User classification methods, devices, computer equipment, storage media, and program products

Abstract

This application relates to a user classification method, user classification device, computer equipment, and storage medium. The method includes: acquiring keyword information from the user's voice text during a voice call between a customer service agent and a user regarding an ongoing financial transaction; the keyword information includes target keywords and their global weights; performing relevance matching between the keyword information and keyword information from multiple template texts in a pre-set corpus to obtain matching results between the keyword information and the keyword information of each template text; based on the matching results, identifying a target template text from the multiple template texts, and classifying the user according to the tag type of the target template text; the tag type represents the user's evaluation level of the financial transaction corresponding to the target template text. This method can improve the efficiency and accuracy of user classification.

Description

Technical Field

[0001] This application relates to the field of artificial intelligence technology, and in particular to a user classification method, user classification device, computer equipment, computer-readable storage medium, and computer program product. Background Technology

[0002] In the banking industry, telephone communication between customer service representatives and customers has become a widely used method. To improve the efficiency of telephone communication, it is necessary to categorize customers based on their call recordings.

[0003] Traditional user classification primarily categorizes users based on the type of financial service they communicate with (such as consultation-type financial services, telemarketing-type financial services, and after-sales-type financial services) and the content of the call. However, current user classification methods cannot intelligently and effectively differentiate user satisfaction with human customer service representatives, leading to inaccurate user classification and low efficiency in subsequent telephone communication. Summary of the Invention

[0004] Therefore, it is necessary to provide a user classification method, a user classification device, a computer device, a computer-readable storage medium, and a computer program product to address the aforementioned technical problems.

[0005] According to a first aspect of the present disclosure, a user classification method is provided, comprising:

[0006] When a customer service agent is having a voice call with a user regarding a financial transaction in progress, keyword information is obtained from the call text generated by the user. The keyword information includes target keywords and the global weight of the target keywords. The target keywords are used to express the main content of the call text, and the global weight represents the degree of contribution of the target keywords to expressing the main content.

[0007] The keyword information is matched with the keyword information of multiple template texts in a preset corpus to obtain the matching result between the keyword information and the keyword information of each template text.

[0008] Based on the matching results, a target template text is determined from the plurality of template texts, and the user is classified according to the tag type of the target template text; the tag type represents the user's evaluation level of the financial business corresponding to the target template text.

[0009] In one exemplary embodiment, obtaining keyword information from the user-generated call text includes:

[0010] The call text is segmented to obtain a word set for the call text;

[0011] Based on the prefixes of each word in the word set, each word is classified to obtain a prefix tree for the call text; the prefix tree is a word clustering tree, and each branch in the prefix tree corresponds to a class of words;

[0012] The prefix tree is scanned using a pre-defined dictionary to expand the words included in each branch of the prefix tree and generate a directed acyclic graph for each branch.

[0013] Based on the word frequency and inverse document frequency of each word after expansion in the directed acyclic graph, the keyword information in the call text is obtained.

[0014] In an exemplary embodiment, obtaining keyword information in the call text based on the term frequency and inverse document frequency of each word after expansion in the directed acyclic graph includes:

[0015] Based on the word frequency and the inverse document frequency, the global weight of each word in the expanded directed acyclic graph is determined;

[0016] Based on the global weight of each word, the words are sorted, and the target words ranked first and their corresponding global weights are used as keyword information in the call text.

[0017] In an exemplary embodiment, the step of performing relevance matching between the keyword information and the keyword information of multiple template texts in a preset corpus to obtain the matching result between the keyword information and the keyword information of each template text includes:

[0018] The target keywords in the keyword information are semantically matched with the corresponding template keywords in the template information to obtain a first matching degree; the semantic matching is used to match the degree of semantic similarity between two words; and

[0019] The global weight of the target keyword is matched with the global weight of the template keyword by distance to obtain a second matching degree; the distance matching is used to match the degree of difference between the global weights corresponding to the two words.

[0020] Based on the first matching degree and the second matching degree, the matching results between the keyword information and each of the template information are obtained.

[0021] In an exemplary embodiment, determining the target template text from the plurality of template texts based on the matching result includes:

[0022] If the matching result is based on the matching score representation, then among the multiple template texts, the candidate template text with the highest corresponding matching score is determined;

[0023] If the matching score of the candidate template text is greater than a preset score, the candidate template text is determined to be the target template text.

[0024] In an exemplary embodiment, determining the target template text from the plurality of template texts based on the matching result includes:

[0025] Emotion detection processing is performed on the audio of the voice call to obtain an emotion detection result for the audio; the emotion detection processing is used to evaluate the voice emotion expressed in the audio.

[0026] Based on the emotion detection results, candidate template texts are determined from the plurality of template texts;

[0027] Based on the matching results, the target template text is determined from the candidate template texts.

[0028] In one exemplary embodiment, after classifying the users, the method further includes:

[0029] Based on the user's user type, obtain the corresponding call strategy for the user type, and feed the call strategy back to the customer service agent;

[0030] The call strategy is used to instruct the customer service agent to execute the corresponding outbound call frequency and call content for the user.

[0031] In one exemplary embodiment, after classifying the users, the method further includes:

[0032] Based on the business type of the financial service, obtain the candidate call strategy corresponding to the business type;

[0033] Based on the user's user type, a target call strategy corresponding to the user type is determined from the candidate call strategies, and the target call strategy is fed back to the customer service agent.

[0034] According to a second aspect of the present disclosure, a user classification device is provided, comprising:

[0035] The information acquisition unit is configured to acquire keyword information from the call text generated by the user when a customer service agent conducts a voice call with the user regarding a financial transaction in progress. The keyword information includes target keywords and global weights of the target keywords. The target keywords are used to express the main content of the call text, and the global weights represent the degree of contribution of the target keywords to expressing the main content.

