Consultation retrieval method and system based on demand tag configuration

Abstract

The present invention discloses a consultation retrieval method and system based on demand tag configuration, which relates to the technical field related to data retrieval processing. The method includes: based on semantic recognition, identifying and classifying the consultation sentence library; based on emotion recognition tags and customer value recognition tags, obtaining the emotion tag library and customer tag library; analyzing the consultation retrieval priority of the demand tag; obtaining historical user consultation record data, and matching according to multiple consultation priorities; randomly adjusting multiple consultation priorities, and matching historical consultation data to calculate the consultation retrieval fitness; and conducting consultation retrieval according to the optimal consultation retrieval scheme. The method solves the technical problems that the existing consultation retrieval cannot accurately analyze the user's demand for information, resulting in inaccurate retrieval results, redundant retrieval information, and low retrieval accuracy and efficiency, and realizes more accurate and efficient information retrieval, achieving the technical effect of improving the efficiency and accuracy of consultation retrieval.

Description

Technical Field

[0001] The present application relates to the field related to data retrieval processing technology, and specifically to a consultation retrieval method and system based on demand tag configuration. Background Art

[0002] In the era of information explosion, it is becoming increasingly difficult for people to obtain and filter the information they need. Both individual users and corporate organizations need an efficient and accurate search tool to help them quickly find information that matches their needs. Consultation retrieval, as a special information retrieval method, focuses on understanding the user's consultation intentions and accurately matching them according to their needs. Through user search needs, the user's consultation questions can be classified into corresponding topics or fields, so as to achieve rapid positioning and screening of relevant information. However, traditional consultation retrieval either searches for information by consulting catalogs, indexes, etc., or performs remote access retrieval through the Internet. Not only is the retrieval efficiency low, the matching degree and accuracy of the retrieval results are not high, but it is also limited by region and time, and cannot accurately and quickly meet the user's needs to obtain information anytime and anywhere.

[0003] Therefore, at the current stage, there is a technical problem in the consulting retrieval related technologies that cannot accurately analyze the user's needs for obtaining information, resulting in inaccurate retrieval results, redundant retrieval information, and low retrieval accuracy and efficiency. Summary of the invention

[0004] This application provides a consulting retrieval method and system based on demand tag configuration, and adopts technical means such as semantic recognition, sentiment classification, and preset retrieval rules to solve the technical problems that the existing consulting retrieval cannot accurately analyze the user's needs for obtaining information, resulting in inaccurate retrieval results, redundant retrieval information, and low retrieval accuracy and efficiency. It realizes more accurate and efficient information retrieval, and achieves the technical effect of improving the efficiency and accuracy of consulting retrieval.

[0005] The present application provides a consultation retrieval method based on demand tag configuration, the method comprising: based on semantic recognition, identifying and classifying a consultation sentence library to obtain a consultation question library; based on emotion recognition tags and customer value recognition tags, obtaining an emotion tag library and a customer tag library, combining the consultation question library to obtain a demand tag library, wherein the demand tag library includes multiple demand tags; analyzing the consultation retrieval priorities of the multiple demand tags to obtain multiple consultation priorities as a basic consultation retrieval scheme; obtaining historical user consultation record data, matching them according to the multiple consultation priorities to obtain matching historical consultation data, and calculating and obtaining a first consultation retrieval fitness; randomly adjusting the multiple consultation priorities in the basic consultation retrieval scheme, matching the historical consultation data to calculate the consultation retrieval fitness, optimizing the basic consultation retrieval scheme until convergence, and obtaining an optimal consultation retrieval scheme; according to the optimal consultation retrieval scheme, conducting consultation retrieval on the consultation needs configured with the multiple demand tags.

[0006] In a possible implementation, based on semantic recognition, the consulting statement library is identified and classified to obtain a consulting question library, and the following processing is performed: based on historical consulting data records, a set of historical consulting statements is obtained to construct a consulting statement library; a set of sample consulting statements is obtained, and consulting questions in the sample consulting statements are identified to obtain a set of sample consulting questions; based on semantic recognition, the sample consulting statement set and the sample consulting question set are used to construct a consulting question identifier, the consulting statement library is identified, a plurality of consulting questions are obtained, and the consulting questions are classified to obtain a consulting question library.

[0007] In a possible implementation, based on emotion recognition tags and customer value recognition tags, an emotion tag library and a customer tag library are obtained, and combined with the consulting question library, the following processing is performed: a sample consulting audio set is obtained, and user emotions in the sample consulting audio are identified to obtain a sample user emotion set; using the sample consulting audio set and the sample user emotion set, an emotion identifier is constructed based on audio recognition to identify the consulting sentence library, obtain multiple user emotions, and classify them to obtain an emotion tag library; based on user service categories of multiple users, a user tag library is constructed; the consulting question library, the emotion tag library and the customer tag library are combined to obtain the demand tag library.

[0008] In a possible implementation, the consultation retrieval priorities of the multiple demand tags are analyzed to obtain multiple consultation priorities, and the following processing is performed: the multiple demand tags in the demand tag library are sorted and prioritized in order of the probability of occurrence of the consultation questions from large to small to obtain multiple question priorities, wherein the higher the ranking, the greater the question priority; the multiple demand tags in the demand tag library are sorted and prioritized in order of the probability of occurrence of the emotion tags from small to large to obtain multiple emotion priorities; the multiple demand tags in the demand tag library are sorted and prioritized in order of customer value from large to small to obtain multiple user priorities; the multiple question priorities, multiple emotion priorities and multiple customer priorities are weightedly calculated to obtain the multiple consultation priorities.

