Data processing method, device, equipment and readable storage medium

Abstract

The present application discloses a data processing method, apparatus, device, and readable storage medium. The method comprises: obtaining target user conversational demand information in response to a target user inputting information into a conversational robot selection interface; generating target user demand features corresponding to the target user conversational demand information; inputting the target user demand features into a recommendation model; determining a target conversational robot from at least two conversational robots in the recommendation model based on a matching relationship and an association relationship between the target user demand features and at least two user demand features; and inputting the target user's conversational operation information into the target conversational robot, triggering the target conversational robot to perform a conversational service function. The present application can improve the efficiency of conversational robot recommendation.

Description

Technical Field

[0001] The present application relates to the field of computer technology, and in particular to a data processing method, apparatus, device, and readable storage medium. Background Art

[0002] With the increasing development of artificial intelligence technology, the application of intelligent conversational robot platforms in various industries has gradually deepened. However, users in different industries have different functional requirements and robot performance requirements for intelligent conversational robots. How to accurately recommend conversational robots that meet their needs to users has attracted widespread attention.

[0003] In the existing technology, the method of recommending conversational robots to users is mainly manual customization, which requires offline communication and then customizing the conversational robot based on the needs after the communication. This consumes a lot of manpower and material resources, resulting in high cost and low recommendation efficiency. Summary of the Invention

[0004] The embodiments of the present application provide a data processing method, apparatus, device, and readable storage medium, which can improve the efficiency of recommending conversational robots.

[0005] An embodiment of the present application provides a data processing method, including:

[0006] Respond to the target user's input operation on the dialogue robot selection interface to obtain the target user's dialogue demand information;

[0007] Generate target user demand features corresponding to target user dialogue demand information;

[0008] Inputting target user demand characteristics into a recommendation model; the recommendation model includes an association relationship between at least two user demand characteristics and at least two dialogue robots;

[0009] In the recommendation model, a target dialogue robot is determined from among the at least two dialogue robots based on a matching relationship and an association relationship between the target user's demand characteristics and the at least two user demand characteristics;

[0010] The target user's dialogue operation information is input into the target dialogue robot, triggering the target dialogue robot to perform dialogue business functions.

[0011] On the one hand, an embodiment of the present application provides another data processing method, including:

[0012] Obtain at least two sample user demand features, and obtain robot identification information of each of the at least two dialogue robots;

[0013] Inputting at least two sample user demand features and robot identification information into an initial recommendation model; the recommendation model includes an initial association relationship between the at least two sample user demand features and the at least two conversational robots;

[0014] Based on the initial association relationship in the initial recommendation model, the predicted conversational robot corresponding to each sample user's demand characteristics is output;

[0015] Obtain a prediction evaluation vector between the predicted conversational robot and at least two sample user demand features;

[0016] Obtain sample evaluation vector labels between at least two sample user demand features and at least two dialogue robots;

[0017] Based on the predicted evaluation vector and the sample evaluation vector label, the initial association relationship in the initial recommendation model is adjusted to obtain a recommendation model containing the association relationship; the recommendation model is used to identify the target dialogue robot that matches the target user's demand characteristics among at least two dialogue robot models.

[0018] In one aspect, the present application provides a data processing device, comprising:

[0019] A demand information acquisition module is used to respond to the target user's information input operation on the dialogue robot selection interface and obtain the target user's dialogue demand information;

[0020] A demand feature generation module, used to generate target user demand features corresponding to target user dialogue demand information;

[0021] A feature input module is used to input target user demand features into a recommendation model; the recommendation model includes an association relationship between at least two user demand features and at least two dialogue robots;

[0022] A conversational robot determination module, configured to determine a target conversational robot from among at least two conversational robots based on a matching relationship and an association relationship between a target user's demand feature and at least two user demand features in a recommendation model;

[0023] The business function execution module is used to input the dialogue operation information of the target user into the target dialogue robot, triggering the target dialogue robot to execute the dialogue business function.

[0024] Among them, information input operations include type input operations, call input operations, and domain input operations;

[0025] The demand information acquisition module includes:

[0026] An information acquisition unit, configured to respond to a type input operation on a dialogue robot selection interface and acquire dialogue intention type information;

[0027] The information acquisition unit is further used to respond to the call input operation on the dialogue robot selection interface and obtain the robot call count information;

[0028] The information acquisition unit is further configured to respond to a domain input operation on the dialogue robot selection interface and acquire dialogue application domain information;

[0029] The demand information generation unit is used to generate target user dialogue demand information based on dialogue intention type information, robot call count information and dialogue application field information.

[0030] The demand feature generation module includes:

[0031] A key information extraction unit is used to extract key fields from the target user's conversation demand information to obtain key user demand information;

[0032] A regularization processing unit, used for regularizing the key user demand information to obtain regularized user demand information;

[0033] The vector conversion unit is used to perform vector conversion on the regular user demand information to obtain the target user demand features corresponding to the target user dialogue demand information.

[0034] Among them, the dialogue robot determination module includes:

[0035] a matching requirement feature acquisition unit, configured to acquire, from among at least two user requirement features, a matching relationship between a target user requirement feature and at least two user requirement features, and to use a user requirement feature with a successful matching relationship as a matching user requirement feature;

[0036] an associated robot determining unit, configured to obtain, from the at least two conversational robots, a conversational robot associated with matching user demand characteristics based on the association relationship, as the associated conversational robot;

[0037] An associated evaluation vector acquisition unit is used to acquire an associated user evaluation vector associated with the associated dialogue robot and the target user's demand characteristics; one associated user evaluation vector corresponds to one associated dialogue robot;

[0038] The associated evaluation vector acquisition unit is further configured to acquire an associated user evaluation vector having a maximum vector modulus from the associated user evaluation vectors;

[0039] The target robot determination unit is used to determine the associated dialogue robot corresponding to the associated user evaluation vector with the largest vector modulus as the target dialogue robot.

[0040] The device further comprises:

[0041] A similarity determination module is used to determine the similarity between the target user demand feature and each of the at least two user demand features;

[0042] The matching relationship determination module is used to determine the matching relationship between the user demand feature with a similarity greater than a first similarity threshold and the target user demand feature as a successful matching relationship.

[0043] The device further comprises:

[0044] The usage data acquisition module is used to obtain the target user's usage behavior data for the target conversational robot;

[0045] A compatibility determination module is used to determine the compatibility between the target conversational robot and the target user's demand characteristics based on the usage behavior data, and to generate a target user evaluation vector associated with the target conversational robot and the target user's demand characteristics based on the compatibility;

[0046] an updated evaluation vector generation module, configured to generate an updated user evaluation vector based on the target user evaluation vector and user evaluation vectors associated with the target dialogue robot and the matching user demand features, if the similarity between the target user demand features and the matching user demand features is greater than a second similarity threshold;

[0047] The evaluation vector updating module is used to update the user evaluation vector associated with the target dialogue robot and the matching user demand characteristics according to the updated user evaluation vector.

[0048] Among them, the business function execution module includes:

[0049] A robot display unit is used to create a robot information management interface and display the target conversational robot in the robot information management interface;

[0050] A text information conversion unit, configured to respond to a target user's robot dialogue operation on the robot information management interface, obtain dialogue operation information, and convert the dialogue operation information into text information;

[0051] The business function triggering unit is used to input text information into the target dialogue robot and trigger the target dialogue robot to perform the dialogue business function associated with the text information.

[0052] On one hand, the present application provides a computer device, including: a processor, a memory, and a network interface;

[0053] The processor is connected to the memory and the network interface, wherein the network interface is used to provide data communication functions, the memory is used to store computer programs, and the processor is used to call the computer program to execute the method in one aspect of the embodiment of the present application.

[0054] On one hand, the present application provides a computer-readable storage medium, which stores a computer program. The computer program includes program instructions. When the program instructions are executed by a processor, the method in one aspect of the embodiment of the present application is executed.

[0055] In one aspect of the present application, a computer program product or computer program is provided. The computer program product or computer program includes computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the computer device to perform the method provided in one aspect of the embodiments of the present application.

[0056] In one aspect, an embodiment of the present application provides another data processing device, including:

[0057] An information acquisition module, configured to acquire at least two sample user demand features and obtain robot identification information of each of the at least two dialogue robots;

[0058] An information input module is used to input at least two sample user demand characteristics and robot identification information into an initial recommendation model; the recommendation model includes an initial association relationship between the at least two sample user demand characteristics and the at least two dialogue robots;

[0059] The prediction robot output module is used to output the predicted conversational robot corresponding to each sample user's demand characteristics based on the initial association relationship in the initial recommendation model;

[0060] A prediction evaluation acquisition module is used to obtain a prediction evaluation vector between the predicted dialogue robot and at least two sample user demand features;

[0061] An evaluation label acquisition module is used to obtain sample evaluation vector labels between at least two sample user demand features and at least two dialogue robots;

[0062] The relationship adjustment module is used to adjust the initial association relationship in the initial recommendation model based on the predicted evaluation vector and the sample evaluation vector label to obtain a recommendation model containing the association relationship; the recommendation model is used to identify the target dialogue robot that matches the target user's demand characteristics among at least two dialogue robot models.

[0063] The relationship adjustment module includes:

[0064] A loss function acquisition unit, used to obtain a loss function;

[0065] A loss value generating unit, configured to generate a loss function value based on the loss function, the predicted evaluation vector, and the sample evaluation vector label;

[0066] The relationship adjustment unit is used to adjust the initial association relationship according to the loss function value if the loss function value does not meet the model convergence condition, so as to obtain a recommendation model including the association relationship.

[0067] The loss value generation unit includes:

[0068] Hidden feature acquisition unit, used to obtain the predicted dialogue robot s i The first hidden feature of the conversational robot j The second hidden feature of predictive conversational robots i The sample user demand feature K output by the initial recommendation model q Corresponding predictive dialogue robot; dialogue robot s j For at least two conversational robots, with sample user demand features K q There are two dialogue robots with sample evaluation vector labels between them; i and j are both integers less than or equal to N, where N is the total number of at least two dialogue robots; q is an integer less than or equal to M, where M is the total number of at least two sample user demand features;

[0069] The loss value generating unit is used to generate a loss function value according to the first hidden feature, the second hidden feature, the predicted evaluation vector, the sample evaluation vector label and the loss function.

[0070] Among them, the loss function acquisition unit includes:

[0071] A vector matrix generating subunit, configured to generate a dialogue robot evaluation vector matrix based on at least two sample user demand features and sample evaluation vector labels of at least two dialogue robots;

[0072] a hidden feature matrix determination subunit, configured to determine hidden feature matrices of at least two dialogue robots based on the dialogue robot evaluation vector matrix;

[0073] The loss function generation subunit is used to generate a loss function based on the dialogue robot evaluation vector matrix and the hidden feature matrix.

[0074] The hidden feature matrix determination subunit is further configured to decompose the dialogue robot evaluation vector matrix to obtain a first decomposition set and a second decomposition set; the first decomposition set includes feedback features of each sample user's demand features for at least two dialogue robots; the second decomposition set includes feedback features of each dialogue robot for at least two sample user's demand features;

[0075] The hidden feature matrix determination subunit is further used to determine the hidden feature matrix based on the dialogue robot evaluation vector matrix, the first decomposition set and the second decomposition set.

[0076] On one hand, the present application provides a computer device, including: a processor, a memory, and a network interface;

[0077] The processor is connected to the memory and the network interface, wherein the network interface is used to provide data communication functions, the memory is used to store computer programs, and the processor is used to call the computer program to execute the method in one aspect of the embodiment of the present application.

[0078] On one hand, the present application provides a computer-readable storage medium, which stores a computer program. The computer program includes program instructions. When the program instructions are executed by a processor, the method in one aspect of the embodiment of the present application is executed.