[0036] The information matching unit is configured to perform relevance matching between the keyword information and the keyword information of multiple template texts in a preset corpus, and obtain the matching result between the keyword information and the keyword information of each template text;

[0037] The user classification unit is configured to perform actions based on the matching results, to determine a target template text from the plurality of template texts, and to classify the user according to the tag type of the target template text; the tag type characterizes the user's evaluation level of the financial business corresponding to the target template text.

[0038] According to a third aspect of the present disclosure, a computer device is provided, comprising:

[0039] processor;

[0040] Memory for storing the executable instructions of the processor;

[0041] The processor is configured to execute the executable instructions to implement the user classification method as described in any of the preceding claims.

[0042] According to a fourth aspect of the present disclosure, a computer-readable storage medium is provided, the computer-readable storage medium including a computer program that, when executed by a processor of a computer device, enables the computer device to perform the user classification method as described in any of the preceding claims.

[0043] According to a fifth aspect of the present disclosure, a computer program product is provided, the computer program product including program instructions that, when executed by a processor of a computer device, enable the computer device to perform the user classification method as described in any of the preceding claims.

[0044] The technical solutions provided by the embodiments of this disclosure have at least the following beneficial effects:

[0045] This method first acquires keyword information from the user's voice call text during a financial transaction being processed. The keyword information includes target keywords and their global weights. Target keywords express the main content of the call text, and global weights represent the degree to which target keywords contribute to expressing the main content. The keyword information is then matched with the keyword information of multiple template texts in a pre-defined corpus to obtain matching results. Based on the matching results, a target template text is identified from the multiple template texts, and users are categorized according to the tag type of the target template text. The tag type represents the user's evaluation level of the financial transaction corresponding to the target template text. In this way, on the one hand, by acquiring the text of voice calls, the keywords and global weights of the user's corresponding calls are obtained, and then the target template text of the corresponding tag type is matched to classify users, thereby optimizing the user classification process, improving the efficiency of user classification, and reducing the consumption of manpower and time costs; on the other hand, users are classified according to their evaluation of the financial business corresponding to the target template text, thereby more effectively grasping the user's evaluation of the financial business, improving the accuracy of user classification, and making subsequent voice calls with users based on their category more efficient.

[0046] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and are not intended to limit this disclosure. Attached Figure Description

[0047] Figure 1 This is an application environment diagram illustrating a user classification method according to an exemplary embodiment;

[0048] Figure 2 This is a flowchart illustrating a user classification method according to an exemplary embodiment;

[0049] Figure 3 This is a flowchart illustrating a step of obtaining keyword information according to an exemplary embodiment;

[0050] Figure 4 This is a flowchart illustrating a step for obtaining keyword information according to an exemplary embodiment;

[0051] Figure 5 This is a flowchart illustrating a step of determining target template text according to an exemplary embodiment;

[0052] Figure 6 This is a flowchart illustrating a user classification method according to another exemplary embodiment;

[0053] Figure 7 This is a schematic diagram of a user classification method according to another exemplary embodiment;

[0054] Figure 8 This is a block diagram illustrating a user classification device according to an exemplary embodiment;

[0055] Figure 9 This is a block diagram illustrating a computer device for user classification according to an exemplary embodiment. Detailed Implementation

[0056] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0057] The terms "first," "second," etc., used in this application are used to distinguish different objects, not to describe a specific order. Furthermore, although the terms "first," "second," etc., are used repeatedly to describe various operations (or various thresholds, or various applications, or various instructions, or various elements), these operations (or thresholds, or applications, or instructions, or elements) should not be limited by these terms. These terms are only used to distinguish one operation (or threshold, or application, or instruction, or element) from another operation (or threshold, or application, or instruction, or element).

[0058] It should be noted that if the technical solution of this application involves personal information, the user must be clearly informed of the rules for processing personal information and their consent obtained before the product corresponding to the technical solution of this application processes the user's personal information. If the technical solution of this application involves sensitive personal information of the user, the user's individual consent must be obtained before the product corresponding to the technical solution of this application processes the user's sensitive personal information, and the requirement of "express consent" must also be met. For example, on the personal information processing device, the personal information processing rules can be clearly indicated by signage / information, or the user's personal authorization can be obtained through pop-up messages or by requesting the individual to upload their personal information. The personal information processing rules may include information such as the personal information processor, the purpose of personal information processing, the processing method, and the types of personal information processed.

[0059] The user classification method provided in this application embodiment can be applied to, for example... Figure 1 In the application environment shown, terminal 102 communicates with server 104 via a communication network. A data storage system can store the data that server 104 needs to process. The data storage system can be integrated onto server 104, or it can be located in the cloud or on other network servers.

[0060] In some embodiments, reference Figure 1 First, when a customer service agent is conducting a voice call with a user regarding an ongoing financial transaction, server 104 acquires keyword information from the user's generated call text. This keyword information includes target keywords and their global weights. The target keywords express the main content of the call text, and the global weights represent the degree to which the target keywords contribute to expressing this main content. Then, server 104 performs relevance matching between the keyword information and the keyword information of multiple template texts in a pre-defined corpus, obtaining matching results between the keyword information and the keyword information of each template text. Finally, based on the matching results, server 104 identifies the target template text among the multiple template texts and categorizes the user according to the tag type of the target template text. The tag type represents the user's evaluation level of the financial transaction corresponding to the target template text.

[0061] In some embodiments, terminal 102 (such as a mobile terminal or a fixed terminal) can be implemented in various forms. Terminal 102 can be a mobile terminal including mobile phones, smartphones, laptops, portable handheld devices, personal digital assistants (PDAs), tablet computers (PADs), etc., or a fixed terminal including automated teller machines (ATMs), automated kiosks, digital TVs, desktop computers, fixed-line computers, etc.

[0062] Hereinafter, it is assumed that terminal 102 is a fixed terminal. However, those skilled in the art will understand that, if there are operations or elements specifically designed for mobile purposes, the construction according to the embodiments disclosed in this application can also be applied to mobile type terminal 102.

[0063] In some embodiments, the data processing component running on server 104 may load any of the various additional server applications and / or middleware applications being executed, such as HTTP (Hypertext Transfer Protocol), FTP (File Transfer Protocol), CGI (Common Gateway Interface), RDBMS (Relational Database Management System), etc.

