A problem analysis customer service system based on intelligent voice

Abstract

The present invention relates to the field of intelligent voice interaction technology, and in particular to a problem analysis customer service system based on intelligent voice, which includes: a voice collection unit, which is used to receive and recognize language text; a preprocessing unit, which is used to obtain dialect keywords and correspond each dialect keyword to a number of common keywords with the same semantics; a voice numbering unit, which is used to generate a common number according to the strokes and tones of each character in each common keyword; a voice reply unit, which is used to determine the optimal solution according to the key value; a voice feedback unit, which is used to receive feedback voice and determine the execution status of the optimal solution, determine the emotion value based on the feedback voice, and decide to re-select the optimal solution or issue an alarm signal; the various units of the present invention work together to effectively recognize various dialects, and the combination of strokes and tones improves the efficiency of retrieval, further improving the stability and accuracy of the problem analysis customer service system based on intelligent voice.

Description

Technical Field

[0001] The present invention relates to the field of intelligent voice interaction technology, and in particular to a problem analysis customer service system based on intelligent voice. Background Art

[0002] With the continuous expansion of corporate business and the increase in the number of customers, traditional manual customer service faces tremendous pressure when handling a large number of customer inquiries, complaints, suggestions and other issues. The intelligent voice customer service system can achieve uninterrupted service, respond to customer needs in a timely manner, improve service efficiency and customer satisfaction, and provide targeted services based on the language habits of different customers.

[0003] Chinese patent application publication number: CN119089153A, discloses an intelligent quality inspection method and system based on deep learning. The invention provides an intelligent quality inspection method and system based on deep learning, wherein an embodiment of the present invention obtains text message and voice data from multiple channels from a database to form a large-scale multi-source data set; pre-processes the multi-source data set based on dialect recognition rules to identify and mark dialect vocabulary; introduces a context analysis algorithm to perform context analysis on the pre-processed multi-source data set to obtain an intermediate data set; uses a natural language processing model based on deep learning to analyze the intermediate data set, extracts basic language features, and obtains a feature set to analyze whether the feature set contains preset information; the technical solution provided by the invention improves the consistency and accuracy of data processing, reduces misjudgments and missed detections, reduces operating costs, quickly adapts to new changes, and provides enterprises and institutions with an efficient, reliable and automated quality inspection solution.

[0004] Chinese patent application publication number: CN119229904A, discloses an online voice anti-harassment intelligent recognition method and system. The invention discloses an online voice anti-harassment intelligent recognition method and system, including: predicting the language type that the user about to call may use, and pre-selecting a language emotion judgment model based on the language type that the customer may use; using the pre-selected language emotion judgment model to perform real-time emotion analysis in real time during the call, and outputting a preset emotion analysis result; obtaining audio of the user's online voice for a certain duration at the beginning of the call, and converting the audio into text, judging the user's language type through the audio and text, and simultaneously performing dual real-time emotion analysis to obtain a real-time emotion analysis result; judging whether the two selected language emotion judgment models are the same, and outputting the final emotion analysis result after judgment. This application combines the predicted language emotion judgment model with the real-time language emotion judgment model to perform online voice anti-harassment intelligent recognition, and the recognition result is more accurate, which can improve the quality of customer after-sales maintenance service.

[0005] However, the above method has the following problems: context analysis of the processed dialect text fails to accurately extract language features, and simply comparing and analyzing user emotions fails to accurately address user needs. Summary of the Invention

[0006] To this end, the present invention provides a customer service system for problem analysis based on intelligent voice to overcome the problems in the prior art that context analysis of processed dialect text fails to accurately extract language features, and simply comparing and analyzing user emotions fails to accurately solve user needs.

[0007] To achieve the above objectives, the present invention provides a problem analysis and customer service system based on intelligent voice, comprising:

[0008] A speech collection unit, which is used to receive and identify a language text as a general text or a dialect text, wherein the general text is composed of a plurality of general keywords, and the dialect text is composed of a plurality of dialect keywords;

[0009] a preprocessing unit connected to the speech collection unit, configured to obtain the dialect keywords based on the comparison between the dialect text and the dialect training model, and to map each dialect keyword to a plurality of common keywords with the same semantics;

[0010] a voice numbering unit, connected to the voice collection unit and the preprocessing unit, for generating universal numbers based on the strokes and tones of each character in each universal keyword, and arranging the universal numbers in the order of the characters in the dialect text or the universal text to form a plurality of universal numbering groups;

[0011] A voice reply unit is connected to the voice numbering unit, and is used to compare the general numbering group with the general training model to obtain a number of general associated sentences, calculate the association probability value between the general number and the adjacent general number in each general associated sentence, and comprehensively obtain the sentence association probability value of each general associated sentence, select the general associated sentence whose sentence association probability value is greater than the preset sentence association probability value as the priority associated sentence, and determine the weight of each general keyword contained in each priority associated sentence, calculate the key value, and determine the optimal solution according to the key value.