[0009] In a possible implementation, historical user consultation record data is obtained, matched according to the multiple consultation priorities to obtain matching historical consultation data, and a first consultation retrieval fitness is calculated, and the following processing is also performed: historical user consultation record data is obtained, the historical user consultation record data includes multiple consultation retrieval data sets, each consultation retrieval data set includes multiple demand tags and multiple consultation retrieval orders for multiple consultations; the multiple consultation retrieval data sets are matched according to the multiple consultation priorities to obtain multiple first matching consultation retrieval data sets whose consultation retrieval orders meet the multiple consultation priorities; based on multiple user evaluation data sets of the multiple first matching consultation retrieval data sets, the first consultation retrieval fitness of the basic consultation retrieval scheme is calculated, wherein each user evaluation data includes an evaluation level.

[0010] In a possible implementation, the first consultation retrieval fitness of the basic consultation retrieval scheme is calculated based on the multiple user evaluation data sets of the multiple first matching consultation retrieval data sets, and the following processing is further performed: a consultation retrieval function for calculating the fitness of the consultation retrieval scheme is constructed, as shown in the following formula:

[0011] ;

[0012] Among them, CSC is the consultation retrieval fitness, M is the number of all consultation needs in multiple first matching consultation retrieval data sets, is the weight assigned according to the consultation priority of the i-th consultation demand, is the user evaluation level in the user evaluation data of the i-th consulting demand, is the consultation priority of the ith consultation demand; according to the consultation retrieval function, combined with the consultation priorities of multiple consultation requests in the multiple first matching consultation retrieval data sets, and multiple user evaluation data sets, the first consultation retrieval fitness is calculated.

[0013] In a possible implementation, multiple consultation priorities within the basic consultation retrieval scheme are randomly adjusted, and the consultation retrieval fitness is calculated by matching historical consultation data to optimize the basic consultation retrieval scheme. The following processing is also performed: randomly select P consultation priorities of P demand tags in the demand tag library for random adjustment to obtain P adjusted consultation priorities and an adjusted basic consultation retrieval scheme, where P is an integer greater than 1 and less than the total number of demand tags; based on the adjusted basic consultation retrieval scheme, the consultation retrieval fitness is calculated by matching historical consultation data to obtain a second consultation retrieval fitness; determine whether the second consultation retrieval fitness is greater than the first consultation retrieval fitness, and if so, continue to randomly adjust the P consultation priorities of the P demand tags; if not, use the P demand tags as taboo tags, reselect the consultation priorities of other P demand tags for random adjustment and optimization, and unban the taboo tags after the number of taboos; continue to optimize until convergence, and output the consultation retrieval scheme with the largest consultation retrieval fitness as the optimal consultation retrieval scheme.

[0014] This application also provides a consultation and retrieval system based on demand tag configuration, including:

[0015] A semantic recognition and classification module, which is used to recognize and classify the consulting statement library based on semantic recognition to obtain a consulting question library;

[0016] A demand label library acquisition module, the demand label library acquisition module is used to obtain an emotion label library and a customer label library based on the emotion recognition mark and the customer value recognition mark, and combine the consultation question library to obtain a demand label library, wherein the demand label library includes a plurality of demand labels;

[0017] A search priority analysis module, wherein the search priority analysis module is used to analyze the consultation search priorities of the multiple demand tags to obtain multiple consultation priorities as a basic consultation search solution;

[0018] A first consultation retrieval fitness calculation module, the first consultation retrieval fitness calculation module is used to obtain historical user consultation record data, match according to the multiple consultation priorities, obtain matching historical consultation data, and calculate and obtain a first consultation retrieval fitness;

[0019] A consultation retrieval scheme optimization module, which is used to randomly adjust multiple consultation priorities in the basic consultation retrieval scheme, match historical consultation data to calculate consultation retrieval fitness, optimize the basic consultation retrieval scheme until convergence, and obtain the optimal consultation retrieval scheme;

[0020] A consultation retrieval module is used to perform consultation retrieval on the consultation requirements configured with the multiple requirement tags according to the optimal consultation retrieval solution.

[0021] The consultation retrieval method and system based on demand tag configuration proposed in this application is intended to identify and classify the consultation sentence library based on semantic recognition; obtain the emotion tag library and customer tag library based on emotion recognition tags and customer value recognition tags; analyze the consultation retrieval priority of the demand tag; obtain historical user consultation record data and match it according to multiple consultation priorities; randomly adjust multiple consultation priorities and match historical consultation data to calculate the consultation retrieval fitness; and conduct consultation retrieval according to the optimal consultation retrieval plan. It solves the technical problems that the existing consultation retrieval cannot accurately analyze the user's needs for obtaining information, resulting in inaccurate retrieval results, redundant retrieval information, and low retrieval accuracy and efficiency, and realizes more accurate and efficient information retrieval, achieving the technical effect of improving the efficiency and accuracy of consultation retrieval. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] In order to more clearly illustrate the technical solution of the embodiment of the present disclosure, the accompanying drawings of the embodiment of the present disclosure will be briefly introduced below. A flow chart is used in the present application to illustrate the operations performed by the system according to the embodiment of the present application. It should be understood that the preceding or following operations are not necessarily performed accurately in order. On the contrary, various steps can be processed in reverse order or simultaneously as needed. At the same time, other operations can also be added to these processes, or one or more operations can be removed from these processes.

[0023] Figure 1 A flowchart of a consultation and retrieval method based on demand tag configuration provided in an embodiment of the present application;

[0024] Figure 2 A schematic diagram of the structure of a consultation and retrieval system based on demand tag configuration provided in an embodiment of the present application.