[0079] In one aspect of the present application, a computer program product or computer program is provided. The computer program product or computer program includes computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the computer device to perform the method provided in one aspect of the embodiments of the present application.

[0080] In an embodiment of the present application, by obtaining the target user's target user dialogue demand information for the dialogue robot, a target user demand feature can be generated, and the target user demand feature is input into the recommendation model. The recommendation model can automatically determine the target dialogue robot that matches the target user demand feature and recommend the target dialogue robot to the target user. Among them, because the recommendation model contains the association relationship between at least two user demand features and at least two dialogue robots, and the target dialogue robot determined by the recommendation model is determined based on the matching relationship between the target user demand feature and the at least two user demand features and the association relationship, the target dialogue robot is also matched with the target user demand feature, that is, the target dialogue robot meets the needs of the target user. It can be seen from this that after obtaining the target user's dialogue demand information, the present application can automatically recommend a dialogue robot to the target user based on the association relationship in the recommendation model, which can improve the recommendation efficiency; and the entire recommendation process does not require human participation, which reduces offline communication time and reduces the manpower and material resources of customizing dialogue robots, thereby reducing costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0081] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0082] Figure 1 This is a schematic diagram of a network architecture provided by an embodiment of the present application;

[0083] Figure 2a This is a schematic diagram of generating user demand characteristics provided by an embodiment of the present application;

[0084] Figure 2b This is a schematic diagram of a scenario provided by an embodiment of the present application;

[0085] Figure 3 This is a flow chart of a data processing method provided in an embodiment of the present application;

[0086] Figure 4 This is a system architecture diagram provided by an embodiment of the present application;

[0087] Figure 5 1 is a flow chart of a data processing method provided in an embodiment of the present application;

[0088] Figure 6 This is a schematic diagram of the association between model training and application provided in an embodiment of the present application;

[0089] Figure 7 is a structural diagram of a data processing device provided in an embodiment of the present application;

[0090] Figure 8 is a schematic diagram of a computer device provided in an embodiment of the present application;

[0091] Figure 9 is a structural diagram of a data processing device provided in an embodiment of the present application;

[0092] Figure 10 is a schematic diagram of a computer device provided in an embodiment of the present application;

[0093] Figure 11 This is a structural diagram of a data processing system provided in an embodiment of the present application. DETAILED DESCRIPTION

[0094] The following will be combined with the drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are only part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0095] Artificial Intelligence (AI) refers to the theories, methods, techniques, and application systems that use digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, to perceive the environment, acquire knowledge, and use that knowledge to achieve optimal results. In other words, AI is a comprehensive technology within computer science that seeks to understand the essence of intelligence and produce new intelligent machines that can respond in a manner similar to human intelligence. AI also involves studying the design principles and implementation methods of various intelligent machines, enabling them to possess the capabilities of perception, reasoning, and decision-making.

[0096] Artificial intelligence (AI) technology is a comprehensive discipline encompassing a wide range of fields, encompassing both hardware and software technologies. Foundational AI technologies generally include sensors, specialized AI chips, cloud computing, distributed storage, big data processing, operating / interaction systems, and mechatronics. AI software technologies primarily encompass computer vision, speech processing, natural language processing, and machine learning / deep learning.

[0097] The solution provided in the embodiments of the present application belongs to the field of natural language processing (NLP) and machine learning (ML) under the artificial intelligence field.

[0098] Natural language processing (NLP) is a key area of research in computer science and artificial intelligence. It studies the theories and methods that enable effective communication between humans and computers using natural language. Natural language processing (NLP) integrates linguistics, computer science, and mathematics. Therefore, research in this field involves natural language—the language we use in everyday life—and is closely linked to the study of linguistics. Natural language processing technologies typically include text processing, semantic understanding, machine translation, robotic question answering, and knowledge graphs.

[0099] Machine learning (ML) is a multidisciplinary field that encompasses probability theory, statistics, approximation theory, convex analysis, and algorithmic complexity theory. It specifically studies how computers can simulate or implement human learning behaviors to acquire new knowledge or skills and reorganize existing knowledge structures to continuously improve their performance. Machine learning is at the core of artificial intelligence and the fundamental way to make computers intelligent. Its applications span all areas of AI. Machine learning and deep learning typically include techniques such as artificial neural networks, belief networks, reinforcement learning, transfer learning, inductive learning, and self-learning.

[0100] See Figure 1 , Figure 1 This is a schematic diagram of a network architecture provided by an embodiment of the present application. Figure 1 As shown, the network architecture may include a business server 1000 and a background server cluster, wherein the background server cluster may include multiple background servers, such as Figure 1 As shown, it may specifically include a backend server 100a, a backend server 100b, a backend server 100c, ..., a backend server 100n. Figure 1 As shown, the backend server 100a, the backend server 100b, the backend server 100c, ..., the backend server 100n can respectively establish a network connection with the above-mentioned business server 1000, so that each backend server can exchange data with the business server 1000 through the network connection, so that the above-mentioned business server 1000 can receive business data from each backend server.

[0101] like Figure 1 Each backend server shown corresponds to a user terminal and can be used to store the business data of the corresponding user terminal. Each user terminal can be integrated with a target application. When the target application is running in each user terminal, the backend server corresponding to each user terminal can store the business data in the application and Figure 1Data interaction is performed between the business servers 1000 shown. Among them, the target application may include an application with the function of displaying data information such as text, images, audio and video. For example, the application can be an object recommendation application, which can be used for users to input demand information and obtain target objects that match the demand information (for example, the application is a conversation robot recommendation application, and after the user enters the conversation demand information, the target conversation robot that meets the conversation demand can be obtained); the business server 1000 in this application can collect business data from the background of these applications (such as the above-mentioned background server cluster), for example, the business data can be the conversation demand information for the conversation robot entered by the user. Based on the collected business data, the business server 1000 can determine the target conversation robot that matches the business data. Furthermore, the business server 1000 can send the target conversation robot to the background server, and the user can view the target conversation robot through the user terminal corresponding to the background server, and then the user can have a conversation with the target conversation robot. For example, the demand information input by user a is "intelligent question-and-answer robot". The business server 1000 determines the dialogue robot M that meets the demand information "intelligent question-and-answer robot" based on the demand information "intelligent question-and-answer robot". The business server 1000 can return the dialogue robot M to the background server of the user terminal used by user a. User a can view the dialogue robot M on the display page of the user terminal, and user a can enter questions on the display page. The dialogue robot M can answer the questions entered by user a.

[0102] In the embodiment of the present application, a user terminal can be selected from multiple user terminals as a target user terminal, and the target user terminal can include: smart phones, tablet computers, desktop computers and other smart terminals with the function of displaying and playing data information. For example, in the embodiment of the present application, Figure 1 The user terminal corresponding to the background server 100 a shown is used as the target user terminal. The target user terminal may be integrated with the above-mentioned target application. At this time, the background server 100 a corresponding to the target user terminal may perform data interaction with the business server 1000 .

[0103] For example, when a user is using a target application (such as an object recommendation application) in a user terminal, the business server 1000 detects and collects the user's demand information for a dialogue robot through the background server corresponding to the user terminal. The business server 1000 can determine one or more dialogue robots that meet the demand information and send the dialogue robot to the background server. The user can then view the dialogue robot on the display page of the user terminal corresponding to the background server, and the user can enter dialogue operation information to communicate with the dialogue robot.

[0104] Optionally, it is understood that the backend server can detect the business data collected from the corresponding user terminals (such as user demand information for the conversation robot) and determine the conversation robot that matches the business data. The user can view the conversation robot determined by the backend server on the display page of the user terminal corresponding to the backend server.

[0105] Optionally, it is understood that the backend server can detect and collect business data (such as user demand information for a conversational robot) from the corresponding user terminal, and the backend server can generate data features (such as user demand features) based on the business data. The backend server can send the data features to the business server 1000, and the business server 1000 can determine a conversational robot that matches the data features and return the conversational robot that matches the data features to the backend server. Thus, the user can view the conversational robot determined by the business server 1000 on the display page of the user terminal corresponding to the backend server.

[0106] It is understood that the methods provided in the embodiments of the present application can be executed by a computer device, including but not limited to a user terminal or a business server. The business server can be an independent physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, CDN, and big data and artificial intelligence platforms.

[0107] The user terminal and the service server may be connected directly or indirectly via wired or wireless communication, which is not limited in this application.

[0108] For easier understanding, see Figure 2a , Figure 2a This is a schematic diagram of generating user demand characteristics provided by an embodiment of the present application. Figure 2a The user terminal E shown can be Figure 1 Any user terminal selected from the user terminal cluster of the corresponding embodiment, for example, the user terminal may be the user terminal 100b.

[0109] like Figure 2a As shown, user E can be the target user, and user E uses user terminal E to input information to the dialogue robot. Figure 2aAs shown, user E can enter the robot name and the robot type (i.e., the conversation intention type) in the conversation robot selection interface 1 of the user terminal E. For the option "Robot Name", the text information entered by user E is "Xiao A"; for the option "Robot Type", user E selects the type "Question and Answer Robot", that is, the conversation intention of user E is "Question and Answer"; Figure 2a As shown, after selecting the robot type, user E can click "Next" in the dialogue robot selection interface 1. Subsequently, user terminal E can respond to this trigger operation of user E and jump to the dialogue robot selection interface 2. In the dialogue robot selection interface 2, user E can select the robot call amount, such as Figure 2a As shown, the robot call volume (i.e., the number of robot calls) selected by user E is "1000-10000". After selecting the robot call volume, user E can click "Next" in the dialogue robot selection interface 2. Subsequently, user terminal E can respond to this trigger operation of user E and jump to the dialogue robot selection interface 3. In the dialogue robot selection interface 3, user E can select the robot business field (i.e., dialogue application field), such as Figure 2a As shown, user E has selected e-commerce as the robot's business domain. Furthermore, if user E has other requirements for the conversational robot, such as speech recognition, machine translation, and text review capabilities, user E can also select advanced capabilities. After completing their selections, user E clicks "Done" on conversational robot selection interface 3. User terminal E responds to this triggering action and generates user E's conversational requirement information, namely, "a question-and-answer robot with an e-commerce business domain and a call volume between 1,000 and 10,000." Furthermore, the user terminal E may perform key field extraction on the conversation demand information of the user E, “a question-and-answer type robot whose business field is e-commerce and whose call volume is between 1,000 and 10,000,” to generate key demand information. For example, if the key field extracted from the conversation demand information, “a question-and-answer type robot whose business field is e-commerce and whose call volume is between 1,000 and 10,000,” is “e-commerce field, low call volume, question-and-answer type,” then the key demand information may be “e-commerce field, low call volume, question-and-answer type.” Subsequently, the user terminal E may perform regularization on the key demand information, and then perform vector conversion, thereby obtaining the user demand characteristics of the user E (i.e., the target user demand characteristics).

[0110] For easier understanding, please refer to Figure 2b , Figure 2b This is a schematic diagram of a scenario provided by an embodiment of the present application. Figure 2bThe user terminal E shown can be Figure 1 Any user terminal selected from the user terminal cluster of the corresponding embodiment, for example, the user terminal may be the user terminal 100b.