[0012] Furthermore, it also includes:

[0013] A voice feedback unit is connected to the voice reply unit, and is used to receive feedback voice, determine the execution status of the optimal solution, and, when the optimal solution does not solve the problem of the language text, determine the emotion value based on the feedback voice, and decide whether to re-select the optimal solution or issue an alarm signal based on the comparison result of the emotion value and the preset emotion value.

[0014] Furthermore, the voice numbering unit includes:

[0015] a numbering subunit, connected to the voice collection unit and the preprocessing unit, respectively, for counting the strokes and tones of the top portion of each character in the universal keyword and generating the universal number by combining it with a numbering model, wherein the numbering model is a model that assigns a one-to-one correspondence between strokes and tones and numerical values;

[0016] An arrangement subunit is connected to the numbering subunit and is used to arrange the universal numbers generated by different universal keywords according to the order of the characters in the dialect text or the universal text corresponding to the universal keywords to generate the universal number group.

[0017] Furthermore, the voice reply unit includes:

[0018] The probability calculation subunit is connected to the voice number unit and is used to compare the general number group with the common training model to obtain a number of general associated sentences, calculate the association probability value between the general number and the adjacent general number in each general associated sentence, and comprehensively obtain the sentence association probability value of each general associated sentence, wherein:

[0019] For a single universal number, the association probability value is the average of the probability value of the universal number being located behind the previous universal number and the probability value of the universal number being located before the next universal number in the universal association sentence.

[0020] The average value of the probability value of the common numbers at both ends of the common associated sentence appearing at both ends of the sentence and the probability value of the adjacent common numbers.

[0021] Furthermore, the voice reply unit further includes:

[0022] The probability comparison subunit is connected to the probability calculation subunit and is used to compare the sentence association probability value with the preset sentence association probability value, and determine whether the general association sentence is a priority association sentence based on the comparison result, wherein:

[0023] If the sentence association probability value is greater than or equal to the preset sentence association probability value, then determining that the general association sentence is the priority association sentence;

[0024] The preset sentence association probability value is negatively correlated with the number of general numbers included in the general associated sentence.

[0025] Furthermore, the voice reply unit further includes:

[0026] The solution output subunit is connected to the probability comparison subunit, and is used to determine the weight of each common keyword contained in each priority association sentence, calculate the key value, and select the maximum key value as the optimal solution.

[0027] Furthermore, the voice feedback unit includes:

[0028] A feedback subunit is connected to the voice reply unit, and is used to receive the feedback voice, select a judgment word in the feedback voice, and determine the execution status of the optimal solution based on the judgment word, wherein the execution status includes whether the optimal solution is not resolved and whether the optimal solution is resolved.

[0029] Furthermore, the voice feedback unit further includes:

[0030] An alarm subunit is connected to the feedback subunit and is used to analyze the emotional vocabulary in the feedback speech to determine the emotional value when the optimal solution does not solve the problem of the language text, and to determine whether to reselect the optimal solution or issue the alarm signal based on the comparison result between the emotional value and the preset emotional value, wherein:

[0031] If the emotion value is less than the preset emotion value, determining to reselect the optimal solution;

[0032] If the emotion value is greater than or equal to the preset emotion value, determining to send the alarm signal;

[0033] The preset emotion value is positively correlated with the total number of user feedbacks.

[0034] Furthermore, the voice collection unit detects special words in the language text, and identifies the language text as a general text or a dialect text based on the usage frequency of general words.

[0035] Furthermore, the pre-processing unit establishes the dialect training model according to the semantics of each dialect, and enters the corresponding dialect text into the dialect training model when the dialect keywords cannot be obtained by comparison.

[0036] Compared with the existing technology, the beneficial effect of the present invention is that the system of the present invention comprehensively processes Mandarin and dialects, and comprehensively converts dialects into Mandarin with the same semantics. Mandarin has a standardized and unified vocabulary, grammar and phonetic system. Using Mandarin as the basis for retrieval can avoid retrieval errors caused by the diversity and arbitrariness of dialect pronunciation and wording, can quickly and accurately find relevant content, reduce information omissions or mismatches caused by dialect differences, and greatly improve retrieval efficiency and accuracy. my country has a vast territory and many dialects. People in different regions use dialects to express themselves very differently. By converting dialects into Mandarin for retrieval, regional language barriers are broken, so that people in different dialect areas can obtain information within a unified language framework, which facilitates the integration of information scattered in different regions and recorded in different dialects. In the process of informatization construction, the retrieval system based on Mandarin helps to establish unified and standardized information management standards, promote the standardization and normalization process of informatization construction, and improve the overall informatization level of society, effectively improving the stability and accuracy of the problem analysis customer service system based on intelligent voice.