[0025] Explanation of the accompanying drawings: semantic recognition and classification module 10, demand tag library acquisition module 20, retrieval priority analysis module 30, first consultation retrieval fitness calculation module 40, consultation retrieval solution optimization module 50, consultation retrieval module 60. DETAILED DESCRIPTION

[0026] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are listed below.

[0027] In order to make the purpose, technical solutions and advantages of the present application clearer, the present application will be further described in detail below in conjunction with the accompanying drawings. The described embodiments should not be regarded as limiting the present application. All other embodiments obtained by ordinary technicians in the field without making creative work are within the scope of protection of this application.

[0028] In the following description, reference is made to "some embodiments", which describe a subset of all possible embodiments, but it is understood that "some embodiments" may be the same subset or different subsets of all possible embodiments, and may be combined with each other without conflict, and the terms "first\second" involved are merely to distinguish similar objects and do not represent a specific ordering of objects. The terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product, or server that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or modules that are not clearly listed or inherent to these processes, methods, products, or devices. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those generally understood by technicians in the technical field of this application. The terms used herein are for the purpose of describing the embodiments of the present application only.

[0029] The present application embodiment provides a consultation retrieval method based on demand tag configuration, such as Figure 1 As shown, the method includes:

[0030] The consultation retrieval method based on demand tag configuration combines the advantages of modern information technology. It can quickly and accurately understand the user's consultation intention, configure demand tags, and accurately match them according to their configured tags. It can classify the user's consultation questions into the fields and topics corresponding to the demand tags, and achieve rapid positioning and screening of relevant information. It provides users with more personalized information recommendations and solutions, meets users' growing information retrieval needs, and helps users solve various problems such as academic research, business decisions, and life services, and achieve more accurate and efficient information consultation retrieval.

[0031] Step S100, based on semantic recognition, the consulting statement library is identified and classified to obtain a consulting question library. Semantic recognition technology is used to deeply analyze and understand a large number of consulting statements, thereby extracting the semantic meanings of these statements, and classifying and organizing them according to these meanings, and finally forming a structured consulting question library. Specifically, semantic recognition deeply analyzes the grammatical structure, lexical meaning, contextual relationship, etc. of the consulting statements to understand the specific intentions and meanings expressed by the statements. According to these meanings, similar consulting statements are classified into the same topic or category to form different consulting question sets, which constitute the consulting question library.

[0032] In a possible implementation, step S100 further includes step S110, based on historical consultation data records, obtaining a set of historical consultation statements and constructing a consultation statement library. By collecting and analyzing past consultation data records, extracting the consultation statements therein, and organizing these statements into a set, and then constructing a consultation statement library for subsequent analysis or processing, specifically, the historical consultation data records include past consultation records collected from various sources (such as customer service centers, online consultation platforms, emails, etc.), for example, customer questions, consultation content, context of conversations and other information; the historical consultation statement set is usually a set of questions raised by customers or doubts expressed. It also includes step S120, obtaining a set of sample consultation statements, and identifying the consultation questions in the sample consultation statements to obtain a set of sample consultation questions. Representative consulting statements are selected from existing consulting records or databases. The sample consulting statements should be able to cover consulting questions in different fields, different topics and different expressions to form a set as a sample consulting statement set. For each sample consulting statement, it is necessary to conduct in-depth analysis and understanding to accurately identify the consulting questions therein. After identification, a sample consulting question set containing multiple consulting questions is obtained, and each question should be associated with the original consulting statement. It also includes step S130, based on semantic recognition, using the sample consulting statement set and the sample consulting question set to construct a consulting question identifier, identify the consulting statement library, obtain multiple consulting questions, and classify them to obtain a consulting question library. Based on semantic recognition, the sample consulting sentence set is analyzed to extract the keywords, phrases and the relationships between them in the sentences. Using the sample consulting sentence set and the corresponding sample consulting question set, a consulting question identifier can be constructed. By learning the characteristics and rules in the sample data, it can automatically identify the consulting questions in new consulting sentences and apply it to the consulting sentence library. By identifying each sentence in the consulting sentence library, the model can automatically extract the consulting questions and associate them with the corresponding sentences. Multiple consulting questions can be quickly and accurately obtained from a large number of consulting sentences, and the identified consulting questions can be classified and organized to form a consulting question library.

[0033] Step S200, based on the emotion recognition mark and the customer value recognition mark, an emotion tag library and a customer tag library are obtained, and combined with the consulting question library, a demand tag library is obtained, and the demand tag library includes multiple demand tags. The emotion recognition mark is used to capture and analyze the emotional tendency in the consulting statement or question, such as positive, negative, neutral, etc. Through the emotion recognition of the consulting statement, an emotion tag library can be formed, which contains various emotion types and their corresponding tags; the customer value recognition mark is based on factors such as the customer's purchase record, activity, feedback quality, etc. to evaluate the customer's value. Specifically, according to the consulting question obtained by semantic recognition of each consulting statement, the user value is identified and marked, for example, according to the user's payment amount or membership level, a corresponding label is added, and multiple emotion tags and customer tags are obtained, which respectively form an emotion tag library and a customer tag library, and then combined with the consulting question library constructed by semantic recognition, the consulting question, emotion tag and customer tag are combined to form a demand tag library. Each demand tag contains the content of the consulting question, the customer's emotional state and the customer's value information, which can more comprehensively reflect the customer's needs. Through the demand tag library, companies can understand customer needs and expectations more accurately and provide personalized services or solutions for different customers.