[0111] like Figure 2b As shown, the user terminal E can Figure 2aThe target user demand characteristics of user E determined in the corresponding embodiment are input into the recommendation model, because the recommendation model includes associations between at least two user demand characteristics and at least two conversational robots. For example, the associations may be {[User demand characteristic a: Conversational Robot 1, Conversational Robot 3], [User demand characteristic b: Conversational Robot 2, Conversational Robot 3], [User demand characteristic c: Conversational Robot 4, Conversational Robot 5}. This association can be understood as a relationship between the conversational robots satisfying the user demand characteristics. For example, Conversational Robot 1 and Conversational Robot 3 both satisfy User demand characteristic a, and Conversational Robot 2 and Conversational Robot 3 both satisfy User demand characteristic b. Among the at least two user demand features, a user demand feature that has a successful matching relationship with the target user demand feature of user E (that is, a user demand feature whose similarity with the target user demand feature is greater than a first similarity threshold) can be obtained as a matching user demand feature; then, based on the association relationship, a dialogue robot associated with the matching user demand feature can be obtained from the at least two dialogue robots as an associated dialogue robot; for example, among user demand feature a, user demand feature b, and user demand feature c, the similarity between user demand feature a and the target user demand feature of user E is 0.6, the similarity between user demand feature b and the target user demand feature of user E is 0.85, and the similarity between user demand feature c and the target user demand feature of user E is 0.95. The similarity between the features is 0.4. Since the first similarity threshold is 0.5, and 0.85>0.6>0.5, it can be seen that the user demand features whose similarity with the target user demand features of user E is greater than the first similarity threshold of 0.5, include user demand feature b and user demand feature a. The matching relationship between user demand feature b and user E's target user demand features can be determined as the matching success relationship, and the matching relationship between user demand feature a and user E's target user demand features can also be determined as the matching success relationship. Then, both user demand feature b and user demand feature a are user demand features that have a matching success relationship with user E's target user demand features, that is, user demand feature b and user demand feature a are both matching user demand features. Because the conversational robots associated with the matching user demand feature b are conversational robots 2 and conversational robots 3, and the conversational robots associated with the matching user demand feature a are conversational robots 1 and conversational robots 2, conversational robots 1, conversational robots 2, and conversational robots 3 can all be determined as associated conversational robots. Among them, the first similarity threshold can be presented in the form of numerical values such as decimals, percentages, and fractions, and this application does not impose any restrictions.

[0112] Furthermore, the associated user evaluation vector associated with the associated dialogue robot and the target user's demand characteristics can be obtained. One associated user evaluation vector corresponds to one associated dialogue robot. For the specific method of obtaining the associated user evaluation vector, please refer to the following Figure 3 Furthermore, among these associated user evaluation vectors, the associated user evaluation vector with the largest vector modulus can be obtained, and the associated dialogue robot corresponding to the associated user evaluation vector with the largest vector modulus is determined as the target dialogue robot.

[0113] Then, if Figure 2b As shown, after determining the target dialogue robot (e.g., dialogue robot 3), the recommendation model can return the target dialogue robot 3 to the backend server of the user terminal E. Subsequently, the user terminal E can create a robot information management interface and display the information of the dialogue robot 3 in the robot information management interface. Figure 2b As shown, user E can view the information of the conversation robot 3 in the robot information management interface, for example, the name of the conversation robot 3 is "Xiao A" (such as Figure 2a In the corresponding embodiment, the robot name entered by user E), the status of the conversation robot 3 is online, and user E can manage his own conversation robot in the robot information management interface. User E can also perform robot conversation operations in the robot information management interface, such as Figure 2b As shown, user E can click "Enter Conversation". User terminal E can respond to this triggering action by creating a robot dialogue interface, in which user E can engage in a conversation with conversation robot 3. User E can input the questions they want to ask by voice or text. User terminal E can use the questions input by user E by voice or text as conversation operation information and convert the conversation operation information into text information. When this text information is input to conversation robot 3, it can trigger conversation robot 3 to perform the conversation service function, that is, conversation robot 3 can answer the questions input by user E.

[0114] For easier understanding, see Figure 3 , Figure 3 This is a flow chart of a data processing method provided by an embodiment of the present application. The method can be performed by a user terminal (for example, the above Figure 1 、 Figure 2a as well as Figure 2b The user terminal shown in FIG) can also be executed by the user terminal and the service server (as shown above Figure 1For ease of understanding, this embodiment uses the method executed by the above-mentioned user terminal as an example to illustrate the specific process of robot recommendation in the user terminal. The method may include at least the following steps S101-S105:

[0115] Step S101: In response to the target user's information input operation on the dialogue robot selection interface, the target user's dialogue demand information is obtained.

[0116] In this application, the information input operation here may include type input operation, call input operation and field input operation. It can be understood that the target user can select the robot type (such as question-and-answer type, chat type, task type and entertainment type, etc.) in the dialogue robot selection interface of the user terminal, as mentioned above. Figure 2a As shown, the target user (user E) can select from the existing robot types in the dialogue robot selection interface 1. It should be understood that the target user can also input their requirements for the dialogue robot by typing keywords (such as Q&A, small talk); the target user can also input their requirements for the dialogue robot by voice input. The user terminal can respond to the target user's type input operation and obtain the target user's dialogue intention type information. Similarly, the target user can enter the number of robot calls in the dialogue robot selection interface of the user terminal (select from the existing call counts or enter the call count by typing or voice input), and the user terminal can respond to the target user's call input operation and obtain the robot call count information. Similarly, the target user can enter the number of robot fields (such as online education, smart transportation, smart media, etc.) in the dialogue robot selection interface of the user terminal (select from the existing robot fields or enter the robot field by typing or voice input), and the user terminal can respond to the target user's field input operation and obtain dialogue application field information. Based on the dialogue intention type information, the robot call count information, and the dialogue application field information, the target user's dialogue requirement information can be generated.

[0117] It should be noted that there is no time sequence or logical order for type input operations, call input operations, and domain input operations. The type input operation can be before or after the call input operation or domain input operation.

[0118] It should be understood that information input operations include but are not limited to type input operations, call input operations, and field input operations. Information input operations can also include input operations of advanced robot capabilities (such as voice recognition capabilities, machine translation capabilities, text review capabilities, etc.), etc., and no examples will be given one by one here.

[0119] Step S102: generating target user demand features corresponding to target user dialogue demand information.

[0120] In this application, the target user conversation demand information can be subjected to key field extraction to obtain key user demand information; then, the key user demand information can be subjected to regularization processing to obtain regular user demand information; the regular user demand information can be subjected to vector conversion to obtain the target user demand features corresponding to the target user demand information.

[0121] Step S103: input the target user demand characteristics into the recommendation model; the recommendation model includes the association relationship between at least two user demand characteristics and at least two dialogue robots.

[0122] In this application, the recommendation model can be a trained recommendation model, and the association relationship in the recommendation model is also a correspondence between user demand characteristics and the dialogue robot obtained through training.

[0123] Step S104: In the recommendation model, a target dialogue robot is determined from among at least two dialogue robots based on a matching relationship and an association relationship between the target user demand feature and at least two user demand features.

[0124] In the present application, among the above-mentioned at least two user demand features, a matching relationship with the target user demand feature can be obtained (for example, a weak similarity relationship, no similarity relationship, a strong similarity relationship, etc.), and the user demand feature with a matching relationship as a successful matching relationship can be used as the matching user demand feature; wherein, the specific method for determining whether the matching relationship is a successful matching relationship can be to determine the similarity between the target user demand feature and each user demand feature of the at least two user demand features, and if the similarity is greater than a first similarity threshold, the user demand feature with a similarity greater than the first similarity threshold can be determined as having a strong similarity with the target user demand feature, and the user demand feature with a similarity greater than the first similarity threshold can be determined as having a strong similarity with the target user demand feature. The matching relationship between the feature and the target user demand feature is determined to be a successful matching relationship; if the similarity is greater than 0 but less than the first similarity threshold, the user demand feature with the similarity greater than 0 but less than the first similarity threshold can be determined as having a weak similarity with the target user demand feature, and the user demand feature with the similarity greater than 0 but less than the first similarity threshold can be considered to have a matching relationship with the target user demand feature, but not a successful matching relationship; if the similarity is less than 0, the user demand feature with the similarity less than 0 can be determined to have no similarity with the target user demand feature, and the user demand feature with the similarity less than 0 can be considered to have no matching relationship with the target user demand feature.

[0125] Furthermore, based on the association between at least two user demand features and at least two dialogue robots in the recommendation model, the dialogue robot associated with the matching user demand feature can be obtained from the at least two dialogue robots, and the associated dialogue robot can be used as the associated dialogue robot. Figure 2b Taking the corresponding embodiment as an example, the association relationship in the recommendation model is {[user demand feature a: dialogue robot 1, dialogue robot 3], [user demand feature b: dialogue robot 2, dialogue robot 3], [user demand feature c: dialogue robot 4, dialogue robot 5}, where, because the matching user demand features are user demand feature a and user demand feature b, and because in the association relationship, the dialogue robots associated with user demand feature a are dialogue robot 1 and dialogue robot 3, and the dialogue robots associated with user demand feature b are dialogue robot 2 and dialogue robot 3, dialogue robot 1, dialogue robot 2 and dialogue robot 3 can all be regarded as associated dialogue robots.

[0126] Furthermore, an associated user evaluation vector associated with the associated conversational robot and the target user's demand characteristics can be obtained; wherein, one associated user evaluation vector corresponds to one associated conversational robot. Here, the associated user evaluation vector can be a predicted evaluation vector generated by the recommendation model based on matching the user demand characteristics with the user evaluation vector of an associated conversational robot. In other words, an associated user evaluation vector can be used to represent the suitability of an associated conversational robot for the target user's demand characteristics as predicted by the recommendation model.

[0127] For example, as mentioned above Figure 2b In the corresponding embodiment, dialogue robot 1, dialogue robot 2 and dialogue robot 3 are all associated dialogue robots. Taking dialogue robot 3 as an example, through the association relationship {[user demand feature a: dialogue robot 1, dialogue robot 3], [user demand feature b: dialogue robot 2, dialogue robot 3], [user demand feature c: dialogue robot 4, dialogue robot 5}, it can be seen that the matching user demand features associated with the associated dialogue robot 3 are matching user demand feature a and matching user demand feature b; among them, the user evaluation vector associated with the associated dialogue robot 3 and the matching user demand feature a is "P", and the user evaluation vector associated with the associated dialogue robot 3 and the matching user demand feature b is is "Q", the recommendation model can generate an associated user evaluation vector (such as "P+Q") based on the user evaluation vector "P" and the user evaluation vector "Q", and the associated user evaluation vector "P+Q" is associated with the target user demand feature of the associated dialogue robot 3 and user E; taking the dialogue robot 1 as an example, it can be seen from the association relationship that the matching user demand feature associated with the associated dialogue robot 1 is matching user demand feature a, and the user evaluation vector associated with the associated dialogue robot 1 and the matching user demand feature a is "T", then based on the user evaluation vector "T", an associated user evaluation vector (such as "T") between the associated dialogue robot 1 and the target user demand feature is generated. 1 ”).

[0128] It can be understood that to obtain the associated user evaluation vector between an associated conversational robot and the target user's demand characteristics, it is necessary to first obtain the matching user demand characteristics associated with the associated conversational robot, then obtain the user evaluation vector between the associated conversational robot and each matching user demand characteristic, and then generate the associated user evaluation vector between the associated conversational robot and the target user's demand characteristics based on these user evaluation vectors. The matching user demand characteristics here can be understood as the user demand characteristics of historical users, and the user evaluation vector between an associated conversational robot and each matching user demand characteristic can be understood as the degree of fit determined by the recommendation model based on historical user usage behavior data for the associated conversational robot, and an evaluation vector generated based on this fit. For example, before user E, user C used user terminal C to input the demand information of the dialogue robot. User terminal C generated user demand feature C based on the demand information of user C. The recommendation model determined that the dialogue robot that best matched the user demand feature C of user C was dialogue robot 1 based on the association relationship. The recommendation model then recommended dialogue robot 1 to user C. In the process of user C using the dialogue robot 1, the usage behavior data of user C was collected (such as the frequency of user C's use of the dialogue robot 1, user C's scoring data of the dialogue robot, and other related behavior data). Based on the usage behavior data of user C, the recommendation model can generate a degree of compatibility between the user demand feature C of user C and the dialogue robot 1, and obtain a user evaluation vector C based on the degree of compatibility. Subsequently, after receiving the target user demand feature of user E, the recommendation model determines that the similarity between the target user demand feature and the user demand feature C is greater than the first similarity threshold. The recommendation model can then use the dialogue robot 1 associated with the user demand feature C as the associated dialogue robot of the target user demand feature, and determine the associated user evaluation vector between the target user demand feature and the associated dialogue robot 1 based on the user evaluation vector C between the user demand feature C and the associated dialogue robot 1.