[0064] In some embodiments, server 104 may be implemented using a standalone server or a server cluster consisting of multiple servers. Server 104 may be adapted to run one or more application services or software components that provide the terminal 102 described in the foregoing disclosure.

[0065] In some embodiments, the application service may include a service interface that provides user classification to users, as well as corresponding program services, etc. The software component may include an application (SDK) or client (APP) that performs the user classification function.

[0066] In some embodiments, the application or client with user classification function provided by server 104 includes a portal port that provides one-to-one application services to users in the foreground and multiple business systems that perform data processing in the background, so as to extend the user classification function application to the APP or client, so that users can use and access the user classification function at any time and any place.

[0067] In some embodiments, users can input corresponding code data or control parameters into the APP or client through a preset input device or automatic control program to execute application services of the computer program in the server 104 and display application services in the user interface.

[0068] In some embodiments, the operating system running the app or client may include various versions of Microsoft... Apple and / or Linux operating system, various commercial or similar Operating systems (including but not limited to various GNU / Linux operating systems, Google) OS and / or mobile operating systems, such as Phone OS OS OS operating systems, as well as other online or offline operating systems, are not specifically limited here.

[0069] In some embodiments, such as Figure 2 As shown, a user classification method is provided, which is applied to... Figure 1 Taking server 104 as an example, the method includes the following steps:

[0070] Step S11: When a customer service agent is having a voice call with a user regarding a financial transaction in progress, obtain keyword information from the user's generated call text.

[0071] In one embodiment, the customer seat is a customer service representative in the banking industry, and the financial business is the business in the banking industry that requires voice calls with bank customers, such as consultation business between bank customers (i.e., users) and customer service representatives, anti-fraud business between bank customers and customer service representatives, etc.

[0072] In one embodiment, when a customer service representative makes a voice call with a user regarding a financial transaction, the server captures the user's voice audio in real time and extracts keyword information from the corresponding call text.

[0073] In one embodiment, the keyword information includes target keywords and the global weights of the target keywords.

[0074] In one embodiment, target keywords are used to express the main content of the call text. Global weights characterize the degree to which target keywords contribute to expressing the main content.

[0075] As an example, a customer service representative conducts a voice call with a user regarding a financial product. The server extracts the following keywords from the call text: "Fund X" with a global weight of 80%, "recently" with a global weight of 40%, "three months" with a global weight of 40%, "increase" with a global weight of 50%, and "amplitude" with a global weight of 60%. Therefore, it can be determined that the user is inquiring about the value increase of Fund X over the past three months.

[0076] Step S12: Perform relevance matching between the keyword information and the keyword information of multiple template texts in the preset corpus to obtain the matching results between the keyword information and the keyword information of each template text.

[0077] In one embodiment, the server obtains the matching results between keyword information and keyword information of each template text, including the following steps:

[0078] Step 1: Perform semantic matching between the target keywords in the keyword information and the corresponding template keywords in the template information to obtain the first matching degree.

[0079] Among them, semantic matching is used to match the degree of semantic similarity between two words.

[0080] As an example, the target keyword in the call text is "purchase", and the context of the target keyword indicates that the user "wants to purchase fund product A". The keyword in the template text S1 is also "purchase", and the context of the target keyword indicates that the user corresponding to the template text S1 "wants to purchase fund product B". Therefore, the first degree of matching between the target keyword in the call text and the keyword in the template text S1 is the same, which can be represented by the similarity as the first degree of matching being 100%.

[0081] As another example, the target keyword in the call text is "purchase", and the context of the target keyword indicates that the user "wants to buy fund product A". In the template text S2, the keyword is "understand", and the context of the target keyword indicates that the user corresponding to the template text S2 "wants to understand fund product B". Therefore, the first matching degree of the target keyword in the call text and the keyword in the template text S2 is close, which can be represented by the similarity as 80%.

[0082] Step 2: Perform distance matching between the global weight of the target keyword and the global weight of the template keyword to obtain the second matching degree.

[0083] Distance matching is used to match the degree of difference between the global weights corresponding to two words.

[0084] As an example, the global weight of the target keyword "purchase" in the call text is 50%, the global weight of the keyword "purchase" in template text S1 is 60%, and the global weight of the keyword "understand" in template text S2 is 50%. Therefore, the second matching degree between the target keyword in the call text and the keyword in template text S1 is close, which can be represented by a second matching degree of 90% through similarity. The second matching degree between the target keyword in the call text and the keyword in template text S2 is also close, which can be represented by a second matching degree of 100% through similarity.

[0085] There is no specific restriction on the order of steps one and two; that is, step one can be executed before step two, step two can be executed before step one, or steps one and two can be executed simultaneously.

[0086] Step 3: Based on the first and second matching degrees, obtain the matching results between keyword information and each template information.

[0087] In one embodiment, the server performs a weighted sum of the first matching degree and the second matching degree according to a preset weight coefficient to obtain the matching result between the keyword information and each template information.

[0088] Step S13: Based on the matching results, determine the target template text among multiple template texts.

[0089] In one embodiment, the server determines the target template text, specifically including the following steps:

[0090] Step 1: If the matching result is based on the matching score, then among multiple template texts, determine the candidate template text with the highest corresponding matching score.

[0091] Step 2: If the matching score of the candidate template text is greater than the preset score, the candidate template text is determined as the target template text.

[0092] As an example, for the call text, template text S1 scores 90 points, template text S2 scores 70 points, template text S3 scores 80 points, template text S4 scores 60 points, and template text S5 scores 95 points. Therefore, the server considers template text S5 as a candidate template text among all the template texts, and since the preset score is 80 points (less than 95 points), the server selects template text S5 as the target template text.

[0093] Step S14: Classify users according to the tag type of the target template text.

[0094] In one embodiment, the tag type characterizes the user's evaluation level of the financial business corresponding to the target template text.