[0034] In a possible implementation, step S200 further includes step S210, obtaining a sample consultation audio set, and identifying the user emotions in the sample consultation audio to obtain a sample user emotion set. The sample consultation audio set refers to an audio data set containing user consultation content collected from various channels (such as telephone recordings, voice chat tools, customer service hotlines, etc.), different consultation scenarios, user groups and emotional expressions, and for each sample consultation audio, analyzing the user emotions and judging the type of emotion expressed by the user, such as happiness, anger, sadness, surprise, calmness, etc., annotating the corresponding emotion labels, and organizing all sample consultation audios annotated with emotion labels and their corresponding emotion labels into a set, namely, a sample user emotion set. It also includes step S220, using the sample consultation audio set and the sample user emotion set, based on audio recognition, constructing an emotion recognizer, identifying the consultation sentence library, obtaining multiple user emotions, and classifying them to obtain an emotion label library. By using the collected and annotated sample consultation audio collection and sample user emotion collection, the emotion recognizer is trained, and the user's emotions can be automatically identified according to the features in the audio signal. After the emotion recognizer is constructed, it is applied to the consultation statement library. Specifically, the data in the consultation statement library is usually in text form, and the text needs to be converted into speech first, and the audio is recognized by the emotion recognizer to obtain multiple user emotions. Finally, these recognized user emotions are classified and sorted to form an emotion tag library. It also includes step S230, building a user tag library based on the user service categories of multiple users. According to the user service category of the user, the user data is classified, integrated, and labeled to build a user tag library, including the user's basic information (such as age, gender, occupation, etc.), behavior data (such as visit frequency, usage time, purchase record, etc.) and feedback data (such as evaluation, complaint, suggestion, etc.). It also includes step S240, combining the consultation question library, emotion tag library and customer tag library to obtain the demand tag library. The obtained consultation question library, emotion tag library and customer tag library are integrated to finally obtain the demand tag library.

[0035] Step S300, analyzing the consultation retrieval priorities of the multiple demand tags, and obtaining multiple consultation priorities as basic consultation retrieval schemes. The consultation retrieval priority analysis of multiple demand tags is usually based on evaluation criteria such as the commercial value, urgency, difficulty of implementation, and user expectations of the demand, with reference to historical data, user feedback, and business goals, etc., to assign a priority score to each demand tag, and sort the multiple demand tags based on the priority score to form multiple consultation priorities. Specifically, after the user raises a consultation question, the corresponding basic answer is first retrieved through artificial intelligence for reply, similar to a customer service robot. The consultation priority is the order priority of retrieving the consultation question when multiple users initiate consultation at the same time. The greater the priority, the consultation search is performed first, and the waiting time is short. Conversely, the consultation search is performed later, and the waiting time is long. The consultation priority, for example, is 10 levels from 1 to 10.

[0036] In a possible implementation, step S300 further includes step S310, sorting and assigning priorities to multiple demand tags in the demand tag library in the order of the probability of occurrence of the consulting question from large to small, and obtaining multiple problem priorities, wherein the higher the sorting, the greater the problem priority. It also includes step S320, sorting and assigning priorities to multiple demand tags in the demand tag library in the order of the probability of occurrence of the emotion tag from small to large, and obtaining multiple emotion priorities. The demand tags are sorted in the order of the probability of occurrence of the emotion tag from small to large, and those demands that are more likely to cause negative emotions are identified so as to give higher priority in subsequent services. Negative emotions often mean that users are less satisfied and need to solve their problems or needs more quickly and accurately. Each demand tag is assigned an emotion priority. For example, emotions with a lower probability of occurrence (such as positive, satisfied, etc.) can be assigned a lower priority, while emotions with a higher probability of occurrence (such as negative, dissatisfied, etc.) can be assigned a higher priority. It also includes step S330, sorting and assigning priorities to multiple demand tags in the demand tag library in the order of customer value from large to small, and obtaining multiple user priorities. The method further includes step S340, in which the weighted calculation of the multiple problem priorities, multiple emotional priorities and multiple customer priorities is performed to obtain the multiple consultation priorities. The weight of each priority reflects the importance of different priorities in the consultation retrieval process. For example, the problem priority may reflect the urgency and importance of the consultation problem, the emotional priority may reflect the emotional state and satisfaction of the customer, and the customer priority may be based on the value or importance of the customer. For each consultation case, the corresponding problem priority, emotional priority and customer priority are multiplied by their respective weights to reflect the priority after comprehensive consideration of multiple factors. Finally, the consultation cases are sorted according to the weighted values ​​to obtain multiple consultation priorities as the basis for consultation retrieval processing, ensuring that more important, more urgent or more valuable cases can be given priority when processing a large number of consultations.

[0037] Step S400, obtain historical user consultation record data, match according to the multiple consultation priorities, obtain matching historical consultation data, and calculate the first consultation retrieval fitness. Obtain historical user consultation record data, match the historical consultation record data according to the determined multiple consultation priorities (obtained by weighted calculation based on problem priority, emotion priority and customer priority), obtain a set of historical consultation data matching the current consultation, and calculate the first consultation retrieval fitness. Specifically, according to the consultation record data of multiple time periods in the historical time, according to the order of consultation retrieval in the historical time, match with multiple consultation priorities, and output the consultation data with consistent order and priority. For example, the priority of the first consultation request for consultation retrieval is the maximum value of the priorities of all consultation requests in the time period, and so on, to calculate the fitness of the basic consultation retrieval scheme.