[0129] It is understood that the associated user evaluation vector "P+Q" is associated with the target user demand characteristics of the associated dialogue robot 3 and user E. The associated user evaluation vector "P+Q" is generated by the recommendation model based on the user evaluation vector "P" matching user demand characteristic a and the user evaluation vector "Q" matching user demand characteristic b. This associated user evaluation vector "P+Q" can be used to represent the suitability of dialogue robot 3 for the target user demand characteristics, as predicted by the recommendation model based on the user evaluation vectors "P" and "Q".

[0130] It can be understood that there is an associated associated user evaluation vector between each associated dialogue robot and the target user demand feature. Among these associated user evaluation vectors (1 or more), the associated user evaluation vector with the largest vector modulus can be obtained; then, the associated dialogue robot corresponding to the associated user evaluation vector with the largest vector modulus can be determined as the target dialogue robot.

[0131] For example, the associated user evaluation vector "T 1 The vector modulus of the associated user evaluation vector "P+Q" between the associated dialogue robot 3 and the target user demand feature of user E is 2, and the vector modulus of the associated user evaluation vector "P+Q" is 3. Since 2<3, the associated user evaluation vector "P+Q" has the largest vector modulus, and the associated dialogue robot 3 corresponding to the associated user evaluation vector "P+Q" can be determined as the target dialogue robot. It should be understood that for the associated user evaluation vector "T 1 ", the associated user evaluation vector "P+Q", the vector modulus 2, and the vector modulus 3 are all examples for ease of understanding and have no practical significance.

[0132] Step S105: input the target user's dialogue operation information into the target dialogue robot, triggering the target dialogue robot to perform dialogue business functions.

[0133] In this application, the user terminal where the target user is located can create a robot information management interface and display the target conversation robot in the robot information management interface; thus, the target user can view the information of the target conversation robot in the robot information management interface, such as Figure 2b Taking the corresponding embodiment as an example, target user E can view the information of target conversational robot 3 in the robot information management interface. Target user E can also click "Enter Conversation" in the robot information management interface to engage in a conversation with target conversational robot 3. User terminal E, where target user E is located, will respond to this robot conversation operation by creating a robot conversation interface. Target user E can enter conversational operation information (e.g., voice input of a question) in this robot conversation interface. User terminal E can convert user E's conversational operation information into text and input this text information into the target conversational robot, thereby triggering the target conversational robot to execute the conversational service function associated with the text information. For example, if user E's voice message asks "What are some ways to cook carrots?", user terminal E converts the voice message into text and inputs it into conversational robot 3. After this, conversational robot 3 can respond to user E with one or more recipes for carrots.

[0134] Optionally, it should be understood that in the process of the target user using the target dialogue robot, the target user's usage behavior data for the target dialogue robot can be collected and obtained; based on the usage behavior data, the actual compatibility between the target dialogue robot and the target user's demand characteristics can be determined, and then a target user evaluation vector associated with the target dialogue robot and the target user's demand characteristics can be generated based on the compatibility; it can be understood that when determining the target dialogue robot for the target user's demand characteristics, the recommendation model will generate an associated user evaluation vector between each associated dialogue robot and the target user's demand characteristics based on the user evaluation vector associated with each associated dialogue robot and the matching user's demand characteristics (that is, The associated user evaluation vector can be a comprehensive evaluation vector determined by the recommendation model based on the user evaluation vector between the matching user demand characteristics and the associated dialogue robot. The associated user evaluation vector is a degree of suitability between each associated dialogue robot for the target user demand characteristics predicted by the recommendation model). Furthermore, after obtaining the associated user evaluation vector, the vector module lengths of the associated user evaluation vectors can be sorted to determine the final target dialogue robot. After the determination is completed, the target user can use the target dialogue robot. According to the target user's usage behavior data for the target dialogue robot, the recommendation model can determine a real fitness degree, and then generate a target user evaluation vector based on the real fitness degree.

[0135] Subsequently, if the similarity between the target user demand feature and the matching user demand feature is greater than the second similarity threshold, an updated user evaluation vector can be generated based on the target user evaluation vector and the user evaluation vector associated with the target dialogue robot and the matching user demand feature. Based on the updated user evaluation vector, the user evaluation vector associated with the target dialogue robot and the matching user demand feature can be updated. The second similarity threshold can be in the form of a decimal (such as 0.7, 0.9), a percentage (such as 60%, 80%), and a percentage. The second similarity threshold can be equal to the above-mentioned first similarity threshold or unequal to the above-mentioned first similarity threshold. For example, with the above Figure 2bTaking the corresponding embodiment as an example, the target dialogue robot is the dialogue machine 3, wherein, because the similarity between the target user demand feature and the matching user demand feature a is 0.6, and the similarity between the target user demand feature and the matching user demand feature b is 0.85, it can be understood that the target user demand feature and the matching user demand feature b have extremely strong similarity, then an updated user evaluation vector (such as the updated user evaluation vector is "S") can be generated based on the target user evaluation vector (such as the target user evaluation vector is "R"), and the user evaluation vector "Q" associated with the target dialogue robot 3 and the matching user demand feature b. Among them, the specific method for generating the updated user evaluation vector "S" can be to determine the mean vector of the target user evaluation vector "R" and the user evaluation vector "Q", and use the mean vector as the updated user evaluation vector "S"; or it can be to weight the target user evaluation vector "R" and the user evaluation vector "Q" respectively, and then fuse the two weighted evaluation vectors to obtain the updated user evaluation vector "S". The specific method for determining and updating the user evaluation vector may be other methods, which will not be given as examples here.

[0136] It should be understood that the user evaluation vector associated with each conversational robot and each user demand feature can be updated multiple times, and each update is based on other user demand features that have a strong similarity to the user demand feature. In other words, the user evaluation vector associated with a conversational robot and a user demand feature is determined by other user evaluation vectors between one or more similar user demand features and this conversational robot, and each user evaluation vector can be updated. In view of this, the user evaluation vector between each conversational robot and each user demand feature has a high accuracy. It is understandable that the recommendation model can cluster user demand features. Through multiple precise classifications (similar user demand features are classified into one category), different conversational robots can be matched with different categories of user demand features, and different categories of evaluations can be performed to obtain user evaluation vectors. The process of updating the user evaluation vector is also a process of optimizing the recommendation model, which can improve the accuracy of the recommendation model.

[0137] In an embodiment of the present application, by obtaining the target user's target user dialogue demand information for the dialogue robot, a target user demand feature can be generated, and the target user demand feature is input into the recommendation model. The recommendation model can automatically determine the target dialogue robot that matches the target user demand feature and recommend the target dialogue robot to the target user. Among them, because the recommendation model contains the association relationship between at least two user demand features and at least two dialogue robots, and the target dialogue robot determined by the recommendation model is determined based on the matching relationship between the target user demand feature and the at least two user demand features and the association relationship, the target dialogue robot is also matched with the target user demand feature, that is, the target dialogue robot meets the needs of the target user. It can be seen from this that after obtaining the target user's dialogue demand information, the present application can automatically recommend a dialogue robot to the target user based on the association relationship in the recommendation model, which can improve the recommendation efficiency; and the entire recommendation process does not require human participation, which reduces offline communication time and reduces the manpower and material resources of customizing dialogue robots, thereby reducing costs. At the same time, because the target conversational robot determined by the recommendation model is determined based on the user evaluation vector of matching user demand characteristics similar to the target user demand characteristics, and the user evaluation vector corresponding to each matching user demand characteristic will be updated multiple times based on the usage behavior data of different users, so each user evaluation vector of matching user demand characteristics can accurately represent the degree of suitability between a conversational robot and the user demand characteristics, then the target conversational robot determined based on the user evaluation vector will also have a higher accuracy (that is, the target conversational robot will be more suitable for the target user demand characteristics).

[0138] For easier understanding, see Figure 4 , Figure 4 This is a system architecture diagram provided by the embodiment of this application. Figure 4 As shown, the recommendation system can include: a data preparation module, a recommendation algorithm module, a real-time recommendation module, an A / B testing module, and a recommendation result storage module. The data preparation module and the recommendation algorithm module constitute the learning subsystem of the recommendation system, while the real-time recommendation module and the A / B testing module constitute the prediction subsystem of the recommendation system.

[0139] The data preparation module may include:

[0140] The Web service submodule can be a module that directly serves users. Its main function is: when the user triggers the recommendation system on the UI interface, the Web service module can trigger the interface of the recommendation system to provide personalized recommendations for the user.

[0141] The data collection submodule can be used to obtain the demand information input by the user. It can also be used to collect the user's usage behavior data when using the conversational robot.

[0142] The ETL submodule can extract key fields from the raw data collected by the data collection module and convert the key fields into structured data. In other words, the ETL module can be used to extract key fields from the raw data and perform standard normalization on the key fields.

[0143] The feature engineering submodule uses various machine learning algorithms to learn user preferences (user demand characteristics) within the recommendation system and recommends conversational bots based on these preferences. The main function of this feature engineering submodule is to convert the data processed by the ETL module into features.

[0144] Among them, the recommendation algorithm module includes:

[0145] The recommendation model and result ranking module input the features converted by the feature engineering submodule into the recommendation model and output one or more conversational robots that match the features. The result ranking module then ranks these conversational robots and ultimately determines a target conversational robot.

[0146] The recommendation result storage module can be used to store each recommendation result. In this application, in order to reduce the delay in returning the recommendation results to the front end (such as the user terminal), a horizontally scalable database such as Redis or CouchBase can be used to store relevant data (such as recommendation results).

[0147] The real-time recommendation module in the prediction subsystem can make real-time recommendations to the needs of the target user based on the relevant data (such as recommendation results) stored in the recommendation result storage module.

[0148] The A / B testing module in the prediction subsystem can be used to test the recommendation results in the real-time recommendation module to see how well the recommendation results meet the needs of target users.

[0149] See Figure 5 , Figure 5 This is a flow chart of a data processing method provided by an embodiment of the present application. The data processing method provided by the embodiment of the present application can be a model training method. After the model training of the recommendation model is completed by using the data processing method, the trained recommendation model can also be used in different application scenarios according to business needs; for example: a dialogue robot recommendation scenario. The data processing (model training) method can be performed by a user terminal (for example, the above Figure 1 、 Figure 2aas well as Figure 2b The user terminal shown in FIG) can also be executed by the service server (as shown above Figure 1 For ease of understanding, this embodiment uses the method executed by the above-mentioned user terminal as an example to illustrate the specific process of training the recommendation model. The method may include at least the following steps S201-S206:

[0150] Step S201: Obtain at least two sample user demand features and obtain robot identification information of each of at least two dialogue robots.

[0151] In this application, the sample user's conversation demand information can be obtained based on the sample user's choices for each step of the conversation robot, so that the sample user demand characteristics can be generated; the structural parameters of the conversation robot (such as version number, framework parameters, network structure and other parameters) can be used as machine identification information.

[0152] Step S202: Input at least two sample user demand features and robot identification information into an initial recommendation model; the recommendation model includes an initial association relationship between at least two sample user demand features and at least two dialogue robots.

[0153] In this application, the initial recommendation model here can be a recommendation model that has not yet been trained. At least two sample user demand features and robot identification information are input into the initial recommendation model to train the initial recommendation model. The initial association relationship here is the association relationship between the sample user demand features that have not yet been trained and the conversational robot.