[0095] The text is tagged with three types: high-intent text, medium-intent text, and low-intent text. High-intent text corresponds to a high level of evaluation from users who have a high evaluation of the financial service. Medium-intent text corresponds to a moderate level of evaluation from users who have a medium evaluation of the financial service. Low-intent text corresponds to a low level of evaluation from users who have a low evaluation of the financial service.

[0096] Furthermore, the server categorizes users based on the text's tag type: high-intent users (indicating a high level of interest in the financial services related to the call text), moderate-intent users (indicating a moderate level of interest in the financial services related to the call text), and low-intent users (indicating a low level of interest in the financial services related to the call text).

[0097] In one embodiment, after classifying users, the server further includes the following steps: obtaining a call policy corresponding to the user type based on the user type, and feeding the call policy back to the customer service agent.

[0098] In one example, each user type has a pre-defined call strategy for its customer service representatives. After determining the user type, the next time a customer service representative calls that user, the representative will conduct a voice call with the user according to the corresponding call strategy.

[0099] The call strategy is used to instruct customer service agents to execute the corresponding outbound call frequency and call content for the user.

[0100] As an example, for high-intent users, the calling strategy is A, with an outbound call frequency of three times per week and the call content is content A1; for general-intent users, the calling strategy is B, with an outbound call frequency of twice per week and the call content is content B1; and for high-intent users, the calling strategy is C, with an outbound call frequency of once per week and the call content is content C1.

[0101] In the aforementioned user classification process, the server first obtains keyword information from the user's voice call text during a customer service conversation regarding an ongoing financial transaction. This keyword information includes target keywords and their global weights. Target keywords express the main content of the call text, while global weights represent the degree to which target keywords contribute to expressing this main content. The server then performs relevance matching between the keyword information and the keyword information of multiple template texts in a pre-defined corpus, obtaining matching results. Based on these matching results, a target template text is identified from the multiple template texts, and users are classified according to the tag type of the target template text. The tag type represents the user's evaluation level of the financial transaction corresponding to the target template text. In this way, on the one hand, by acquiring the text of voice calls, the keywords and global weights of the user's corresponding calls are obtained, and then the target template text of the corresponding tag type is matched to classify users, thereby optimizing the user classification process, improving the efficiency of user classification, and reducing the consumption of manpower and time costs; on the other hand, users are classified according to their evaluation of the financial business corresponding to the target template text, thereby more effectively grasping the user's evaluation of the financial business, improving the accuracy of user classification, and making subsequent voice calls with users based on their category more efficient.

[0102] Those skilled in the art will understand that the methods disclosed in the above-described specific embodiments can be implemented in more concrete ways. For example, the implementation described above of a server classifying users based on the tag type of the target template text is merely illustrative.

[0103] For example, the server obtains keyword information from the user-generated call text; or the server performs correlation matching between the keyword information and the keyword information of multiple template texts in a preset corpus, etc. These are just one way of setting up the data. In actual implementation, there may be other ways of dividing the data. For example, the keyword information of the call text and the keyword information of the template text can be combined or set into another system, or some features can be ignored or not executed.

[0104] In one exemplary embodiment, see Figure 3 , Figure 3 This is a schematic flowchart illustrating an embodiment of obtaining keyword information in this application. In step S11, the process by which the server obtains keyword information from the user-generated call text can be implemented in the following way:

[0105] Step S111: Perform word segmentation on the call text to obtain a word set for the call text.

[0106] In one embodiment, the server performs word segmentation on the call text based on a preset word segmentation tool to separate the individual words in the call text and obtain a word set for the call text.

[0107] Step S112: Based on the prefixes of each word in the word set, classify each word to obtain a prefix tree for the call text.

[0108] In one embodiment, the prefix tree is represented as a trie tree model, which is a word clustering tree, and each branch in the prefix tree corresponds to a class of words.

[0109] Specifically, the server categorizes and aggregates words with the same prefix in the word set to obtain word classes, and then puts the words in each class into a prefix tree to form a prefix tree with multiple branches.

[0110] Step S113: Perform a word graph scan on the prefix tree based on a preset dictionary to expand the words included in each branch of the prefix tree and generate a directed acyclic graph for each branch.

[0111] In one embodiment, a word graph scan is performed on the trie tree based on a preset dictionary. This involves placing words from the dictionary into branches of the trie tree. If the first few words of each word in each branch are the same, it means that they have the same prefix. This expands each branch to perform fast word lookup and generate a directed acyclic graph for each branch.

[0112] Step S114: Based on the word frequency and inverse document frequency of each word after expansion in the directed acyclic graph, obtain the keyword information in the call text.

[0113] In one embodiment, dynamic programming is used to find the maximum probability path for each directed acyclic graph (DAG) to identify the maximum partitioning combination (i.e., the target keyword and its global weight). For example, only one of the keywords "stock" and "fund" can be chosen as the target keyword, and the specific selection must be considered from an overall perspective to find a globally optimal solution. The specific method is a dynamic programming algorithm.

[0114] In one exemplary embodiment, see Figure 4 , Figure 4This is a schematic flowchart illustrating an embodiment of obtaining keyword information in this application. In step S114, the process by which the server obtains keyword information from the call text based on the term frequency and inverse document frequency of each word after expansion in the directed acyclic graph, can be implemented in the following way:

[0115] Step a1: Determine the global weight of each word in the extended directed acyclic graph based on word frequency and inverse document frequency.

[0116] Specifically, the TF-IDF algorithm is used for word frequency and inverse document frequency to convert each word into a weight vector.

[0117] In one embodiment, Term Frequency (TF) = the number of times a word appears in the text, or TF = the number of times a word appears in the text / the total number of words in the text. IDF weight, or Inverse Document Frequency, is simulated using a pre-defined corpus to represent the language's usage environment. The larger the IDF, the more concentrated the distribution of this feature in the text, and the stronger the ability of the segmented word to distinguish the text's content attributes. Specifically, Inverse Document Frequency (IDF) = log(total number of texts in the corpus / number of texts containing the word + 1).