[0038] In a possible implementation, step S400 further includes step S410, obtaining historical user consultation record data, the historical user consultation record data includes multiple consultation retrieval data sets, each consultation retrieval data set includes multiple demand tags and multiple consultation retrieval orders of multiple consultations. The historical user consultation record data refers to the user consultation data accumulated in the past period of time, which contains a large amount of user consultation information, such as consultation time, consultation content, consultation results, etc.; the consultation retrieval data set is a subset divided from the historical user consultation record data, and each consultation retrieval data set focuses on a certain type of specific consultation topic or problem type; each consultation retrieval data set includes multiple demand tags and multiple consultation retrieval orders of multiple consultations, and the consultation retrieval order refers to the order in which users perform retrieval or query in multiple consultations, which is affected by multiple factors, such as the urgency of user needs, the complexity of consultation content, and the satisfaction of historical consultation results. It also includes step S420, matching the multiple consultation retrieval data sets according to the multiple consultation priorities, and obtaining multiple first matching consultation retrieval data sets whose consultation retrieval orders meet the multiple consultation priorities. It also includes step S430, which calculates the first consultation retrieval fitness of the basic consultation retrieval scheme according to the multiple user evaluation data sets of the multiple first matching consultation retrieval data sets, wherein each user evaluation data includes an evaluation level. The effectiveness of the basic consultation retrieval scheme is evaluated in combination with the user's evaluation data on the matched consultation retrieval data set, and each user evaluation data includes an evaluation level, which reflects the user's satisfaction and recognition of the retrieval results. The user evaluation data is the rating after the user's consultation, such as 10 levels of 1-10. According to the user evaluation data of the multiple first matching consultation retrieval data sets, which usually include direct feedback from users, such as scoring, rating, and comments, the average evaluation score of each first matching consultation retrieval data set is calculated, and the first consultation retrieval fitness of the basic consultation retrieval scheme is calculated based on these average evaluation scores.

[0039] In a possible implementation, step S430 further includes step S431, constructing a consultation retrieval function for calculating the fitness of the consultation retrieval solution, as shown in the following formula:

[0040] ;

[0041] Among them, CSC is the consultation retrieval fitness, M is the number of all consultation needs in multiple first matching consultation retrieval data sets, is the weight assigned according to the consultation priority of the i-th consultation demand, is the user evaluation level in the user evaluation data of the i-th consulting demand, is the consultation priority of the i-th consultation demand. The greater the consultation priority, the greater the corresponding weight. The larger the value of . It also includes step S432, according to the consultation retrieval function, combined with the consultation priorities of multiple consultation requests in the multiple first matching consultation retrieval data sets, and multiple user evaluation data sets, to calculate and obtain the first consultation retrieval fitness. The consultation retrieval function is usually used to quickly and accurately find the historical consultation data that best matches the current consultation request in a large amount of consultation data. According to the consultation retrieval function, combined with the consultation priorities of multiple consultation requests in the multiple first matching consultation retrieval data sets, and multiple user evaluation data sets, specifically, the user evaluation data is weighted according to the consultation priority, so that consultation requests with higher priorities occupy a greater weight in the fitness calculation, and the first consultation retrieval fitness is calculated in combination with the matching degree and user evaluation data returned by the consultation retrieval function, reflecting the comprehensive performance of the current consultation retrieval scheme in terms of matching degree and user satisfaction.

[0042] Step S500, randomly adjust the multiple consultation priorities in the basic consultation retrieval scheme, match the historical consultation data to calculate the consultation retrieval fitness, optimize the basic consultation retrieval scheme until convergence, and obtain the optimal consultation retrieval scheme. Randomly adjust the consultation priorities in the basic consultation retrieval scheme to generate a variety of candidate schemes in the search space, use these randomly adjusted consultation priorities to match the historical consultation data, and calculate the corresponding consultation retrieval fitness. According to the calculated consultation retrieval fitness, evaluate the pros and cons of the current retrieval scheme. If the fitness is improved, it means that the random adjustment has produced a positive effect, and this part of the adjustment is retained. On the contrary, if the fitness decreases, it is necessary to return to the previous scheme and repeat it. By constantly making random adjustments, matching historical data, calculating fitness and evaluating the effect, we can gradually optimize the basic consultation retrieval scheme. As the number of iterations increases, the retrieval scheme will gradually converge to a better state, that is, the optimal consultation retrieval scheme.

[0043] In a possible implementation, step S500 further includes step S510, randomly selecting P consultation priorities of P demand tags in the demand tag library for random adjustment, obtaining P adjusted consultation priorities, and obtaining an adjusted basic consultation retrieval scheme, where P is an integer greater than 1 and less than the total number of demand tags. P demand tags are randomly selected from the demand tag library, and the corresponding consultation priorities of the P randomly selected demand tags are randomly adjusted, that is, the priority order or value of these demand tags is changed to obtain P adjusted consultation priorities, and based on the adjusted priorities, an adjusted basic consultation retrieval scheme is formed, aiming to optimize the overall retrieval effect by changing the priorities of some demand tags. It also includes step S520, based on the adjusted basic consultation retrieval scheme, the consultation retrieval fitness is calculated by matching the historical consultation data to obtain a second consultation retrieval fitness. The second consultation retrieval fitness is the consultation retrieval fitness corresponding to the adjusted basic consultation retrieval scheme. It also includes step S530, judging whether the second consultation retrieval fitness is greater than the first consultation retrieval fitness. If so, the P consultation priorities of the P demand tags are continuously randomly adjusted. If not, the P demand tags are used as taboo tags, and the consultation priorities of the other P demand tags are reselected for random adjustment and optimization, and the taboo tags are unbanned after the taboo times. If the second consultation retrieval fitness is greater than the first consultation retrieval fitness, it means that the random adjustment of the consultation priority of the demand tag improves the retrieval effect, and the consultation priority is continued to be randomly adjusted; if the second consultation retrieval fitness is less than the first consultation retrieval fitness, it means that the retrieval effect is worse, and the P demand tags are used as taboo tags, and the consultation priorities of other demand tags are reselected for random adjustment and optimization. Taboo tags are tags whose priorities are not adjusted within the taboo times, so as to find a better solution more quickly. When the taboo times are reached, the taboo tags are unbanned, that is, their consultation priorities can be adjusted. It also includes step S530, continuing the optimization until convergence, and outputting the consulting retrieval scheme with the largest consulting retrieval fitness as the optimal consulting retrieval scheme. Specifically, within the current basic consulting retrieval scheme, the consulting priorities of P demand tags are adjusted to generate new candidate consulting retrieval schemes, the new candidate consulting retrieval schemes are used to match historical consulting data, and their consulting retrieval fitness is calculated based on user evaluation data or other indicators, and the consulting retrieval fitness of the new scheme is compared with the fitness of the current optimal scheme. If the fitness of the new scheme is higher, it is replaced with the current optimal scheme, and the optimization is repeated to continuously generate new candidate schemes and evaluate their fitness until a certain convergence condition is met, for example, the fitness improvement reaches below a small threshold, or the number of iterations reaches a preset number, and the consulting retrieval scheme with the largest consulting retrieval fitness obtained during the iteration process is output as the optimal consulting retrieval scheme.