[0037] Furthermore, the present invention receives user feedback, determines whether the customer problem is solved, and makes an emotional judgment on the feedback. By receiving user feedback and judging whether the problem is solved, the enterprise or service organization can accurately understand whether the customer's actual needs are met. If the problem is properly solved, the customer will feel valued and cared for, thereby improving satisfaction with the product or service. Making an emotional judgment on the feedback can further understand the customer's feelings in the problem-solving process. If the customer's emotion is positive, it means that the entire service process experience is good; if the emotion is negative, targeted measures can be taken to make improvements, soothe the customer's emotions in time, and further improve satisfaction. By judging whether the customer's problem is solved, it is possible to discover the deficiencies in the product or service, such as functional defects, cumbersome processes, etc., so as to make targeted optimization and improvements. Judging feedback emotions can help companies understand the severity and scope of the problem more deeply. In the process of analyzing user feedback, in addition to solving existing problems, it is also possible to discover customers' potential needs. Judging feedback emotions can help companies better grasp the urgency of customer needs and give priority to meeting those needs that are strongly reflected by customer emotions, thereby occupying a favorable position in market competition. The data accumulated from user feedback and its emotional judgment can provide strong support for the company's decision-making. Through the analysis of these data, companies can understand customer behavior patterns, demand trends, and satisfaction changes, etc., so as to formulate more scientific and reasonable strategic plans, product development plans and marketing strategies, realize data-driven refined management and decision-making, and further improve the stability and accuracy of the problem analysis customer service system based on intelligent voice.

[0038] Furthermore, the present invention generates numbers by combining the upper strokes and tones of the characters, and searches in the database according to the numbers. Traditional text search methods, such as by pinyin, radicals, etc., may have the problem of slow search speed when processing large amounts of text data. Combining the upper strokes and tones of the characters to generate numbers establishes a unique and relatively concise encoding system for the characters. Through number search, the database can quickly locate the corresponding text information, greatly reducing the search time and improving the efficiency of information acquisition. It is especially suitable for large-scale text databases. In some cases, pinyin search may lead to inaccurate search results due to the existence of homophones, and radical search may also have deviations due to differences in radical judgment. The numbers generated by combining the upper strokes and tones have higher uniqueness. The uniqueness of the text can more accurately distinguish different characters. Even if homophones or characters with similar shapes are encountered, the target characters can be found quickly and accurately through numbering, which reduces the mismatch of search results and improves the accuracy of retrieval. This retrieval method provides a new dimension for text retrieval and complements the traditional retrieval method. In the digital age, the processing and management of text information are increasingly dependent on computer technology. The method of generating numbers based on the upper strokes and tones of the text facilitates computers to quickly identify, classify and retrieve text, providing an effective means for text information processing, helping to promote the digitalization and intelligent development of text information, improve the level of automation in text information processing, and further enhance the stability and accuracy of the problem analysis customer service system based on intelligent voice.

[0039] Furthermore, the present invention can quickly find the same or similar problems and their solutions in historical data by analyzing the probability of adjacent word combinations. When encountering a new problem, the present invention can quickly match the same or similar problems in historical data by analyzing the probability of adjacent word combinations. In this way, there is no need to analyze and find solutions from scratch, and existing successful experiences can be directly borrowed, which greatly shortens the time to solve the problem. The solutions in historical data can be reused to avoid repeated investment of resources, whether it is human, material or time costs, which can be effectively controlled. The solutions accumulated in historical data are often tested and optimized in practice. By finding solutions to similar problems, the excellent experience and methods therein can be borrowed to make the solutions to new problems more complete and reliable. When facing complex problems and needing to make decisions, referring to the processing conditions and results of similar problems in historical data can provide information for decision-making. It provides a strong basis. After analyzing the probability of adjacent text combinations to find similar cases, decision makers can understand the advantages and disadvantages of different solutions and the possible impacts, so as to make more informed decisions. In the field of marketing, companies can refer to the results of similar marketing activities in the past to formulate more effective marketing strategies. By analyzing adjacent text combinations in a large amount of historical data, not only can solutions to the same or similar problems be found, but also potential problems and trends can be discovered. In the service industry, based on the questions asked by customers and the solutions to similar problems, the deficiencies in the service process can be discovered, and then optimized and improved. In product research and development, referring to user feedback problems and historical solutions can help understand product defects and user needs, carry out targeted upgrades and improvements to products, improve user satisfaction, and further improve the stability and accuracy of the problem analysis customer service system based on intelligent voice. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 This is a structural diagram of the intelligent voice-based problem analysis and customer service system of the present invention;