[0154] Step S203: Output the predicted dialogue robot corresponding to each sample user's demand feature through the initial association relationship in the initial recommendation model.

[0155] In this application, because the initial association relationship includes the correspondence between the sample user demand features and the dialogue robot, the predicted dialogue robot corresponding to each sample user demand feature can be determined through the initial association relationship.

[0156] Step S204: Obtain a prediction evaluation vector between the predicted dialogue robot and at least two sample user demand features.

[0157] In this application, the initial recommendation model automatically predicts an evaluation vector between a conversational robot and a sample user's demand feature as a predicted evaluation vector. For example, for sample user demand feature a, the initial recommendation model predicts conversational robots A and B. The initial recommendation model can predict a predicted evaluation vector between sample user demand feature a and predicted conversational robot A, and can also predict a predicted evaluation vector between sample user demand feature a and predicted conversational robot B.

[0158] Step S205: Obtain sample evaluation vector labels between at least two sample user demand features and at least two dialogue robots.

[0159] In this application, for each sample user demand feature, the conversation robot that should be associated will be manually labeled as the labeled conversation robot. At the same time, an evaluation vector will be labeled between each sample user demand feature and the associated labeled conversation robot. This labeled evaluation vector can be used as a sample evaluation vector label. For example, for sample user demand feature a, the manually labeled conversation robot is conversation robot C, then this conversation robot C is the labeled conversation robot corresponding to the sample user demand feature a. An evaluation vector can be determined between the sample user demand feature a and the labeled conversation robot C as the labeled evaluation vector between the sample user demand feature a and the labeled conversation robot C, that is, the sample evaluation vector label.

[0160] Step S206: Adjust the initial association relationship in the initial recommendation model based on the predicted evaluation vector and the sample evaluation vector label to obtain a recommendation model containing the association relationship; the recommendation model is used to identify the target dialogue robot that matches the target user's demand characteristics among at least two dialogue robot models.

[0161] In the present application, a loss function value can be generated based on the predicted evaluation vector and the sample evaluation vector label. The specific method can be to first obtain the loss function, and then generate the loss function value based on the loss function, the predicted evaluation vector and the sample evaluation vector label. The loss function here can be generated based on the sample user demand feature and the sample evaluation vector label between the dialogue robot. The specific method can be to generate a dialogue robot evaluation vector matrix based on the sample evaluation vector label between the at least two sample user demand features and the at least two dialogue robots; then, the dialogue robot evaluation vector matrix can be decomposed to obtain a first decomposition set and a second decomposition set, wherein the first decomposition set can include the feedback features of each sample user demand feature for the at least two dialogue robots; the second decomposition set can include the feedback features of each dialogue robot for the at least two sample user demand features; based on the dialogue robot evaluation vector matrix, the first decomposition set and the second decomposition set, a hidden feature matrix can be determined; then, based on the dialogue robot evaluation vector matrix and the hidden feature matrix, a loss function can be generated.

[0162] It can be understood that a sample user demand feature corresponds to one or more labeled conversational robots. Therefore, there will be a labeled evaluation vector (i.e., a sample evaluation vector label) between a sample user demand feature and each associated labeled conversational robot. Based on these sample evaluation vector labels, a conversational robot evaluation vector matrix can be generated. For example, as shown in Table 1, the sample user demand features may include sample user demand feature 1, sample user demand feature 2, sample user demand feature 3, and sample user demand feature 4, and the conversational robots may include conversational robot A, conversational robot B, conversational robot C, and conversational robot D. For sample user demand feature 1, the corresponding labeled conversational robots are conversational robot B and conversational robot C, and the labeled evaluation vector between sample user demand feature 1 and conversational robot B is "a," while the labeled evaluation vector between sample user demand feature 1 and conversational robot C is "c." Similarly, Table 1 shows the labeled conversational robots corresponding to sample user demand feature 2, sample user demand feature 3, and sample user demand feature 4, as well as the labeled evaluation vectors between them and the corresponding labeled conversational robots.

[0163] Table 1

[0164] Conversational Robot A Conversational Robot B Conversational Robot C Conversational Robot D <![CDATA[Sample user requirement feature 1]]> 0 a c 0 Sample user demand characteristics 2 b g 0 0 Sample user demand characteristics 3 u v y 0 <![CDATA[Sample user requirement feature 4]]> z h 0 x

[0165] According to the labeled evaluation vectors in Table 1, the evaluation vector matrix of the dialogue robot can be obtained. The evaluation vector matrix of the dialogue robot can be shown as A1:

[0166] Dialogue robot evaluation vector matrix A1

[0167] According to the evaluation vector matrix A1 of the dialogue robot, the hidden feature matrix can be obtained. The specific method for obtaining the hidden feature matrix can be shown as formula (1):

[0168]

[0169] Among them, S (u) It can be used to represent the first decomposition set obtained by matrix decomposition of the dialogue robot evaluation vector matrix, S (u) It can include m data, that is, in, (i can be 1, 2, ..., m) can be used to characterize the sample user i (sample user demand feature i), the n-dimensional feedback vector (feedback feature) on all dialogue robots; S (i) It can be used to represent the second decomposition set obtained by matrix decomposing the dialogue robot evaluation vector matrix, S (i) It can include n data, that is, in, (j can be 1, 2, ..., n) can be used to represent the m-dimensional feedback vector (feedback feature) of the dialogue robot j on all users (all sample user demand features); R can be used to represent the dialogue robot evaluation vector matrix; It can be a loss function for hidden feature extraction, and can extract the hidden feature matrix U and the hidden feature matrix V from the dialogue robot evaluation vector matrix R; It can be a functional factor connecting users (sample user demand features) and hidden features (e.g., hidden feature matrix U), or connecting the dialogue robot and hidden features (e.g., hidden feature matrix V); β and δ are trade-off parameters; λ is the regularization parameter; Y is the additional auxiliary information matrix of the dialogue robot.

[0170] That is, through the dialogue robot evaluation vector matrix R, the first decomposition set S (u) , the second decomposition set S (i) As well as additional auxiliary information Y, hidden features (such as hidden features U and hidden features V) can be learned. The process of learning hidden features can be implemented by formula (1).

[0171] Furthermore, based on the dialogue robot evaluation vector R and the hidden feature matrix, the correlation matrix between the sample user demand features and the hidden features can be obtained. The specific method for obtaining the correlation matrix can be shown in formula (2):

[0172]

[0173] Among them, Iij It can be used to represent the non-empty real indicator matrix in the dialogue robot evaluation vector R; u i and v j It can be used to represent the hidden features extracted from the sample user demand features and the hidden layer of the conversation robot respectively; It can be used to characterize the correlation matrix between sample user demand features and hidden features.

[0174] Furthermore, based on the dialogue robot evaluation vector matrix and the correlation matrix, a loss function can be generated. The specific method can be shown in formula (3):

[0175]

[0176] Among them, L can be used to represent the loss function; α1 and α2 can be used to represent the trade-off parameters; Auxiliary information matrix that can be used to characterize the demand characteristics of sample users; Auxiliary information matrix that can be used to characterize the conversational robot; g reg Can be used to characterize the regularization term, g reg It can be shown as formula (4):

[0177]

[0178] Among them, W l 、V l 、W l ' and V l ′ It can be the weight matrix in the Denoising AutoEncoder (DAE).

[0179] The denoising autoencoder (DAE) model is a deep learning model. For ease of understanding, the DAE model is explained below: The DAE model uses unsupervised learning to efficiently extract and represent features from high-dimensional data. Its key feature is that the encoder creates one or more hidden layers containing a low-dimensional vector representation of the input data. The DAE model also contains a decoder, which reconstructs the input data from the low-dimensional vectors in the hidden layer. Subsequently, after training the neural network, the DAE model generates a low-dimensional vector representing the input data in the hidden layer. This helps with data classification, visualization, and storage. It should be noted that the DAE model adds random noise to the input data, specifically after the input layer, to enhance robustness.

[0180] It should be understood that in this application, after matrix decomposition of the dialogue robot evaluation vector matrix R to obtain the first decomposition set and the second decomposition set, the process of obtaining the hidden feature matrix according to the dialogue robot evaluation vector matrix R, the first decomposition set and the second decomposition set is mainly intended to connect the user (sample user demand feature) and the dialogue robot through the hidden features of the hidden layer. Among them, the hidden feature can be understood as the feature obtained by performing latent semantic matrix decomposition on the dialogue robot evaluation vector matrix R. For ease of understanding, the latent semantic matrix decomposition will be explained below using the latent semantic model (Latent Factor Model, LFM) as an example. The core idea of the latent semantic model is to connect user interests and items through latent features, which can reflect the process of performing latent semantic matrix decomposition on the dialogue robot evaluation vector matrix R to obtain hidden features, thereby connecting users and dialogue robots through hidden features. The user interests here can be understood as the sample user demand features in this application, and the items can be understood as dialogue robots.

[0181] The process of linking user interests and items through latent features can be divided into three parts: mapping items to latent categories; then determining the user's interest in the latent categories; and finally, selecting and recommending items from the categories of interest to the user. This is an automatic clustering based on user behavior statistics. Therefore, the LFM model proposes automatically identifying those categories based on the data and then making personalized recommendations. Latent semantic analysis technology uses automatic clustering based on user behavior statistics to effectively link user interests and items through latent features. Since its inception, latent semantic analysis technology has produced many well-known models and methods, including those related to recommendation technology, such as the latent class model, latent topic model, and matrix factorization.

[0182] The LFM model draws on the idea of linear regression to seek the optimal implicit vector representation of users and items by minimizing the square of the observed data. The matrix decomposition method calculates the interest of user u in item i through formula (5):

[0183]

[0184] Among them, r ui is the real rating of user u on item i, (∥q i ∥ 2 +∥p u ∥ 2 ) is the regularization term to prevent overfitting, and λ is the regularization coefficient. Assuming that the input rating matrix is R, which is an m×n dimensional matrix, then by optimizing formula (5), the user's feature matrix p can be obtained.u and item feature matrix q i The optimization method can be cross least squares or stochastic gradient descent.

[0185] It should be understood that, comparing formula (1) with formula (5), the dialogue robot evaluation vector matrix R in formula (1) can correspond to r in formula (5) ui ; In formula (1) It can be corresponded to (∥q i ∥ 2 +∥p u ∥ 2 ); Formula (5) is obtained by the user's feature matrix p u and item feature matrix q i To calculate user u's interest in item i, formula (1) uses the hidden feature matrix U and the hidden feature matrix V to calculate the connection between the sample user's demand characteristics and the dialogue robot. Furthermore, formula (1) can be used to optimize the obtained hidden feature matrix U and the hidden feature matrix V multiple times, thereby optimizing the connection between the sample user's demand characteristics and the dialogue robot.

[0186] Furthermore, after generating the loss function L, the initial recommendation model can be trained according to the loss function L to obtain the recommendation model. The specific method is that after the initial recommendation model outputs the predicted dialogue robot corresponding to each sample user demand feature, the predicted dialogue robot s can be obtained. i The first hidden feature of the conversational robot j The second hidden feature of ; Among them, predicting the conversation robot s i The sample user demand feature K output by the initial recommendation model q Corresponding predictive dialogue robot; dialogue robot s j For at least two conversational robots, with sample user demand features K q A conversation robot with a sample evaluation vector label between them; i and j are both integers less than or equal to N, N is the total number of at least two conversation robots; q is an integer less than or equal to M, M is the total number of at least two sample user demand features; for example, for sample user demand feature a, the conversation robot predicted by the initial recommendation model is conversation robot B, and the manually labeled conversation robot is conversation robot C. The conversation robot C is the labeled conversation robot corresponding to the sample user demand feature a, and there is a labeled evaluation vector (i.e., sample evaluation vector label) between the conversation robot C and the sample user demand feature a, then the predicted evaluation vector between the sample user demand feature a and the predicted conversation robot B, as well as the sample evaluation vector label between the sample user demand feature a and the labeled conversation robot C can be obtained.