[0044] Step S600, according to the optimal consultation retrieval scheme, the consultation needs configured with the multiple demand tags are searched for consultation. When a user raises a consultation need, the system will perform a consultation search operation according to the determined optimal consultation retrieval scheme to find the most relevant and valuable consultation data. Specifically, the user's consultation needs are analyzed, key information is extracted, and the key information is converted into demand tags. According to the optimal consultation retrieval scheme obtained through the optimization process, the consultation priority of each demand tag is determined. According to the determined demand tags and the corresponding consultation priorities, the historical consultation data is searched, and the retrieved historical consultation data is sorted based on the matching degree, and the results with the highest matching degree and the most value to the current consultation needs are displayed first.

[0045] In the above, refer to Figure 1 The consultation retrieval method based on demand tag configuration according to an embodiment of the present invention is described in detail. Figure 2 A consultation and retrieval system based on demand tag configuration according to an embodiment of the present invention is described.

[0046] The consulting retrieval system based on demand tag configuration according to the embodiment of the present invention is used to solve the technical problem that the existing consulting retrieval cannot accurately analyze the user's demand for obtaining information, resulting in inaccurate retrieval results, redundant retrieval information, and low retrieval accuracy and efficiency, and realizes more accurate and efficient information retrieval, achieving the technical effect of improving the efficiency and accuracy of consulting retrieval. The consulting retrieval system based on demand tag configuration includes: a semantic recognition and classification module 10, a demand tag library acquisition module 20, a retrieval priority analysis module 30, a first consulting retrieval fitness calculation module 40, a consulting retrieval scheme optimization module 50, and a consulting retrieval module 60.

[0047] A semantic recognition and classification module 10, wherein the semantic recognition and classification module 10 is used to recognize and classify the consulting statement library based on semantic recognition to obtain a consulting question library;

[0048] A demand label library acquisition module 20, the demand label library acquisition module 20 is used to obtain an emotion label library and a customer label library based on the emotion recognition mark and the customer value recognition mark, and combine the consultation question library to obtain a demand label library, wherein the demand label library includes a plurality of demand labels;

[0049] A search priority analysis module 30, wherein the search priority analysis module 30 is used to analyze the consultation search priorities of the multiple demand tags to obtain multiple consultation priorities as a basic consultation search solution;

[0050] A first consultation retrieval fitness calculation module 40, which is used to obtain historical user consultation record data, match according to the multiple consultation priorities, obtain matching historical consultation data, and calculate and obtain a first consultation retrieval fitness;

[0051] A consultation retrieval scheme optimization module 50, which is used to randomly adjust multiple consultation priorities in the basic consultation retrieval scheme, match historical consultation data to calculate consultation retrieval fitness, optimize the basic consultation retrieval scheme until convergence, and obtain the optimal consultation retrieval scheme;

[0052] The consultation retrieval module 60 is used to perform consultation retrieval on the consultation requirements configured with the multiple requirement tags according to the optimal consultation retrieval solution.

[0053] The specific configuration of the semantic recognition and classification module 10 will be described in detail below. The semantic recognition and classification module 10 further includes: based on the historical consultation data records, obtaining a set of historical consultation statements and constructing a consultation statement library; obtaining a set of sample consultation statements, and identifying the consultation questions in the sample consultation statements to obtain a set of sample consultation questions; based on semantic recognition, using the sample consultation statement set and the sample consultation question set to construct a consultation question identifier, identifying the consultation statement library, obtaining multiple consultation questions, and classifying them to obtain a consultation question library.

[0054] The specific configuration of the demand tag library acquisition module 20 will be described in detail below. The demand tag library acquisition module 20 may further include: obtaining a sample consultation audio set, and identifying the user emotions in the sample consultation audio to obtain a sample user emotion set; using the sample consultation audio set and the sample user emotion set, based on audio recognition, constructing an emotion identifier, recognizing the consultation statement library, obtaining multiple user emotions, and classifying them to obtain an emotion tag library; constructing a user tag library based on user service categories of multiple users; combining the consultation question library, the emotion tag library and the customer tag library to obtain the demand tag library.

