handled and collect customer feedback on the service through questionnaires.

6. The method for generating a summary of a power customer service work order based on voice dialogue according to claim 4, characterized in that: Submitting the automatically generated work order summary to manual review and publishing the reviewed work order summary to the power service system specifically includes: Submit the automatically generated work order summary to the human reviewer for review and confirmation; After approval, the work order summary is published to the work order management module of the power service system. The system assigns the work order to the corresponding processing team based on its classification and priority. The system records the log of manual review, including the name of the reviewer, review time and review opinion.

7. A power customer service work order summary generation system based on voice dialogue, characterized in that: include: Voice receiving module: used to receive the voice conversation between customers and customer service staff and perform voice recognition; Speech and text processing module: The speech and text processing module is electrically connected to the speech receiving module, and the speech and text processing module is used to pre-process the recognized text; Keyword extraction module: The keyword extraction module is electrically connected to the speech text processing module, and the keyword extraction module is used to extract key information and power-related entities from the processed text; Semantic analysis module: The semantic analysis module is electrically connected to the keyword extraction module, and the semantic analysis module is used to perform semantic understanding in combination with the context and analyze customer needs and problems; Work order generation module: The work order generation module is electrically connected to the semantic analysis module, and the work order generation module is used to automatically generate a summary of the power customer service work order and perform preliminary classification; Work order review module: The work order review module is electrically connected to the work order generation module, and the work order review module is used to manually review and optimize the generated work order summary; Work order publishing module: The work order publishing module is electrically connected to the work order review module, and the work order publishing module is used to publish the reviewed work order to the power service system for subsequent processing.

8. The power customer service work order summary generation system based on voice dialogue according to claim 7 is characterized in that: The speech text processing module specifically includes: Text cleaning unit: cleans the recognized text to remove useless characters, spaces, special symbols and repeated words; Spelling Correction Unit: Automatically check and correct spelling errors in recognized texts; Word segmentation and part-of-speech tagging unit: segment the text and mark the part of speech for each word; Entity recognition unit: recognizes entity information in text, including names of people, places, times, dates, and organization names; Stop word removal unit: removes stop words from the text.

9. The power customer service work order summary generation system based on voice dialogue according to claim 8 is characterized in that: The keyword extraction module specifically includes: Power entity recognition unit: Identify and extract power-related entities in text, extract power-related proper nouns, and identify the relationship between entities based on context; Power business related term extraction unit: Identify and extract professional terms, terminology and abbreviations in the power industry; Word vector analysis unit: Use deep learning technology to analyze word vectors and extract key information from texts; Topic modeling and clustering: cluster similar keywords through cluster analysis method to extract power-related information with similar background; Keyword filtering and priority sorting unit: Filter and sort the extracted keywords and give the keywords according to the priority.

10. The power customer service work order summary generation system based on voice dialogue according to claim 9, characterized in that: The semantic analysis module specifically includes: Context analysis unit: Analyze the contextual relationship between words and sentences in the text, and identify the different semantics of the same word in different contexts; Sentiment analysis unit: Analyzes the sentiment tendency in the text and identifies the emotional state of the customer; Intent Identification Unit: Identify customer intent based on context and keywords.

Images