[0041] Figure 2 This is a schematic diagram of the structure of a voice reply unit according to an embodiment of the present invention;

[0042] Figure 3 A decision diagram for priority-related statements according to an embodiment of the present invention;

[0043] Figure 4 This is a diagram for determining the emotion value according to an embodiment of the present invention. DETAILED DESCRIPTION

[0044] In order to make the objects and advantages of the present invention more clearly understood, the present invention is further described below in conjunction with embodiments; it should be understood that the specific embodiments described herein are merely used to explain the present invention and are not intended to limit the present invention.

[0045] The preferred embodiments of the present invention are described below with reference to the accompanying drawings. It should be understood by those skilled in the art that these embodiments are only used to explain the technical principles of the present invention and are not intended to limit the scope of protection of the present invention.

[0046] It should be noted that, in the description of the present invention, terms such as "up", "down", "left", "right", "inside", and "outside" indicating directions or positional relationships are based on the directions or positional relationships shown in the accompanying drawings. This is only for the convenience of description and does not indicate or imply that the device or element must have a specific orientation, be constructed and operated in a specific orientation. Therefore, it cannot be understood as a limitation on the present invention.

[0047] Furthermore, it should be noted that, in the description of the present invention, unless otherwise expressly specified or limited, the terms "mounted," "connected," and "connected" should be understood in a broad sense. For example, they may refer to fixed connections, detachable connections, or integral connections; mechanical connections or electrical connections; direct connections or indirect connections through an intermediate medium; and internal communication between two components. Those skilled in the art will understand the specific meanings of the above terms in the present invention based on specific circumstances.

[0048] See also Figure 1 As shown in FIG, which is a structural diagram of the problem analysis and customer service system based on intelligent voice of the present invention, an embodiment of the present invention provides a problem analysis and customer service system based on intelligent voice, including:

[0049] A speech collection unit, which is used to receive and identify language text as general text or dialect text, wherein the general text is composed of a plurality of general keywords, and the dialect text is composed of a plurality of dialect keywords;

[0050] A preprocessing unit connected to the speech collection unit is used to obtain dialect keywords based on the comparison of the dialect text with the dialect training model, and to map each dialect keyword to a number of common keywords with the same semantics;

[0051] a voice numbering unit, connected to the voice collection unit and the preprocessing unit, for generating universal numbers based on the strokes and tones of each character in each universal keyword, and arranging the universal numbers in the order of the characters in the dialect text or the universal text to form a plurality of universal numbering groups;

[0052] The voice reply unit is connected to the voice numbering unit and is used to compare the general number group with the general training model to obtain several general associated sentences, calculate the association probability value between the general number and the adjacent general number in each general associated sentence, and comprehensively obtain the sentence association probability value of each general associated sentence, select the general associated sentence with a sentence association probability value greater than the preset sentence association probability value as the priority associated sentence, and determine the weight of each general keyword contained in each priority associated sentence, calculate the key value, and determine the optimal solution according to the key value.

[0053] It can be understood that the universal keywords existing in the language text identified as the dialect text and the universal keywords corresponding to the dialect keywords are uniformly generated with universal numbers for sorting.

[0054] Specifically, the system of the present invention comprehensively processes Mandarin and dialects, and comprehensively converts dialects into Mandarin with the same semantics. Mandarin has a standardized and unified vocabulary, grammar and phonetic system. Using Mandarin as the basis for retrieval can avoid retrieval errors caused by the diversity and arbitrariness of dialect pronunciation and wording, and can quickly and accurately find relevant content, reduce information omissions or mismatches caused by dialect differences, and greatly improve retrieval efficiency and accuracy. my country has a vast territory and many dialects. People in different regions use dialects to express themselves very differently. By converting dialects into Mandarin for retrieval, regional language barriers are broken, so that people in different dialect areas can obtain information within a unified language framework, which facilitates the integration of information scattered in different regions and recorded in different dialects. In the process of informatization construction, a retrieval system based on Mandarin helps to establish unified and standardized information management standards, promote the standardization and normalization process of informatization construction, and improve the overall informatization level of society, effectively improving the stability and accuracy of the problem analysis customer service system based on intelligent voice.

[0055] Specifically, it also includes:

[0056] The voice feedback unit is connected to the voice reply unit to receive feedback voice and determine the execution status of the optimal solution. When the optimal solution does not solve the problem of the language text, the emotion value is determined based on the feedback voice, and the optimal solution is re-selected or an alarm signal is issued according to the comparison result between the emotion value and the preset emotion value.