[0118] In one embodiment, the TF-IDF value (Term Frequency Document Frequency) is TF-IDF = Term Frequency (TF) × Inverse Document Frequency (IDF).

[0119] Step a2: Based on the global weight of each word, sort the words and use the target words in the preset positions and their corresponding global weights as keyword information in the call text.

[0120] In one embodiment, actual analysis shows that the TF-IDF value is directly proportional to the frequency of a word in the text; the higher the importance of a word in the text, the larger its TF-IDF value. The TF-IDF value of each word in the text is calculated and sorted, and the target word with the highest frequency in a preset position is the target keyword to be extracted.

[0121] In one exemplary embodiment, see Figure 5 , Figure 5 This is a flowchart illustrating an embodiment of determining the target template text in this application. In step S13, the process by which the server determines the target template text from multiple template texts based on the matching results can be implemented in the following way:

[0122] Step S131: Perform emotion detection processing on the audio of the voice call to obtain the emotion detection results for the audio.

[0123] Among them, emotion detection processing is used to evaluate the emotional state expressed in the audio of a call.

[0124] As an example, the server inputs the audio features of the call speech into a pre-trained audio quality inspection model. This model performs quality checks on the call speech to assess the emotion conveyed, obtaining a corresponding emotion detection value, which is then used as the emotion detection result. The emotion detection value is directly proportional to the quality assessment of the speech emotion; that is, a higher emotion detection value indicates a better expression of emotion, while a lower value indicates a worse expression of emotion.

[0125] Step S132: Based on the emotion detection results, candidate template texts are identified from multiple template texts.

[0126] In one embodiment, the emotion detection result is represented based on the emotion detection value. The server selects a preset number of candidate template texts based on the emotion detection value corresponding to each template text.

[0127] Step S133: Based on the matching results, determine the target template text from the candidate template texts.

[0128] In one embodiment, the matching result between the keyword information of the call text and the keyword information of each template text is represented by the matching value. The server selects a preset number of target template texts based on the matching value corresponding to each candidate template text.

[0129] In one exemplary embodiment, after step S14, that is, after the server has categorized the users, it can also be implemented in the following way:

[0130] Step 1: Based on the business type of the financial business, obtain the candidate call strategies corresponding to the business type.

[0131] In one embodiment, the server extracts call policies of the same type as the financial business from a preset policy library as candidate call policies.

[0132] Step 2: Based on the user's user type, determine the target call strategy corresponding to the user type from the candidate call strategies, and feed the target call strategy back to the customer service agent.

[0133] In one embodiment, the server selects the call strategy that is the same as the call strategy type corresponding to the user type from among the candidate call strategies as the target call strategy.

[0134] To more clearly illustrate the user classification method provided in this disclosure, a specific embodiment is described below. In an exemplary embodiment, reference is made to... Figure 6 and Figure 7 , Figure 6 This is a flowchart illustrating a user classification method according to another exemplary embodiment. Figure 7 This is a block diagram illustrating a user classification method according to another exemplary embodiment. The user classification method is used in a server and specifically includes the following:

[0135] Step S21: Obtain the target conversation text from the most recent marketing call.

[0136] Step S22: Preprocess the target dialogue text using a pre-trained NLP (Natural Language Processing) model to obtain the target text features of the target dialogue text.

[0137] The preprocessing includes the following steps:

[0138] Step ①: Use the preset word segmentation tool to segment the text to be processed, and obtain the word set of natural language in the text to be processed.

[0139] Step 2: Based on the word set, construct a trie tree for the text to be processed.

[0140] In this context, the prefix tree is a tree structure of a set of words, and the words in each branch of the tree structure have the same first few words, that is, the words in each branch have the same prefix.

[0141] The prefix tree construction process is as follows: classify each word in the word set according to its prefix to obtain the classified word groups; use each word group to construct the branches of the trie tree structure to obtain the word tree.

[0142] Step 3: By comparing the prefix tree with the preset dictionary, perform a word graph scan to generate a directed acyclic graph consisting of all possible word combinations of the characters in the word set.

[0143] Among them, word graph scanning is used to expand the words in the prefix tree to obtain a directed acyclic graph that represents multiple words.

[0144] Step 4: Extract keywords from each word in the directed acyclic graph based on the TF-IDF algorithm to obtain a preset number of target keywords and the importance of the target keywords in the text to be processed.

[0145] Step 5: Based on the target keywords and their corresponding importance, generate a feature matrix corresponding to each target keyword, and use the feature matrix as the target text feature of the target dialogue text.

[0146] Step S23: Perform similarity matching between the target text features and the historical text features of multiple historical dialogue texts in the preset corpus to obtain the matching results.

[0147] Among them, multiple historical dialogue texts in the pre-set corpus are used to train the NLP model.

[0148] Similarity matching includes performing similarity matching on the feature matrices corresponding to the two text features.

[0149] Similarity matching includes keyword matching and importance matching in text features.

[0150] The matching results include the similarity between the target dialogue text and each historical dialogue text.

[0151] Step S24: Among all historical dialogue texts, identify the most similar dialogue text to the target dialogue text.

[0152] Step S25: Based on the type of similar dialogue text, classify the target dialogue text into dialogue text of the same type as the similar dialogue text.

[0153] The types of dialogue texts include high intention type, medium intention type and low intention type.

[0154] If the similarity between the target dialogue text and similar dialogue texts is less than a preset threshold, the target dialogue text will not be classified, and it will be added to the corpus to update the NLP model.

[0155] Step S26: Based on the type of the target dialogue text, adopt the corresponding outbound marketing strategy for the target customer corresponding to the target dialogue text.

[0156] Different outbound marketing strategies include different outbound call frequencies and marketing scripts. For example, if the target dialogue text is of a high-intent type and the target customer is a high-intent customer, then marketing strategy A is adopted; if the target dialogue text is of a general-intent type and the target customer is a general-intent customer, then marketing strategy B is adopted; if the target dialogue text is of a low-intent type and the target customer is a low-intent customer, then marketing strategy C is adopted.