[0187] Furthermore, the first hidden feature, the second hidden feature, the predicted evaluation vector, and the sample evaluation vector label are substituted into the aforementioned loss function (e.g., loss function L) to generate a loss function value. If the loss function value does not meet the model convergence condition, the initial association relationship of the initial recommendation model can be adjusted based on the loss function value to obtain a recommendation model that includes the association relationship.

[0188] In the model training process of the embodiment of the present application, by marking the association between the sample user demand features and the dialogue robot, and then marking an evaluation vector as a sample evaluation vector label between each sample user demand feature and each corresponding labeled dialogue robot; a dialogue robot evaluation vector matrix is generated through the sample evaluation vector label, and then the dialogue robot evaluation vector matrix is matrix decomposed, and the sample user demand features and the hidden features of the dialogue robot are extracted to generate a loss function. The initial recommendation model can be trained according to the loss function, so that the initial association relationship in the initial recommendation model can be closer and closer to the labeled association relationship between the sample user demand features and the dialogue robot. Therefore, the recommendation model obtained after the training is completed can more accurately recommend a suitable dialogue robot for the target user demand features.

[0189] For easier understanding, see Figure 6 , Figure 6 This is a schematic diagram of the association between model training and application provided in the embodiment of this application. Figure 6 As shown, the sample user demand features and the dialogue robot identification information are used as training data and input into the initial recommendation model. The initial recommendation model can output one or more predicted dialogue robots corresponding to each sample user demand feature based on the initial association relationship. The initial recommendation model will also predict a predicted evaluation vector between a sample user demand feature and its corresponding predicted dialogue robot; based on the predicted evaluation vector between each sample user demand feature and each corresponding predicted dialogue robot, as well as the sample evaluation vector label between each sample user demand feature and each corresponding labeled dialogue robot and the loss function in the initial recommendation model, a loss function value (i.e., prediction error) can be generated. Based on the loss function value, the initial association relationship in the initial recommendation model can be adjusted. It can be understood that by minimizing the loss function value, the prediction result (predicted dialogue robot) output by the initial recommendation model can be made more and more accurate.

[0190] It should be understood that when the loss function value meets the model convergence condition (for example, when the loss function value is less than the error threshold), a round of model training can be completed and a recommendation model can be obtained. The recommendation model can be put into an application scenario (for example, a conversational robot recommendation scenario), and the application data can be input into the recommendation model. The prediction results can be obtained through the recommendation model.

[0191] Further, see Figure 7 , Figure 7 This is a structural diagram of a data processing device provided in an embodiment of the present application. The data processing device may be a computer program (including program code) running on a computer device, for example, the data processing device is an application software; the data processing device may be used to execute Figure 3 As shown in the method. Figure 7 As shown, the data processing device 1 may include: a demand information acquisition module 11, a demand feature generation module 12, a feature input module 13, a dialogue robot determination module 14 and a business function execution module 15.

[0192] The demand information acquisition module 11 is used to respond to the target user's information input operation on the dialogue robot selection interface and obtain the target user's dialogue demand information;

[0193] A demand feature generating module 12 is used to generate target user demand features corresponding to target user dialogue demand information;

[0194] The feature input module 13 is used to input the target user's demand features into the recommendation model; the recommendation model includes the association relationship between at least two user demand features and at least two dialogue robots;

[0195] A conversational robot determination module 14 is configured to determine a target conversational robot from among at least two conversational robots based on a matching relationship and an association relationship between a target user's demand feature and at least two user demand features in a recommendation model;

[0196] The business function execution module 15 is used to input the dialogue operation information of the target user into the target dialogue robot, and trigger the target dialogue robot to execute the dialogue business function.

[0197] The specific implementation of the demand information acquisition module 11, the demand feature generation module 12, the feature input module 13, the dialogue robot determination module 14 and the business function execution module 15 can be found in Figure 3 The description of steps S101 to S105 in the corresponding embodiment will not be repeated here.

[0198] Among them, information input operations include type input operations, call input operations, and domain input operations;

[0199] See Figure 7 The demand information acquisition module 11 may include: an information acquisition unit 111 and a demand information generation unit 112 .

[0200] An information acquisition unit 111 is configured to respond to a type input operation on a dialogue robot selection interface and acquire dialogue intention type information;

[0201] The information acquisition unit 111 is further configured to respond to a call input operation on the dialogue robot selection interface and acquire information on the number of times the robot has been called;

[0202] The information acquisition unit 111 is further configured to respond to a domain input operation on the dialogue robot selection interface and acquire dialogue application domain information;

[0203] The demand information generating unit 112 is used to generate target user dialogue demand information based on dialogue intention type information, robot call count information, and dialogue application field information.

[0204] The specific implementation of the information acquisition unit 111 and the demand information generation unit 112 can be found in the above Figure 3 The description of step S101 in the corresponding embodiment will not be repeated here.

[0205] See Figure 7 The demand feature generation module 12 may include: a key information extraction unit 121 , a regularization processing unit 122 and a vector conversion unit 123 .

[0206] The key information extraction unit 121 is used to extract key fields from the target user's conversation demand information to obtain key user demand information;

[0207] Regularization processing unit 122, used to perform regularization processing on key user demand information to obtain regularized user demand information;

[0208] The vector conversion unit 123 is used to perform vector conversion on the regular user demand information to obtain target user demand features corresponding to the target user dialogue demand information.

[0209] The specific implementation of the key information extraction unit 121, the regularization processing unit 122 and the vector conversion unit 123 can be found in the above Figure 3 The description of step S102 in the corresponding embodiment will not be repeated here.

[0210] See Figure 7The dialogue robot determination module 14 may include: a matching requirement feature acquisition unit 141, an associated robot determination unit 142, an associated evaluation vector acquisition unit 143, an associated evaluation vector acquisition unit 144 and a target robot determination unit 145.

[0211] The matching requirement feature acquisition unit 141 is configured to acquire a matching relationship between a target user requirement feature and at least two user requirement features, and use a user requirement feature with a successful matching relationship as a matching user requirement feature;

[0212] An associated robot determining unit 142 is configured to obtain, from the at least two conversational robots, a conversational robot associated with matching user demand characteristics based on the association relationship, as an associated conversational robot;

[0213] The associated evaluation vector acquisition unit 143 is used to acquire an associated user evaluation vector associated with the associated dialogue robot and the target user's demand characteristics; one associated user evaluation vector corresponds to one associated dialogue robot;

[0214] The associated evaluation vector acquisition unit 144 is further configured to acquire an associated user evaluation vector having a maximum vector modulus among the associated user evaluation vectors;

[0215] The target robot determining unit 145 is configured to determine the associated dialogue robot corresponding to the associated user evaluation vector having the largest vector modulus as the target dialogue robot.

[0216] The specific implementation of the matching requirement feature acquisition unit 141, the associated robot determination unit 142, the associated evaluation vector acquisition unit 143, the associated evaluation vector acquisition unit 144 and the target robot determination unit 145 can be found in the above Figure 3 The description of step S104 in the corresponding embodiment will not be repeated here.

[0217] See Figure 7 The data processing device 1 may further include: a similarity determination module 16 and a matching relationship determination module 17 .

[0218] A similarity determination module 16 is configured to determine the similarity between the target user's demand feature and each of the at least two user demand features;

[0219] The matching relationship determination module 17 is configured to determine the matching relationship between the user demand feature and the target user demand feature whose similarity is greater than a first similarity threshold as a successful matching relationship.

[0220] The specific implementation of the similarity determination module 16 and the matching relationship determination module 17 can be found in the above Figure 3 The description of step S104 in the corresponding embodiment will not be repeated here.

[0221] See Figure 7 The data processing device 1 may further include: a usage data acquisition module 18 , a fitness determination module 19 , an updated evaluation vector generation module 20 , and an evaluation vector updating module 21 .

[0222] Usage data acquisition module 18, used to obtain the target user's usage behavior data for the target conversation robot;

[0223] A compatibility determination module 19 is configured to determine the compatibility between the target conversational robot and the target user's demand characteristics based on the usage behavior data, and to generate a target user evaluation vector associated with the target conversational robot and the target user's demand characteristics based on the compatibility;

[0224] An updated evaluation vector generating module 20 is configured to generate an updated user evaluation vector based on the target user evaluation vector and the user evaluation vectors associated with the target dialogue robot and the matching user demand features if the similarity between the target user demand features and the matching user demand features is greater than a second similarity threshold;

[0225] The evaluation vector updating module 21 is used to update the user evaluation vector associated with the target dialogue robot and the matching user demand feature according to the updated user evaluation vector.

[0226] The specific implementation of the data acquisition module 18, the fitness determination module 19, the updated evaluation vector generation module 20 and the evaluation vector update module 21 can be found in the above Figure 3 In the corresponding embodiment, the description of updating the user evaluation vector in step S105 will not be repeated here.

[0227] See Figure 7 The business function execution module 15 may include: a robot display unit 151 , a text information conversion unit 152 and a business function triggering unit 153 .

[0228] The robot display unit 151 is used to create a robot information management interface and display the target conversation robot in the robot information management interface;

[0229] A text information conversion unit 152 is configured to respond to a target user's robot dialogue operation on the robot information management interface, obtain dialogue operation information, and convert the dialogue operation information into text information;

[0230] The service function triggering unit 153 is configured to input text information into a target dialogue robot and trigger the target dialogue robot to execute a dialogue service function associated with the text information.

[0231] The specific implementation of the robot display unit 151, the text information conversion unit 152 and the business function triggering unit 153 can be found in the above Figure 3 The description of step S105 in the corresponding embodiment will not be repeated here.

[0232] In an embodiment of the present application, by obtaining the target user's target user dialogue demand information for the dialogue robot, a target user demand feature can be generated, and the target user demand feature is input into the recommendation model. The recommendation model can automatically determine the target dialogue robot that matches the target user demand feature and recommend the target dialogue robot to the target user. Among them, because the recommendation model contains the association relationship between at least two user demand features and at least two dialogue robots, and the target dialogue robot determined by the recommendation model is determined based on the matching relationship between the target user demand feature and the at least two user demand features and the association relationship, the target dialogue robot is also matched with the target user demand feature, that is, the target dialogue robot meets the needs of the target user. It can be seen from this that after obtaining the target user's dialogue demand information, the present application can automatically recommend a dialogue robot to the target user based on the association relationship in the recommendation model, which can improve the recommendation efficiency; and the entire recommendation process does not require human participation, which reduces offline communication time and reduces the manpower and material resources of customizing dialogue robots, thereby reducing costs. At the same time, because the target conversational robot determined by the recommendation model is determined based on the user evaluation vector of matching user demand characteristics similar to the target user demand characteristics, and the user evaluation vector corresponding to each matching user demand characteristic will be updated multiple times based on the usage behavior data of different users, so each user evaluation vector of matching user demand characteristics can accurately represent the degree of suitability between a conversational robot and the user demand characteristics, then the target conversational robot determined based on the user evaluation vector will also have a higher accuracy (that is, the target conversational robot will be more suitable for the target user demand characteristics).