[0057] Specifically, the present invention determines whether the customer's problem is solved by receiving the user's feedback, judges the emotion of the feedback. By receiving the user's feedback and judging whether the problem is solved, the enterprise or service institution can accurately understand whether the actual needs of the customer are met. If the problem is properly solved, the customer will feel valued and cared for, thus enhancing the satisfaction with the product or service. Judging the emotion of the feedback can further understand the customer's feelings during the problem-solving process. If the customer's emotion is positive, it indicates that the entire service process experience is good; if the emotion is negative, targeted measures can be taken for improvement to soothe the customer's emotion in a timely manner and further improve the satisfaction. Judging whether the customer's problem is solved can discover the deficiencies in the product or service, such as functional defects, cumbersome processes, etc., and thus optimize and improve them targeted. Judging the emotion of the feedback can help the enterprise more deeply understand the severity and scope of the problem. In the process of analyzing user feedback, in addition to solving existing problems, potential customer needs may also be discovered. Judging the emotion of the feedback can help the enterprise better grasp the urgency of customer needs and prioritize meeting the needs strongly reflected by the customer's emotion, so as to gain an advantageous position in the market competition. The data accumulated from user feedback and its emotion judgment can provide strong support for the enterprise's decision-making. By analyzing these data, the enterprise can understand the customer's behavior patterns, demand trends, and satisfaction changes, etc., and thus formulate more scientific and reasonable strategic plans, product R & D plans, and marketing strategies, achieving data-driven refined management and decision-making, and further enhancing the stability and accuracy of the problem analysis customer service system based on intelligent voice.

[0058] Specifically, the voice numbering unit includes:

[0059] A numbering subunit, which is respectively connected to the voice acquisition unit and the preprocessing unit, and is used to count the strokes and tones at the topmost part of each character in the general keywords, and generate a general number in combination with a numbering model. The numbering model is a model that corresponds strokes and tones to numerical values one by one; [[ID=,8]]

[0060] In a specific embodiment, the stroke at the topmost part of the character is obtained by moving a horizontal line downward from above the regular script character, and all the basic strokes that are first touched are the topmost strokes. If multiple basic strokes are touched, they are arranged in order from left to right.

[0061] It can be understood that in the numbering model, the basic strokes include: dot, horizontal line, vertical line, left-falling stroke, and right-falling stroke, corresponding to 1 to 5 in sequence, and the tones include high level, rising tone, falling-rising tone, and falling tone, corresponding to 6 to 9 in sequence.

[0062] In a specific embodiment, the character "我" is numbered. By moving a horizontal line downward from above the regular script character, the basic strokes first touched include: left-falling stroke, right-falling stroke, and dot, and its tone is falling-rising tone. Therefore, the number of "我" is 4518.

[0063] The arrangement subunit is connected to the numbering subunit and is used to arrange the universal numbers generated by different universal keywords according to the order of the words in the dialect text or universal text corresponding to the universal keywords to generate a universal number group.

[0064] In a specific embodiment, the text in the general text is "Who am I", where "I" is numbered 4518, "is" is numbered 27, and "who" is numbered 1417, so the number group of "Who am I" is 4518271417.

[0065] Specifically, the present invention generates numbers by combining the upper strokes and tones of the characters, and searches in the database according to the numbers. Traditional text search methods, such as by pinyin, radicals, etc., may have the problem of slow search speed when processing large amounts of text data. By combining the upper strokes and tones of the characters to generate numbers, a unique and relatively concise coding system is established for the characters. Through number search, the database can quickly locate the corresponding text information, greatly reducing the search time and improving the efficiency of information acquisition. It is especially suitable for large-scale text databases. In some cases, pinyin search may lead to inaccurate search results due to the existence of homophones, and radical search may also have deviations due to differences in radical judgment. The numbers generated by combining the upper strokes and tones have higher uniqueness. The uniqueness of the text can more accurately distinguish different characters. Even if homophones or characters with similar shapes are encountered, the target characters can be found quickly and accurately through numbering, which reduces the mismatch of search results and improves the accuracy of retrieval. This retrieval method provides a new dimension for text retrieval and complements the traditional retrieval method. In the digital age, the processing and management of text information are increasingly dependent on computer technology. The method of generating numbers based on the upper strokes and tones of the text facilitates computers to quickly identify, classify and retrieve text, providing an effective means for text information processing, helping to promote the digitalization and intelligent development of text information, improve the level of automation in text information processing, and further enhance the stability and accuracy of the problem analysis customer service system based on intelligent voice.