[0157] Specifically, (1) for high-intent customers, the frequency of outbound marketing calls and the types of marketing products can be appropriately increased, and subsequent marketing messages to customers can be delivered in a low-profile and highly friendly manner to maintain the activity of high-intent customers; (2) for general-intent customers, marketing messages can be optimized to establish direction and focus with customers, ensure that more effective information is conveyed during the marketing period, make full use of time, and keep customers in sync with us; inquiry methods, hypothetical methods, and other methods can also be used to address customers' doubts and considerations and guide customers to make decisions; (3) for low-intent customers, the marketing approach can be optimized based on the matching rate between the customer's multiple marketing call texts and the model samples. If the recent marketing texts are developing in a general or intentional direction, the marketing approach for general-intent customers can be adopted; if the matching rate with low-intent samples is constantly increasing, or there is obvious rejection, the marketing frequency can be appropriately reduced.

[0158] In this way, on the one hand, by acquiring the text of voice calls, the keywords and global weights of the user's corresponding calls are obtained, and then the target template text of the corresponding tag type is matched to classify users, thereby optimizing the user classification process, improving the efficiency of user classification, and reducing the consumption of manpower and time costs; on the other hand, users are classified according to their evaluation of the financial business corresponding to the target template text, thereby more effectively grasping the user's evaluation of the financial business, improving the accuracy of user classification, and making subsequent voice calls with users based on their category more efficient.

[0159] It should be understood that, although Figures 2-7 The steps in the flowchart are shown sequentially as indicated by the arrows, but these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order in which these steps are executed, and they can be performed in other orders. Figures 2-7 At least some of the steps in the process may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but may be executed at different times. The execution order of these steps or stages is not necessarily sequential, but may be executed in turn or alternately with other steps or at least some of the steps or stages in other steps.

[0160] It is understood that the same / similar parts between the various embodiments of the methods described above in this specification can be referred to each other. Each embodiment focuses on the differences from other embodiments, and relevant parts can be referred to the description of other method embodiments.

[0161] Figure 8 This is a block diagram of a user classification device provided in an embodiment of this application. (Refer to...) Figure 8The user classification device 20 includes: an information acquisition unit 21, an information matching unit 22, and a user classification unit 23.

[0162] The information acquisition unit 21 is configured to acquire keyword information from the call text generated by the user when a customer service agent is conducting a voice call with the user regarding a financial transaction in progress. The keyword information includes target keywords and the global weight of the target keywords. The target keywords are used to express the main content of the call text, and the global weight represents the degree of contribution of the target keywords to expressing the main content.

[0163] The information matching unit 22 is configured to perform relevance matching between the keyword information and the keyword information of multiple template texts in a preset corpus, and obtain the matching result between the keyword information and the keyword information of each template text.

[0164] The user classification unit 23 is configured to determine the target template text from the plurality of template texts based on the matching results, and classify the user according to the tag type of the target template text; the tag type represents the user's evaluation level of the financial business corresponding to the target template text.

[0165] In one exemplary embodiment, obtaining keyword information from the user-generated call text includes:

[0166] The call text is segmented to obtain a word set for the call text;

[0167] Based on the prefixes of each word in the word set, each word is classified to obtain a prefix tree for the call text; the prefix tree is a word clustering tree, and each branch in the prefix tree corresponds to a class of words;

[0168] The prefix tree is scanned using a pre-defined dictionary to expand the words included in each branch of the prefix tree and generate a directed acyclic graph for each branch.

[0169] Based on the word frequency and inverse document frequency of each word after expansion in the directed acyclic graph, the keyword information in the call text is obtained.

[0170] In an exemplary embodiment, obtaining keyword information in the call text based on the term frequency and inverse document frequency of each word after expansion in the directed acyclic graph includes:

[0171] Based on the word frequency and the inverse document frequency, the global weight of each word in the expanded directed acyclic graph is determined;

[0172] Based on the global weight of each word, the words are sorted, and the target words ranked first and their corresponding global weights are used as keyword information in the call text.

[0173] In an exemplary embodiment, the step of performing relevance matching between the keyword information and the keyword information of multiple template texts in a preset corpus to obtain the matching result between the keyword information and the keyword information of each template text includes:

[0174] The target keywords in the keyword information are semantically matched with the corresponding template keywords in the template information to obtain a first matching degree; the semantic matching is used to match the degree of semantic similarity between two words; and

[0175] The global weight of the target keyword is matched with the global weight of the template keyword by distance to obtain a second matching degree; the distance matching is used to match the degree of difference between the global weights corresponding to the two words.

[0176] Based on the first matching degree and the second matching degree, the matching results between the keyword information and each of the template information are obtained.

[0177] In an exemplary embodiment, determining the target template text from the plurality of template texts based on the matching result includes:

[0178] If the matching result is based on the matching score representation, then among the multiple template texts, the candidate template text with the highest corresponding matching score is determined;

[0179] If the matching score of the candidate template text is greater than a preset score, the candidate template text is determined to be the target template text.

[0180] In an exemplary embodiment, determining the target template text from the plurality of template texts based on the matching result includes:

[0181] Emotion detection processing is performed on the audio of the voice call to obtain an emotion detection result for the audio; the emotion detection processing is used to evaluate the voice emotion expressed in the audio.

[0182] Based on the emotion detection results, candidate template texts are determined from the plurality of template texts;

[0183] Based on the matching results, the target template text is determined from the candidate template texts.

[0184] In one exemplary embodiment, after classifying the users, the method further includes:

[0185] Based on the user's user type, obtain the corresponding call strategy for the user type, and feed the call strategy back to the customer service agent;

[0186] The call strategy is used to instruct the customer service agent to execute the corresponding outbound call frequency and call content for the user.

[0187] In one exemplary embodiment, after classifying the users, the method further includes:

[0188] Based on the business type of the financial service, obtain the candidate call strategy corresponding to the business type;

[0189] Based on the user's user type, a target call strategy corresponding to the user type is determined from the candidate call strategies, and the target call strategy is fed back to the customer service agent.