[0055] The specific configuration of the retrieval priority analysis module 30 will be described in detail below. The retrieval priority analysis module 30 may further include: sorting and assigning priorities to the multiple demand tags in the demand tag library in the order of the probability of occurrence of the consulting question from large to small, and obtaining multiple problem priorities, wherein the higher the ranking, the greater the problem priority; sorting and assigning priorities to the multiple demand tags in the demand tag library in the order of the probability of occurrence of the emotion tag from small to large, and obtaining multiple emotion priorities; sorting and assigning priorities to the multiple demand tags in the demand tag library in the order of customer value from large to small, and obtaining multiple user priorities; performing weighted calculation on the multiple problem priorities, multiple emotion priorities and multiple customer priorities, and obtaining the multiple consulting priorities.

[0056] The specific configuration of the first consultation retrieval fitness calculation module 40 will be described in detail below. The first consultation retrieval fitness calculation module 40 further includes: obtaining historical user consultation record data, the historical user consultation record data includes multiple consultation retrieval data sets, each consultation retrieval data set includes multiple demand tags and multiple consultation retrieval orders of multiple consultations; matching the multiple consultation retrieval data sets according to the multiple consultation priorities to obtain multiple first matching consultation retrieval data sets whose consultation retrieval orders meet the multiple consultation priorities; calculating the first consultation retrieval fitness of the basic consultation retrieval scheme according to multiple user evaluation data sets of the multiple first matching consultation retrieval data sets, wherein each user evaluation data includes an evaluation level.

[0057] The specific configuration of the first consultation retrieval fitness calculation module 40 will be described in detail below. The first consultation retrieval fitness calculation module 40 further includes: constructing a consultation retrieval function for calculating the fitness of the consultation retrieval solution, as shown in the following formula:

[0058] ;

[0059] Among them, CSC is the consultation retrieval fitness, M is the number of all consultation needs in multiple first matching consultation retrieval data sets, is the weight assigned according to the consultation priority of the i-th consultation demand, is the user evaluation level in the user evaluation data of the i-th consulting demand, is the consultation priority of the ith consultation demand; according to the consultation retrieval function, combined with the consultation priorities of multiple consultation requests in the multiple first matching consultation retrieval data sets, and multiple user evaluation data sets, the first consultation retrieval fitness is calculated.

[0060] The specific configuration of the consultation retrieval scheme optimization module 50 will be described in detail below. The consultation retrieval scheme optimization module 50 may further include: randomly selecting P consultation priorities of P demand labels in the demand label library for random adjustment, obtaining P adjusted consultation priorities, and obtaining an adjusted basic consultation retrieval scheme, where P is an integer greater than 1 and less than the number of all demand labels; based on the adjusted basic consultation retrieval scheme, matching historical consultation data to calculate the consultation retrieval fitness, and obtaining a second consultation retrieval fitness; judging whether the second consultation retrieval fitness is greater than the first consultation retrieval fitness, if so, continuing to randomly adjust the P consultation priorities of the P demand labels, if not, taking the P demand labels as taboo labels, reselecting the consultation priorities of the other P demand labels for random adjustment and optimization, and unbanning the taboo labels after the number of taboos; continuing to optimize until convergence, and outputting the consultation retrieval scheme with the largest consultation retrieval fitness as the optimal consultation retrieval scheme.

[0061] The consultation and retrieval system based on demand tag configuration provided by the embodiment of the present invention can execute the consultation and retrieval method based on demand tag configuration provided by any embodiment of the present invention, and has the corresponding functional modules and beneficial effects of the execution method.

[0062] Although the present application makes various references to certain modules in the system according to the embodiments of the present application, any number of different modules may be used and run on the user terminal and / or server, and the various units and modules included are only divided according to functional logic, but are not limited to the above division, as long as the corresponding functions can be achieved; in addition, the specific names of the functional units are only for the convenience of distinguishing each other, and are not used to limit the scope of protection of the present invention.

[0063] The above specific implementations do not constitute a limitation on the protection scope of this application. It should be understood by those skilled in the art that various modifications, combinations and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions and improvements made within the spirit and principles of this application should be included in the protection scope of this application.

Claims

1. The consulting retrieval method based on demand tag configuration is characterized by: The method comprises: Based on semantic recognition, the consulting sentence library is identified and classified to obtain the consulting question library; Based on the emotion recognition mark and the customer value recognition mark, an emotion tag library and a customer tag library are obtained, and combined with the consulting question library to obtain a demand tag library, wherein the demand tag library includes a plurality of demand tags; Analyze the consultation retrieval priorities of the multiple demand tags to obtain multiple consultation priorities as a basic consultation retrieval solution; Acquiring historical user consultation record data, matching according to the multiple consultation priorities to obtain matching historical consultation data, and calculating and obtaining a first consultation retrieval fitness, specifically comprising: acquiring historical user consultation record data, the historical user consultation record data comprising multiple consultation retrieval data sets, each consultation retrieval data set comprising multiple demand tags and multiple consultation retrieval orders for multiple consultations; According to the plurality of consultation priorities, the plurality of consultation retrieval data sets are matched to obtain a plurality of first matching consultation retrieval data sets whose consultation retrieval order meets the plurality of consultation priorities; Calculating and obtaining the first consultation retrieval fitness of the basic consultation retrieval scheme according to the plurality of user evaluation data sets of the plurality of first matching consultation retrieval data sets, wherein each user evaluation data includes an evaluation grade; Randomly adjust multiple consultation priorities in the basic consultation retrieval scheme, match historical consultation data to calculate consultation retrieval fitness, optimize the basic consultation retrieval scheme until convergence, and obtain the optimal consultation retrieval scheme; According to the optimal consultation search solution, consultation search is performed on the consultation requirements configured with the multiple requirement tags.