[0066] See also Figure 2 As shown in FIG, which is a schematic diagram of the structure of a voice response unit according to an embodiment of the present invention, the voice response unit includes:

[0067] The probability calculation subunit is connected to the voice number unit and is used to compare the general number group with the common training model to obtain several general related sentences, calculate the association probability value between the general number and the adjacent general number in each general related sentence, and comprehensively obtain the sentence association probability value of each general related sentence, wherein,

[0068] For a single universal number, the association probability value is the average of the probability value of the universal number in the universal conjunction sentence being behind the previous universal number and the probability value of the universal number being in front of the next universal number.

[0069] The average of the probability values ​​of the common numbers at both ends of a common association sentence and the probability values ​​of adjacent common numbers.

[0070] It can be understood that the probability value of a universal number in a universal conjunction sentence being located after the previous universal number is the percentage of the number of times the universal number is located after the previous universal number divided by the number of times the universal number and the previous universal number appear in the same universal number group in the common training model;

[0071] The probability value of a universal number in a universal conjunction sentence being located before the next universal number is the percentage of the number of times the universal number is located before the next universal number divided by the number of times the universal number and the next universal number appear in the same universal number group in the common training model;

[0072] The probability value of the common number at both ends of the common association sentence appearing at both ends of the sentence is the percentage of the number of times the common number appears at both ends of the sentence divided by the number of times the common number appears in the common training model;

[0073] The probability value of the common number and the adjacent common number at both ends of the common association sentence is the percentage of the number of times the common number appears at both ends of the sentence next to the adjacent common number divided by the number of times the common number appears at both ends of the sentence in the common training model;

[0074] In a specific embodiment, the number group for the text "Who am I" in the general text is 4518271417. For "yes", the associated probability value is 24.75% for the probability value behind the general number of "I" and 13.68% for the probability value in front of the general number of "who", with an average value of 19.22%.

[0075] In a specific embodiment, when the sentence association probability values ​​of each common association sentence are comprehensively obtained, the number group of the text "Who am I" in the common text is 4518271417. For "I", its association probability value is 18.42%, for "is", its association probability value is 19.22%, and for "who", its association probability value is 20.63%. The sentence association probability value of "Who am I" is (18.42%+19.22%+20.63%) / 3=19.42%.

[0076] Specifically, the present invention quickly finds the same or similar problems and their solutions in historical data by analyzing the probability of adjacent word combinations. When encountering a new problem, the present invention can quickly match the same or similar problems in historical data by analyzing the probability of adjacent word combinations. In this way, there is no need to analyze and find solutions from scratch, and existing successful experiences can be directly borrowed, which greatly shortens the time to solve the problem. The solutions in historical data are reused to avoid repeated investment of resources, whether it is human, material or time costs, all of which can be effectively controlled. The solutions accumulated in historical data are often tested and optimized in practice. By finding solutions to similar problems, the excellent experience and methods therein can be borrowed to make the solutions to new problems more complete and reliable. When facing complex problems and need to make decisions, referring to the processing and results of similar problems in historical data can provide information for decision-making. It provides a strong basis. After analyzing the probability of adjacent text combinations to find similar cases, decision makers can understand the advantages and disadvantages of different solutions and the possible impacts, so as to make more informed decisions. In the field of marketing, companies can refer to the results of similar marketing activities in the past to formulate more effective marketing strategies. By analyzing adjacent text combinations in a large amount of historical data, not only can solutions to the same or similar problems be found, but also potential problems and trends can be discovered. In the service industry, based on the questions asked by customers and the solutions to similar problems, the deficiencies in the service process can be discovered, and then optimized and improved. In product research and development, referring to user feedback problems and historical solutions can help understand product defects and user needs, carry out targeted upgrades and improvements to products, improve user satisfaction, and further improve the stability and accuracy of the problem analysis customer service system based on intelligent voice.

[0077] See also Figure 3 As shown, it is a decision diagram for priority related sentences in an embodiment of the present invention, and the voice reply unit also includes:

[0078] The probability comparison subunit is connected to the probability calculation subunit and is used to compare the sentence association probability value with the preset sentence association probability value, and determine whether the general association sentence is a priority association sentence based on the comparison result, wherein:

[0079] If the statement association probability value is greater than or equal to the preset statement association probability value, the general association statement is determined to be the priority association statement;

[0080] If the statement association probability value is less than the preset statement association probability value, the general association statement is determined to be a non-priority association statement;

[0081] In a specific embodiment, the preset sentence association probability value is set to 15%. If the sentence association probability value is 16.27% and is greater than the preset sentence association probability value, the general association sentence is determined to be the priority association sentence.

[0082] If the sentence association probability value of 7.8% is less than the preset sentence association probability value, the general association sentence is determined to be a non-priority association sentence.