[0190] Figure 9 This is a block diagram of a computer device 30 provided in an embodiment of this application. For example, the computer device 30 can be an electronic device, an electronic component, or a server array, etc. (Refer to...) Figure 9 The computer device 30 includes a processor 31, which may be a collection of processors, including one or more processors. The computer device 30 also includes memory resources represented by memory 32, on which computer programs, such as application programs, are stored. The computer programs stored in memory 32 may include one or more modules, each corresponding to a set of executable instructions. Furthermore, the processor 31 is configured to implement the user classification method described above when executing the executable instructions.

[0191] In some embodiments, computer device 30 is an electronic device whose computing system can run one or more operating systems, including any operating systems discussed above and any commercially available server operating system. Computer device 30 can also run any of a variety of additional server applications and / or middleware applications, including HTTP (Hypertext Transfer Protocol) servers, FTP (File Transfer Protocol) servers, CGI (Common Gateway Interface) servers, super servers, database servers, etc. Exemplary database servers include, but are not limited to, commercially available database servers from companies such as IBM.

[0192] In some embodiments, the processor 31 typically controls the overall operation of the computer device 30, such as operations associated with display, data processing, data communication, and recording operations. The processor 31 may include one or more processor components to execute computer programs to perform all or part of the steps of the methods described above. Furthermore, the processor components may include one or more modules to facilitate interaction between the processor components and other components. For example, the processor components may include a multimedia module to facilitate control of the interaction between the user computer device 30 and the processor 31 using multimedia components.

[0193] In some embodiments, the processor component in processor 31 may also be referred to as a CPU (Central Processing Unit). The processor component may be an electronic chip with signal processing capabilities. The processor may also be a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. A general-purpose processor may be a microprocessor or any conventional processor component. Furthermore, the processor component may be implemented using integrated circuit chips.

[0194] In some embodiments, memory 32 is configured to store various types of data to support operation on computer device 30. Examples of such data include instructions for any application or method operating on computer device 30, acquired data, messages, images, videos, etc. Memory 32 can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk, optical disk, or graphene storage.

[0195] In some embodiments, the memory 32 can be a memory module, TF card, etc., and can store all information in the computer device 30, including the input raw data, computer programs, intermediate running results, and final running results. In some embodiments, it stores and retrieves information according to the location specified by the processor. In some embodiments, the computer device 30 has a memory function and can ensure normal operation because of the memory 32. In some embodiments, the memory 32 of the computer device 30 can be classified into main memory (RAM) and auxiliary memory (external memory) according to its purpose, or it can be classified into external memory and internal memory. External memory is usually magnetic media or optical discs, which can store information for a long time. RAM refers to the storage component on the motherboard, which is used to store the currently executing data and programs, but it is only used to temporarily store programs and data. The data will be lost when the power is turned off or disconnected.

[0196] In some embodiments, the computer device 30 may further include: a power supply component 33 configured to perform power management of the computer device 30, a wired or wireless network interface 34 configured to connect the computer device 30 to a network, and an input / output (I / O) interface 35. The computer device 30 may operate on an operating system stored in memory 32, such as Windows Server, Mac OS X, Unix, Linux, FreeBSD, or similar.

[0197] In some embodiments, power supply component 33 provides power to various components of computer device 30. Power supply component 33 may include a power management system, one or more power supplies, and other components associated with generating, managing, and distributing power to computer device 30.

[0198] In some embodiments, the wired or wireless network interface 34 is configured to facilitate wired or wireless communication between the computer device 30 and other devices. The computer device 30 may access wireless networks based on communication standards, such as WiFi, carrier networks (such as 2G, 3G, 4G, or 5G), or combinations thereof.

[0199] In some embodiments, the wired or wireless network interface 34 receives broadcast signals or broadcast-related information from an external broadcast management system via a broadcast channel. In one exemplary embodiment, the wired or wireless network interface 34 also includes a near-field communication (NFC) module to facilitate short-range communication. For example, the NFC module may be implemented based on radio frequency identification (RFID) technology, Infrared Data Association (IrDA) technology, ultra-wideband (UWB) technology, Bluetooth (BT) technology, and other technologies.

[0200] In some embodiments, the input / output (I / O) interface 35 provides an interface between the processor 31 and peripheral interface modules, such as a keyboard, click wheel, buttons, etc. These buttons may include, but are not limited to, a home button, volume buttons, a power button, and a lock button.

[0201] This application provides a block diagram of a computer-readable storage medium. The computer-readable storage medium stores a computer program, which, when executed by a processor, implements the user classification method described above.

[0202] If the integrated units of the various functional units in the various embodiments of this application are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. The computer-readable storage medium includes a number of instructions in a computer program to cause a computer device (which may be a personal computer, system server, or network device, etc.), an electronic device (e.g., MP3, MP4, etc., or a mobile phone, tablet computer, wearable device, etc., or a desktop computer, etc.) or a processor to execute all or part of the steps of the methods of the various embodiments of this application.

[0203] This application provides a block diagram of a computer program product. The computer program product includes program instructions that can be executed by a server's processor to implement the user classification method described above.

[0204] Those skilled in the art will understand that embodiments of this application may provide user classification methods, user classification devices 20, computer equipment 30, computer-readable storage media, or computer program products. Therefore, this application may take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application may take the form of a computer program product embodied on one or more computer program instructions (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0205] This application is described with reference to flowchart illustrations and / or block diagrams of user classification methods, user classification devices 20, computer devices 30, computer-readable storage media, or computer program products according to embodiments of this application. It should be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by the computer program product. These computer program products can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing device to produce a machine, such that program instructions executable by the processor of the computer or other programmable data processing device generate instructions for implementing the user classification method in the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0206] These computer program products may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that program instructions stored in the computer program product produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0207] These program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process, thereby providing program instructions that execute on the computer or other programmable apparatus for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0208] It should be noted that the various methods, apparatuses, electronic devices, computer-readable storage media, computer program products, etc. described above may also include other implementation methods according to the description of the method embodiments. For specific implementation methods, please refer to the description of the relevant method embodiments, which will not be elaborated here.