[0233] Further, see Figure 8 , Figure 8 This is a schematic diagram of a computer device provided in an embodiment of the present application. Figure 8 As shown, the computer device 1000 can be Figure 3In the corresponding embodiment, the user terminal, the computer device 1000 may include: at least one processor 1001, such as a CPU, at least one network interface 1004, a user interface 1003, a memory 1005, and at least one communication bus 1002. The communication bus 1002 is used to realize the connection and communication between these components. The user interface 1003 may include a display screen (Display), a keyboard (Keyboard), and the network interface 1004 may optionally include a standard wired interface, a wireless interface (such as a WI-FI interface). The memory 1005 may be a high-speed RAM memory, or a non-volatile memory (non-volatile memory), such as at least one disk storage. The memory 1005 may optionally also be at least one storage device located away from the aforementioned processor 1001. As Figure 8 As shown, the memory 1005 as a computer storage medium may include an operating system, a network communication module, a user interface module, and a device control application program.

[0234] exist Figure 8 In the computer device 1000 shown, the network interface 1004 is mainly used for network communication with the service server; the user interface 1003 is mainly used to provide an interface for user input; and the processor 1001 can be used to call the device control application stored in the memory 1005 to achieve:

[0235] Respond to the target user's input operation on the dialogue robot selection interface to obtain the target user's dialogue demand information;

[0236] Generate target user demand features corresponding to target user dialogue demand information;

[0237] Inputting target user demand characteristics into a recommendation model; the recommendation model includes an association relationship between at least two user demand characteristics and at least two dialogue robots;

[0238] In the recommendation model, a target dialogue robot is determined from among the at least two dialogue robots based on a matching relationship and an association relationship between the target user's demand characteristics and the at least two user demand characteristics;

[0239] The target user's dialogue operation information is input into the target dialogue robot, triggering the target dialogue robot to perform dialogue business functions.

[0240] It should be understood that the computer device 1000 described in the embodiment of the present application can execute the above Figure 3 The description of the data processing method in the corresponding embodiment can also be performed as described above. Figure 7The description of the data processing device 1 in the corresponding embodiment will not be repeated here. In addition, the description of the beneficial effects of adopting the same method will not be repeated here either.

[0241] In addition, it should be noted that: the embodiment of the present application also provides a computer-readable storage medium, and the computer-readable storage medium stores a computer program executed by the computer device 1000 for data processing mentioned above, and the computer program includes program instructions. When the above-mentioned processor executes the above-mentioned program instructions, it can execute the above-mentioned Figure 3 Therefore, the description of the data processing method in the corresponding embodiment will not be repeated here. In addition, the description of the beneficial effects of adopting the same method will not be repeated. For technical details not disclosed in the computer-readable storage medium embodiment involved in this application, please refer to the description of the method embodiment of this application.

[0242] The computer-readable storage medium may be the data processing device provided in any of the aforementioned embodiments or the internal storage unit of the computer device, such as the hard disk or memory of the computer device. The computer-readable storage medium may also be an external storage device of the computer device, such as a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, etc. equipped on the computer device. Furthermore, the computer-readable storage medium may also include both the internal storage unit of the computer device and an external storage device. The computer-readable storage medium is used to store the computer program and other programs and data required by the computer device. The computer-readable storage medium may also be used to temporarily store data that has been output or is to be output.

[0243] In one aspect of the present application, a computer program product or computer program is provided. The computer program product or computer program includes computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the computer device to perform the method provided in one aspect of the embodiments of the present application.

[0244] Further, see Figure 9 , Figure 9 This is a structural diagram of a data processing device provided in an embodiment of the present application. The data processing device may be a computer program (including program code) running on a computer device, for example, the data processing device is an application software; the data processing device may be used to execute Figure 5 As shown in the method. Figure 9As shown, the data processing device 2 may include: an information acquisition module 200 , an information input module 210 , a prediction robot output module 220 , a prediction evaluation acquisition module 230 , an evaluation label acquisition module 240 and a relationship adjustment module 250 .

[0245] An information acquisition module 200 is configured to acquire at least two sample user demand features and obtain robot identification information of each of the at least two conversational robots;

[0246] An information input module 210 is configured to input at least two sample user demand characteristics and robot identification information into an initial recommendation model; the recommendation model includes an initial association relationship between the at least two sample user demand characteristics and the at least two conversational robots;

[0247] The prediction robot output module 220 is used to output the predicted dialogue robot corresponding to each sample user's demand characteristics based on the initial association relationship in the initial recommendation model;

[0248] A prediction evaluation acquisition module 230 is used to obtain a prediction evaluation vector between the predicted dialogue robot and at least two sample user demand features;

[0249] An evaluation label acquisition module 240 is used to obtain sample evaluation vector labels between at least two sample user demand features and at least two dialogue robots;

[0250] The relationship adjustment module 250 is used to adjust the initial association relationship in the initial recommendation model based on the predicted evaluation vector and the sample evaluation vector label to obtain a recommendation model containing the association relationship; the recommendation model is used to identify the target dialogue robot that matches the target user's demand characteristics among at least two dialogue robot models.

[0251] The specific implementation of the information acquisition module 200, the information input module 210, the prediction robot output module 220, the prediction evaluation acquisition module 230, the evaluation label acquisition module 240 and the relationship adjustment module 250 can be found in Figure 5 The description of steps S201 to S206 in the corresponding embodiment will not be repeated here.

[0252] See Figure 9 The relationship adjustment module 250 may include: a loss function acquisition unit 2501, a loss value generation unit 2502 and a relationship adjustment unit 2503.

[0253] A loss function acquisition unit 2501 is used to acquire a loss function;

[0254] A loss value generating unit 2502 is configured to generate a loss function value based on the loss function, the predicted evaluation vector, and the sample evaluation vector label;

[0255] The relationship adjustment unit 2503 is used to adjust the initial association relationship according to the loss function value if the loss function value does not meet the model convergence condition, so as to obtain a recommendation model including the association relationship.

[0256] The specific implementation of the loss function acquisition unit 2501, the loss value generation unit 2502 and the relationship adjustment unit 2503 can be found in the above Figure 5 The description of adjusting the initial association relationship in step S206 in the corresponding embodiment will not be repeated here.

[0257] See Figure 9 The loss value generating unit 2502 may include: a hidden feature acquiring unit 25021 and a loss value generating unit 25022.

[0258] Hidden feature acquisition unit 25021, used to obtain the predicted dialogue robot s i The first hidden feature of the conversational robot j The second hidden feature of predictive conversational robots i The sample user demand feature K output by the initial recommendation model q Corresponding predictive dialogue robot; dialogue robot s j For at least two conversational robots, with sample user demand features K q There are two dialogue robots with sample evaluation vector labels between them; i and j are both integers less than or equal to N, where N is the total number of at least two dialogue robots; q is an integer less than or equal to M, where M is the total number of at least two sample user demand features;

[0259] The loss value generating unit 25022 is used to generate a loss function value according to the first hidden feature, the second hidden feature, the predicted evaluation vector, the sample evaluation vector label and the loss function.

[0260] The specific implementation of the hidden feature acquisition unit 25021 and the loss value generation unit 25022 can be found in the above Figure 5 The description of the loss value generation in step S206 in the corresponding embodiment will not be repeated here.

[0261] Among them, the loss function acquisition unit 2501 may include: a vector matrix generation subunit 25011, a hidden feature matrix determination subunit 25012 and a loss function generation subunit 25013.

[0262] A vector matrix generation subunit 25011 is configured to generate a dialogue robot evaluation vector matrix based on at least two sample user demand features and sample evaluation vector labels of at least two dialogue robots;

[0263] A hidden feature matrix determination subunit 25012 is configured to determine hidden feature matrices of at least two dialogue robots based on the dialogue robot evaluation vector matrix;

[0264] The loss function generation subunit 25013 is used to generate a loss function based on the dialogue robot evaluation vector matrix and the hidden feature matrix.

[0265] The specific implementation of the vector matrix generating subunit 25011, the hidden feature matrix determining subunit 25012 and the loss function generating subunit 25013 can be found in the above Figure 5 The description of the loss function generation in step S206 in the corresponding embodiment will not be repeated here.

[0266] The hidden feature matrix determination subunit 25012 is further configured to decompose the dialogue robot evaluation vector matrix to obtain a first decomposition set and a second decomposition set; the first decomposition set includes feedback features of each sample user's demand features for at least two dialogue robots; and the second decomposition set includes feedback features of each dialogue robot for at least two sample user's demand features.

[0267] The hidden feature matrix determination subunit 25012 is further used to determine the hidden feature matrix based on the dialogue robot evaluation vector matrix, the first decomposition set and the second decomposition set.

[0268] In the model training process of the embodiment of the present application, by marking the association between the sample user demand features and the dialogue robot, and then marking an evaluation vector as a sample evaluation vector label between each sample user demand feature and each corresponding labeled dialogue robot; a dialogue robot evaluation vector matrix is generated through the sample evaluation vector label, and then the dialogue robot evaluation vector matrix is matrix decomposed, and the sample user demand features and the hidden features of the dialogue robot are extracted to generate a loss function. The initial recommendation model can be trained according to the loss function, which can make the initial association relationship in the initial recommendation model closer and closer to the labeled association relationship between the sample user demand features and the dialogue robot. Therefore, the recommendation model obtained after the training is completed can accurately recommend a suitable dialogue robot for the target user demand features.

[0269] Further, see Figure 10 , Figure 10 This is a schematic diagram of a computer device provided in an embodiment of the present application. Figure 10As shown, the computer device 4000 can be the above Figure 5 In the corresponding embodiment, the user terminal, the computer device 4000 may include: at least one processor 4001, such as a CPU, at least one network interface 4004, a user interface 4003, a memory 4005, and at least one communication bus 4002. The communication bus 4002 is used to realize the connection and communication between these components. The user interface 4003 may include a display screen (Display), a keyboard (Keyboard), and the network interface 4004 may optionally include a standard wired interface, a wireless interface (such as a WI-FI interface). The memory 4005 may be a high-speed RAM memory, or a non-volatile memory (non-volatile memory), such as at least one disk storage. The memory 4005 may optionally also be at least one storage device located away from the aforementioned processor 4001. As Figure 10 As shown, the memory 4005 as a computer storage medium may include an operating system, a network communication module, a user interface module, and a device control application.

[0270] exist Figure 10 In the computer device 4000 shown, the network interface 4004 is mainly used to communicate with the user terminal over the network; the user interface 4003 is mainly used to provide an interface for user input; and the processor 4001 can be used to call the device control application stored in the memory 4005 to achieve:

[0271] Obtain at least two sample user demand features, and obtain robot identification information of each of the at least two dialogue robots;

[0272] Inputting at least two sample user demand features and robot identification information into an initial recommendation model; the recommendation model includes an initial association relationship between the at least two sample user demand features and the at least two conversational robots;

[0273] Based on the initial association relationship in the initial recommendation model, the predicted conversational robot corresponding to each sample user's demand characteristics is output;

[0274] Obtain a prediction evaluation vector between the predicted conversational robot and at least two sample user demand features;

[0275] Obtain sample evaluation vector labels between at least two sample user demand features and at least two dialogue robots;

[0276] Based on the predicted evaluation vector and the sample evaluation vector label, the initial association relationship in the initial recommendation model is adjusted to obtain a recommendation model containing the association relationship; the recommendation model is used to identify the target dialogue robot that matches the target user's demand characteristics among at least two dialogue robot models.

[0277] It should be understood that the computer device 4000 described in the embodiment of the present application can execute the above Figure 5 The description of the data processing method in the corresponding embodiment can also be performed as described above. Figure 9 The description of the data processing device 2 in the corresponding embodiment will not be repeated here. In addition, the description of the beneficial effects of adopting the same method will not be repeated here either.

[0278] In addition, it should be noted that: the embodiment of the present application also provides a computer-readable storage medium, and the computer-readable storage medium stores a computer program executed by the computer device 4000 for data processing mentioned above, and the computer program includes program instructions. When the processor executes the above program instructions, it can execute the above Figure 5 The description of the data processing method in the corresponding embodiment will therefore not be repeated here. In addition, the description of the beneficial effects of adopting the same method will not be repeated. For technical details not disclosed in the computer-readable storage medium embodiment involved in this application, please refer to the description of the method embodiment of this application.