[0083] The association probability value of the preset statement is negatively correlated with the number of common numbers contained in the common association statement.

[0084] It is understandable that the more general numbers a general association sentence contains, the lower the probability that different general numbers are located in adjacent positions. Therefore, the preset sentence association probability value is negatively correlated with the number of general numbers contained in the general association sentence.

[0085] Specifically, the voice response unit also includes:

[0086] The solution output subunit is connected to the probability comparison subunit to determine the weight of each common keyword contained in each priority association sentence, calculate the key value, and select the maximum key value as the optimal solution.

[0087] It is understandable that the weight of each general keyword included in each priority associated sentence is determined according to the search frequency of each general keyword in the common training model.

[0088] In a specific embodiment, for the priority associated statement of "Who am I", where the weight of "I" is set to 24%, the weight of "is" is set to 10%, and the weight of "who" is set to 33%, the key value of "Who am I" is (24%+10%+33%) / 3=22.33%.

[0089] In a specific embodiment, the key values ​​of the priority association statements of A, B, and C are 22.33%, 24.72%, and 20.59% respectively, and the priority association statement of B is selected as the optimal solution.

[0090] Specifically, the voice feedback unit includes:

[0091] The feedback subunit is connected to the voice response unit, and is used to receive the feedback voice, select the judgment words in the feedback voice, and determine the execution status of the optimal solution based on the judgment words, wherein the execution status includes whether the optimal solution is not solved and whether the optimal solution is solved.

[0092] See also Figure 4 As shown in FIG. , which is a judgment diagram of the emotion value according to an embodiment of the present invention, the voice feedback unit further includes:

[0093] The alarm subunit is connected to the feedback subunit and is used to analyze the emotional words in the feedback speech to determine the emotional value when the optimal solution does not solve the problem of the language text, and to determine whether to re-select the optimal solution or issue an alarm signal based on the comparison result between the emotional value and the preset emotional value, wherein,

[0094] If the emotion value is less than the preset emotion value, it is determined to reselect the optimal solution;

[0095] If the emotion value is greater than or equal to the preset emotion value, an alarm signal is issued;

[0096] It can be understood that the sentiment value is the character ratio of emotional words in the feedback voice, where emotional words are words that express the user's current emotional state, for example, very urgent, no time, please wait as soon as possible.

[0097] In a specific embodiment, the preset emotion value is set to 20%. If the emotion value is 12% and is less than the preset emotion value, it is determined that the optimal solution is to be reselected.

[0098] If the emotion value is 34% and is greater than the preset emotion value, it is determined that an alarm signal is issued.

[0099] The preset emotion value is positively correlated with the total number of user feedbacks.

[0100] It is understandable that the more times a user provides feedback on the same issue, generally speaking, the greater the emotional fluctuations, the more emotional words used, and the larger the emotional value. Therefore, the preset emotional value is positively correlated with the total number of user feedbacks.

[0101] Specifically, the speech collection unit detects special words in the language text, and identifies the language text as a general text or a dialect text based on the usage frequency of general words.

[0102] It is understood that special words are words that do not have the same pronunciation in the general vocabulary. If the frequency of use of general words is low and the language text contains special words, the language text is identified as a dialect text. If the frequency of use of general words is high and the language text does not contain special words or contains only a few special words, the language text is identified as a general text.

[0103] Specifically, the preprocessing unit establishes a dialect training model based on the semantics of each dialect, and when the dialect keywords cannot be obtained by comparison, the corresponding dialect text is entered into the dialect training model.

[0104] Thus far, the technical solutions of the present invention have been described in conjunction with the preferred embodiments shown in the accompanying drawings. However, it will be readily understood by those skilled in the art that the scope of protection of the present invention is obviously not limited to these specific embodiments. Without departing from the principles of the present invention, those skilled in the art may make equivalent changes or substitutions to the relevant technical features, and the technical solutions after such changes or substitutions will fall within the scope of protection of the present invention.

[0105] The foregoing description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Those skilled in the art will readily appreciate that the present invention is susceptible to various modifications and variations. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present invention are intended to be within the scope of protection of the present invention.