[0209] Other embodiments of this disclosure will readily occur to those skilled in the art upon consideration of the specification and practice of the invention disclosed herein. This disclosure is intended to cover any variations, uses, or adaptations of this disclosure that follow the general principles of this disclosure and include common knowledge or customary techniques in the art not disclosed herein. The specification and examples are to be considered exemplary only, and the true scope and spirit of this disclosure are indicated by the claims.

[0210] It should be understood that this disclosure is not limited to the precise structures described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of this disclosure is limited only by the appended claims.

Claims

1. A user classification method, characterized in that, The method includes: When a customer service agent is having a voice call with a user regarding a financial transaction in progress, keyword information is obtained from the call text generated by the user. The keyword information includes target keywords and the global weight of the target keywords. The target keywords are used to express the main content of the call text, and the global weight represents the degree of contribution of the target keywords to expressing the main content. The keyword information is matched with the keyword information of multiple template texts in a preset corpus to obtain the matching result between the keyword information and the keyword information of each template text. Based on the matching results, a target template text is determined from the plurality of template texts, and the user is classified according to the tag type of the target template text; the tag type represents the user's evaluation level of the financial business corresponding to the target template text; The step of performing relevance matching between the keyword information and the keyword information of multiple template texts in a preset corpus to obtain the matching result between the keyword information and the keyword information of each template text includes: The target keywords in the keyword information are semantically matched with the corresponding template keywords in the template information to obtain a first matching degree; the semantic matching is used to match the degree of semantic similarity between two words; and The global weight of the target keyword is matched with the global weight of the template keyword by distance to obtain a second matching degree; the distance matching is used to match the degree of difference between the global weights corresponding to the two words. Based on the first matching degree and the second matching degree, the matching results between the keyword information and each of the template information are obtained.

2. The method according to claim 1, characterized in that, The step of obtaining keyword information from the user-generated call text includes: The call text is segmented to obtain a word set for the call text; Based on the prefixes of each word in the word set, each word is classified to obtain a prefix tree for the call text; the prefix tree is a word clustering tree, and each branch in the prefix tree corresponds to a class of words; The prefix tree is scanned using a pre-defined dictionary to expand the words included in each branch of the prefix tree and generate a directed acyclic graph for each branch. Based on the word frequency and inverse document frequency of each word after expansion in the directed acyclic graph, the keyword information in the call text is obtained.

3. The method according to claim 2, characterized in that, The keyword information in the call text is obtained based on the term frequency and inverse document frequency of each word after expansion in the directed acyclic graph, including: Based on the word frequency and the inverse document frequency, the global weight of each word in the expanded directed acyclic graph is determined; Based on the global weight of each word, the words are sorted, and the target words ranked first and their corresponding global weights are used as keyword information in the call text.

4. The method according to claim 1, characterized in that, The step of determining the target template text from the plurality of template texts based on the matching results includes: If the matching result is based on the matching score representation, then among the multiple template texts, the candidate template text with the highest corresponding matching score is determined; If the matching score of the candidate template text is greater than a preset score, the candidate template text is determined to be the target template text.

5. The method according to claim 1, characterized in that, The step of determining the target template text from the plurality of template texts based on the matching results includes: Emotion detection processing is performed on the audio of the voice call to obtain an emotion detection result for the audio; the emotion detection processing is used to evaluate the voice emotion expressed in the audio. Based on the emotion detection results, candidate template texts are determined from the plurality of template texts; Based on the matching results, the target template text is determined from the candidate template texts.

6. The method according to claim 1, characterized in that, Following the classification of the users, the following is also included: Based on the user's user type, obtain the corresponding call strategy for the user type, and feed the call strategy back to the customer service agent; The call strategy is used to instruct the customer service agent to execute the corresponding outbound call frequency and call content for the user.

7. The method according to claim 1, characterized in that, Following the classification of the users, the following is also included: Based on the business type of the financial service, obtain the candidate call strategy corresponding to the business type; Based on the user's user type, a target call strategy corresponding to the user type is determined from the candidate call strategies, and the target call strategy is fed back to the customer service agent.

8. A user classification device, characterized in that, include: The information acquisition unit is configured to acquire keyword information from the call text generated by the user when a customer service agent conducts a voice call with the user regarding a financial transaction in progress. The keyword information includes target keywords and global weights of the target keywords. The target keywords are used to express the main content of the call text, and the global weights represent the degree of contribution of the target keywords to expressing the main content. The information matching unit is configured to perform relevance matching between the keyword information and the keyword information of multiple template texts in a preset corpus, and obtain the matching result between the keyword information and the keyword information of each template text; The user classification unit is configured to perform actions based on the matching results, to determine a target template text from the plurality of template texts, and to classify the user according to the tag type of the target template text; the tag type characterizes the user's evaluation level of the financial business corresponding to the target template text. The information matching unit is further configured to perform semantic matching between the target keyword in the keyword information and the corresponding template keyword in the template information to obtain a first matching degree; the semantic matching is used to match the degree of similarity in meaning between two words; and to perform distance matching between the global weight of the target keyword and the global weight of the template keyword to obtain a second matching degree; the distance matching is used to match the degree of difference between the global weights corresponding to two words. Based on the first matching degree and the second matching degree, the matching results between the keyword information and each of the template information are obtained.

9. A computer device, characterized in that, include: processor; Memory for storing the executable instructions of the processor; The processor is configured to execute the executable instructions to implement the user classification method as described in any one of claims 1 to 7.

10. A computer-readable storage medium comprising program data, characterized in that, When the program data is executed by the processor of a computer device, the computer device is able to perform the user classification method as described in any one of claims 1 to 7.

11. A computer program product comprising program instructions that, when executed by a processor of a computer device, enable the computer device to perform the user classification method as described in any one of claims 1 to 7.

Images