[0279] The computer-readable storage medium may be the data processing device provided in any of the aforementioned embodiments or the internal storage unit of the computer device, such as the hard disk or memory of the computer device. The computer-readable storage medium may also be an external storage device of the computer device, such as a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, etc. equipped on the computer device. Furthermore, the computer-readable storage medium may also include both the internal storage unit of the computer device and an external storage device. The computer-readable storage medium is used to store the computer program and other programs and data required by the computer device. The computer-readable storage medium may also be used to temporarily store data that has been output or is to be output.

[0280] In one aspect of the present application, a computer program product or computer program is provided. The computer program product or computer program includes computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the computer device to perform the method provided in one aspect of the embodiments of the present application.

[0281] For further information, see Figure 11 , Figure 11 This is a structural diagram of a data processing system provided in an embodiment of the present application. The data processing system 3 may include a data processing device 1a and a data processing device 2a. The data processing device 1a may be the above-mentioned Figure 7 The data processing device 1 in the corresponding embodiment can be understood that the data processing device 1a can be integrated into the above Figure 3 The data processing device 2a can be the user terminal in the embodiment described above. Figure 9 The data processing device 2 in the corresponding embodiment can be understood as follows: the data processing device 2a can be integrated into the above Figure 5 Therefore, the user terminal in the corresponding embodiment is not described in detail here. In addition, the description of the beneficial effects of adopting the same method is not described in detail. For technical details not disclosed in the data transmission system embodiment involved in this application, please refer to the description of the method embodiment of this application.

[0282] The terms "first", "second", etc. in the description, claims, and drawings of the embodiments of the present application are used to distinguish different objects, rather than to describe a specific order. In addition, the terms "comprising" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, apparatus, product, or device comprising a series of steps or units is not limited to the listed steps or modules, but may optionally include steps or modules not listed, or may optionally include other step units inherent to these processes, methods, apparatuses, products, or devices.

[0283] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described in terms of function in the above description. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0284] The methods and related devices provided by the embodiments of the present application are described with reference to the method flow charts and / or structural diagrams provided by the embodiments of the present application. Specifically, each process and / or block in the method flow charts and / or structural diagrams, as well as the combination of processes and / or blocks in the flow charts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the steps in the process. Figure 1 Schematic diagram of one or more processes and / or structures Figure 1 These computer program instructions can also be stored in a computer-readable memory that can guide a computer or other programmable data processing device to work in a specific way, so that the instructions stored in the computer-readable memory produce a product including the instruction device, which implements the function specified in the process. Figure 1 Schematic diagram of one or more processes and / or structures Figure 1 These computer program instructions can also be loaded onto a computer or other programmable data processing device, so that a series of operation steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process in the process. Figure 1 The flow or flows and / or structures illustrate the steps of the functions specified in one block or multiple blocks.

[0285] The above disclosure is only a preferred embodiment of the present application, and certainly cannot be used to limit the scope of rights of the present application. Therefore, equivalent changes made according to the claims of the present application are still within the scope covered by the present application.

Claims

1. A data processing method, characterized in that: include: Respond to the target user's input operation on the dialogue robot selection interface to obtain the target user's dialogue demand information; generating target user demand features corresponding to the target user dialogue demand information; Inputting the target user demand characteristics into a recommendation model; the recommendation model includes an association relationship between at least two user demand characteristics and at least two dialogue robots; By using the recommendation model, a matching relationship between the target user demand feature and the at least two user demand features is obtained from the at least two user demand features, and the user demand feature with a successful matching relationship is used as the matching user demand feature; based on the association relationship, a conversational robot associated with the matching user demand feature is obtained from the at least two conversational robots as an associated conversational robot; an associated user evaluation vector associated with the associated conversational robot and the target user demand feature is obtained; one associated user evaluation vector corresponds to one associated conversational robot; an associated user evaluation vector with the largest vector modulus is obtained from the associated user evaluation vectors; and the associated conversational robot corresponding to the associated user evaluation vector with the largest vector modulus is determined as the target conversational robot; The target user's dialogue operation information is input into the target dialogue robot to trigger the target dialogue robot to perform dialogue business functions.

2. The method according to claim 1, characterized in that The information input operation includes a type input operation, a call input operation, and a domain input operation; The response to the information input operation on the dialogue robot selection interface to obtain the target user dialogue demand information includes: In response to a type input operation on the dialogue robot selection interface, obtaining dialogue intention type information; In response to a call input operation on the dialogue robot selection interface, obtaining robot call count information; Responding to a domain input operation on the dialogue robot selection interface, obtaining dialogue application domain information; The target user dialogue demand information is generated based on the dialogue intention type information, the robot call count information, and the dialogue application field information.

3. The method according to claim 1, characterized in that Generating target user demand features corresponding to the target user dialogue demand information includes: Extract key fields from the target user conversation demand information to obtain key user demand information; Regularizing the key user demand information to obtain regularized user demand information; The regular user demand information is vector-converted to obtain target user demand features corresponding to the target user dialogue demand information.

4. The method according to claim 1, wherein Also includes: Determining a similarity between the target user demand feature and each of the at least two user demand features; The matching relationship between the user demand feature whose similarity is greater than the first similarity threshold and the target user demand feature is determined as the successful matching relationship.

5. The method according to claim 4, characterized in that Also includes: Obtaining usage behavior data of the target user for the target conversational robot; Determining, based on the usage behavior data, a degree of compatibility between the target conversational robot and the target user's demand characteristics, and generating, based on the degree of compatibility, a target user evaluation vector associated with the target conversational robot and the target user's demand characteristics; If the similarity between the target user demand feature and the matching user demand feature is greater than a second similarity threshold, generating an updated user evaluation vector based on the target user evaluation vector and the user evaluation vector associated with the target dialogue robot and the matching user demand feature; According to the updated user evaluation vector, the user evaluation vector associated with the target dialogue robot and the matching user demand feature is updated.

6. The method according to claim 1, characterized in that The step of inputting the target user's dialogue operation information into the target dialogue robot to trigger the target dialogue robot to perform a dialogue service function includes: Creating a robot information management interface, and displaying the target conversational robot in the robot information management interface; Responding to the target user's robot dialogue operation on the robot information management interface, obtaining dialogue operation information, and converting the dialogue operation information into text information; The text information is input into the target dialogue robot, triggering the target dialogue robot to execute a dialogue service function associated with the text information.

7. A data processing method, characterized in that: include: Obtain at least two sample user demand features, and obtain robot identification information of each of the at least two dialogue robots; Inputting the at least two sample user demand features and the robot identification information into an initial recommendation model; the initial recommendation model includes an initial association relationship between the at least two sample user demand features and the at least two dialogue robots; Outputting a predicted conversational robot corresponding to each sample user's demand characteristics through the initial association relationship in the initial recommendation model; Obtaining a prediction evaluation vector between the prediction dialogue robot and the at least two sample user demand features; Obtaining sample evaluation vector labels between the at least two sample user demand features and the at least two dialogue robots; Adjusting the initial association relationship in the initial recommendation model based on the predicted evaluation vector and the sample evaluation vector label to obtain a recommendation model that includes an association relationship between at least two user demand characteristics and the at least two conversational robots; the recommendation model is used to identify a target conversational robot that matches the target user demand characteristics among the at least two conversational robot models; The process of the recommendation model identifying a target dialogue robot that matches the target user demand feature includes: obtaining a matching relationship between the target user demand feature and the at least two user demand features among the at least two user demand features, and taking the user demand feature with a successful matching relationship as the matching user demand feature; according to the association relationship, obtaining a dialogue robot associated with the matching user demand feature among the at least two dialogue robots as an associated dialogue robot; obtaining an associated user evaluation vector associated with the associated dialogue robot and the target user demand feature; one associated user evaluation vector corresponds to one associated dialogue robot; obtaining an associated user evaluation vector with the largest vector modulus among the associated user evaluation vectors; and determining the associated dialogue robot corresponding to the associated user evaluation vector with the largest vector modulus as the target dialogue robot.

8. The method according to claim 7, characterized in that The adjusting the initial association relationship in the initial recommendation model according to the predicted evaluation vector and the sample evaluation vector label to obtain a recommendation model including the association relationship includes: Get the loss function; Generate a loss function value according to the loss function, the predicted evaluation vector, and the sample evaluation vector label; If the loss function value does not meet the model convergence condition, the initial association relationship is adjusted according to the loss function value to obtain a recommendation model including the association relationship.

9. The method according to claim 8, characterized in that Generating a loss function value according to the loss function, the predicted evaluation vector, and the sample evaluation vector label includes: Get predictive conversational bots i The first hidden feature of the conversational robot j The second hidden feature of the predictive dialogue robot s i The sample user demand feature K output by the initial recommendation model q Corresponding prediction dialogue robot; the dialogue robot s j For the at least two dialogue robots, the sample user demand feature K q A conversational robot with a sample evaluation vector label between them; i and j are both integers less than or equal to N, N is the total number of the at least two conversational robots; q is an integer less than or equal to M, M is the total number of the at least two sample user demand features; The loss function value is generated according to the first hidden feature, the second hidden feature, the predicted evaluation vector, the sample evaluation vector label, and the loss function.

10. The method according to claim 8, characterized in that The obtaining of the loss function includes: Generating a dialogue robot evaluation vector matrix based on the sample evaluation vector labels between the at least two sample user demand features and the at least two dialogue robots; Determining hidden feature matrices of the at least two dialogue robots based on the dialogue robot evaluation vector matrix; A loss function is generated according to the dialogue robot evaluation vector matrix and the hidden feature matrix.

11. The method according to claim 10, characterized in that The step of determining the hidden feature matrices of the at least two dialogue robots based on the dialogue robot evaluation vector matrix includes: Decomposing the dialogue robot evaluation vector matrix to obtain a first decomposition set and a second decomposition set; the first decomposition set includes feedback features of each sample user demand feature for the at least two dialogue robots; the second decomposition set includes feedback features of each dialogue robot for the at least two sample user demand features; The hidden feature matrix is determined according to the dialogue robot evaluation vector matrix, the first decomposition set, and the second decomposition set.

12. A data processing device, characterized in that: include: A demand information acquisition module is used to respond to the target user's information input operation on the dialogue robot selection interface and obtain the target user's dialogue demand information; A demand feature generating module, configured to generate target user demand features corresponding to the target user dialogue demand information; A feature input module, configured to input the target user's demand features into a recommendation model; the recommendation model includes an association relationship between at least two user demand features and at least two dialogue robots; A dialogue robot determination module is configured to obtain, from the at least two user demand features, a matching relationship between the target user demand feature and the at least two user demand features, and use the user demand feature with a successful matching relationship as the matching user demand feature; based on the association relationship, obtain, from the at least two dialogue robots, a dialogue robot associated with the matching user demand feature as the associated dialogue robot; obtain an associated user evaluation vector associated with the associated dialogue robot and the target user demand feature; one associated user evaluation vector corresponds to one associated dialogue robot; obtain an associated user evaluation vector with a maximum vector modulus length from the associated user evaluation vectors; and determine the associated dialogue robot corresponding to the associated user evaluation vector with the maximum vector modulus length as the target dialogue robot; The business function execution module is used to input the dialogue operation information of the target user into the target dialogue robot, triggering the target dialogue robot to execute the dialogue business function.

13. A computer device, characterized in that: include: processor, memory, and network interface; The processor is connected to the memory and the network interface, wherein the network interface is used to provide a network communication function, the memory is used to store program code, and the processor is used to call the program code to execute the method described in any one of claims 1-11.

14. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a computer program, wherein the computer program includes program instructions. When the program instructions are executed by a processor, the method according to any one of claims 1 to 11 is executed.

Images