Claims

1. A problem analysis customer service system based on intelligent voice, characterized by: include: A speech collection unit, which is used to receive and identify a language text as a general text or a dialect text, wherein the general text is composed of a plurality of general keywords, and the dialect text is composed of a plurality of dialect keywords; a preprocessing unit connected to the speech collection unit, configured to obtain the dialect keywords based on the comparison between the dialect text and the dialect training model, and to map each dialect keyword to a plurality of common keywords with the same semantics; a voice numbering unit, connected to the voice collection unit and the preprocessing unit, for generating universal numbers based on the strokes and tones of each character in each universal keyword, and arranging the universal numbers in the order of the characters in the dialect text or the universal text to form a plurality of universal numbering groups; A voice reply unit is connected to the voice numbering unit, and is used to compare the general numbering group with the general training model to obtain a number of general associated sentences, calculate the association probability value between the general number and the adjacent general number in each general associated sentence, and comprehensively obtain the sentence association probability value of each general associated sentence, select the general associated sentence whose sentence association probability value is greater than the preset sentence association probability value as the priority associated sentence, and determine the weight of each general keyword contained in each priority associated sentence, calculate the key value, and determine the optimal solution according to the key value.

2. The intelligent voice-based problem analysis customer service system according to claim 1, characterized in that: Also includes: A voice feedback unit is connected to the voice reply unit, and is used to receive feedback voice, determine the execution status of the optimal solution, and, when the optimal solution does not solve the problem of the language text, determine the emotion value based on the feedback voice, and decide whether to re-select the optimal solution or issue an alarm signal based on the comparison result of the emotion value and the preset emotion value.

3. The intelligent voice-based problem analysis customer service system according to claim 2, characterized in that: The voice numbering unit includes: a numbering subunit, connected to the voice collection unit and the preprocessing unit, respectively, for counting the strokes and tones of the top portion of each character in the universal keyword and generating the universal number by combining it with a numbering model, wherein the numbering model is a model that assigns a one-to-one correspondence between strokes and tones and numerical values; An arrangement subunit is connected to the numbering subunit and is used to arrange the universal numbers generated by different universal keywords according to the order of the characters in the dialect text or the universal text corresponding to the universal keywords to generate the universal number group.

4. The intelligent voice-based problem analysis customer service system according to claim 3, characterized in that: The voice reply unit includes: The probability calculation subunit is connected to the voice number unit and is used to compare the general number group with the common training model to obtain a number of general associated sentences, calculate the association probability value between the general number and the adjacent general number in each general associated sentence, and comprehensively obtain the sentence association probability value of each general associated sentence, wherein: For a single universal number, the association probability value is the average of the probability value of the universal number being located behind the previous universal number and the probability value of the universal number being located before the next universal number in the universal association sentence. The average value of the probability value of the common numbers at both ends of the common associated sentence appearing at both ends of the sentence and the probability value of the adjacent common numbers.

5. The intelligent voice-based problem analysis customer service system according to claim 4 is characterized in that: The voice reply unit also includes: The probability comparison subunit is connected to the probability calculation subunit and is used to compare the sentence association probability value with the preset sentence association probability value, and determine whether the general association sentence is a priority association sentence based on the comparison result, wherein: If the sentence association probability value is greater than or equal to the preset sentence association probability value, then determining that the general association sentence is the priority association sentence; The preset sentence association probability value is negatively correlated with the number of general numbers included in the general associated sentence.

6. The intelligent voice-based problem analysis customer service system according to claim 5, characterized in that: The voice reply unit also includes: The solution output subunit is connected to the probability comparison subunit, and is used to determine the weight of each common keyword contained in each priority association sentence, calculate the key value, and select the maximum key value as the optimal solution.

7. The intelligent voice-based problem analysis customer service system according to claim 6, characterized in that: The voice feedback unit includes: A feedback subunit is connected to the voice reply unit, and is used to receive the feedback voice, select a judgment word in the feedback voice, and determine the execution status of the optimal solution based on the judgment word, wherein the execution status includes whether the optimal solution is not resolved and whether the optimal solution is resolved.

8. The intelligent voice-based problem analysis customer service system according to claim 7, characterized in that: The voice feedback unit also includes: An alarm subunit is connected to the feedback subunit and is used to analyze the emotional vocabulary in the feedback speech to determine the emotional value when the optimal solution does not solve the problem of the language text, and to determine whether to reselect the optimal solution or issue the alarm signal based on the comparison result between the emotional value and the preset emotional value, wherein: If the emotion value is less than the preset emotion value, determining to reselect the optimal solution; If the emotion value is greater than or equal to the preset emotion value, determining to send the alarm signal; The preset emotion value is positively correlated with the total number of user feedbacks.

9. The intelligent voice-based problem analysis customer service system according to claim 8, characterized in that: The speech collection unit detects special words in the language text, and identifies whether the language text is a general text or a dialect text based on the usage frequency of general words.

10. The intelligent voice-based problem analysis customer service system according to claim 9, characterized in that: The preprocessing unit establishes the dialect training model according to the semantics of each dialect, and enters the corresponding dialect text into the dialect training model when the dialect keywords cannot be obtained by comparison.

Images