2. The consultation retrieval method based on demand tag configuration according to claim 1 is characterized in that: Based on semantic recognition, the consulting statement library is identified and classified to obtain the consulting question library, including: Based on historical consultation data records, obtain a collection of historical consultation statements and build a consultation statement library; A set of sample consulting statements is obtained, and consulting questions in the sample consulting statements are identified to obtain a set of sample consulting questions; based on semantic recognition, the sample consulting statement set and the sample consulting question set are used to construct a consulting question identifier, the consulting statement library is identified, a plurality of consulting questions are obtained, and the consulting questions are classified to obtain a consulting question library.

3. The consultation retrieval method based on demand tag configuration according to claim 2 is characterized in that: Based on the emotion recognition mark and the customer value recognition mark, an emotion tag library and a customer tag library are obtained, which are combined with the consulting question library, including: Obtain a sample consultation audio set, and identify user emotions in the sample consultation audio, to obtain a sample user emotion set; Using the sample consultation audio set and the sample user emotion set, based on audio recognition, an emotion recognizer is constructed to recognize the consultation sentence library, obtain multiple user emotions, and classify them to obtain an emotion tag library; Build a user tag library based on user service categories of multiple users; The consulting question library, the emotion tag library and the customer tag library are combined to obtain the demand tag library.

4. The consultation retrieval method based on demand tag configuration according to claim 1 is characterized in that: The consultation retrieval priorities of the multiple demand tags are analyzed to obtain multiple consultation priorities, including: According to the order of the probability of occurrence of the consulting questions from large to small, the multiple demand tags in the demand tag library are sorted and prioritized to obtain multiple problem priorities, wherein the higher the sorting, the greater the problem priority; Sorting and assigning priorities to multiple demand tags in the demand tag library according to the order of occurrence probability of the emotion tags from small to large, to obtain multiple emotion priorities; Sorting and assigning priorities to the multiple demand tags in the demand tag library in descending order of customer value to obtain multiple user priorities; The multiple problem priorities, the multiple emotion priorities, and the multiple customer priorities are weightedly calculated to obtain the multiple consultation priorities.

5. The consultation retrieval method based on demand tag configuration according to claim 1 is characterized in that: Calculating and obtaining the first consultation retrieval fitness of the basic consultation retrieval scheme according to the plurality of user evaluation data sets of the plurality of first matching consultation retrieval data sets includes: Construct a consultation retrieval function to calculate the fitness of the consultation retrieval solution, as follows: ； Among them, CSC is the consultation retrieval fitness, M is the number of all consultation needs in multiple first matching consultation retrieval data sets, is the weight assigned according to the consultation priority of the i-th consultation demand, is the user evaluation level in the user evaluation data of the i-th consulting demand, is the consultation priority of the i-th consultation demand; According to the consultation retrieval function, combined with the consultation priorities of multiple consultation requests in the multiple first matching consultation retrieval data sets and multiple user evaluation data sets, a first consultation retrieval fitness is calculated.

6. The consultation retrieval method based on demand tag configuration according to claim 1 is characterized in that: The multiple consultation priorities in the basic consultation retrieval scheme are randomly adjusted, and the consultation retrieval fitness is calculated by matching the historical consultation data, so as to optimize the basic consultation retrieval scheme, including: Randomly select P consultation priorities of P demand tags in the demand tag library for random adjustment to obtain P adjusted consultation priorities and an adjusted basic consultation retrieval scheme, where P is an integer greater than 1 and less than the number of all demand tags; Based on the adjustment of the basic consultation retrieval scheme, the consultation retrieval fitness is calculated by matching the historical consultation data to obtain a second consultation retrieval fitness; Determine whether the second consultation retrieval fitness is greater than the first consultation retrieval fitness. If so, continue to randomly adjust the P consultation priorities of the P demand tags. If not, use the P demand tags as taboo tags, reselect the consultation priorities of the other P demand tags for random adjustment and optimization, and unban the taboo tags after the taboo times. The optimization is continued until convergence, and the consulting retrieval solution with the largest consulting retrieval fitness is output as the optimal consulting retrieval solution.

7. The consulting retrieval system based on demand tag configuration is characterized by: The system is used to implement the consultation and retrieval method based on demand tag configuration according to any one of claims 1 to 6, and the system includes: A semantic recognition and classification module, which is used to recognize and classify the consulting statement library based on semantic recognition to obtain a consulting question library; A demand label library acquisition module, the demand label library acquisition module is used to obtain an emotion label library and a customer label library based on the emotion recognition mark and the customer value recognition mark, and combine the consultation question library to obtain a demand label library, wherein the demand label library includes a plurality of demand labels; A retrieval priority analysis module, the retrieval priority analysis module is used to analyze the consultation retrieval priorities of the multiple demand tags to obtain multiple consultation priorities as a basic consultation retrieval solution; A first consultation retrieval fitness calculation module, the first consultation retrieval fitness calculation module is used to obtain historical user consultation record data, match according to the multiple consultation priorities, obtain matching historical consultation data, and calculate and obtain a first consultation retrieval fitness; A consultation retrieval scheme optimization module, which is used to randomly adjust multiple consultation priorities in the basic consultation retrieval scheme, match historical consultation data to calculate consultation retrieval fitness, optimize the basic consultation retrieval scheme until convergence, and obtain the optimal consultation retrieval scheme; A consultation retrieval module is used to perform consultation retrieval on the consultation requirements configured with the multiple requirement tags according to the optimal consultation retrieval solution.

Images