Business data recommendation method, storage medium and device

Abstract

The embodiment of the present application discloses a business data recommendation method, storage medium and device. The method includes: a first device responds to an information input operation for a target business, and sends the business consultation information determined by the information input operation to a second device; the second device obtains M target material data matching the business consultation information in the target business, displays the M target material data, where M is a positive integer; the second device responds to a selection operation for the M target material data, and determines the target material data determined by the selection operation as consultation feedback information among the M target material data, and the second device sends the consultation feedback information to the first device. The present application can improve the accuracy and efficiency of business data recommendation.

Description

Technical Field

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

[0002] In recent years, with the rapid development of various industries, the service businesses provided by various industries to users have also increased day by day, and the number is huge. How to provide consulting services to users quickly and accurately and help users carry out the business they need has become a problem that institutions / personnel in various industries need to solve.

[0003] In the prior art, when a user asks a question online, relevant business personnel are required to perform manual inquiries to obtain answers to the user's question and reply to the answer to the user; however, relevant business personnel usually need to face a large amount of consulting work, and manually querying the answers to questions takes a lot of time, which results in too low processing efficiency of the consulting business; and the answers retrieved by the business personnel are personally subjective and may differ from the answers the user wants, which results in too low accuracy of the consulting business response. Summary of the Invention

[0004] The technical problem to be solved by the embodiments of the present application is to provide a business data recommendation method, storage medium and device, which can improve the accuracy and efficiency of business data recommendation.

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

[0006] The first device responds to the information input operation for the target business and sends the business consultation information determined by the information input operation to the second device;

[0007] The second device obtains M target material data matching the business consultation information in the target business and displays the M target material data; M is a positive integer;

[0008] In response to the selection operation on the M target material data, the second device determines the target material data determined by the selection operation as the consultation feedback information among the M target material data;

[0009] The second device sends the consultation feedback information to the first device.

[0010] The second device obtains M target material data matching the business consultation information in the target business, including:

[0011] The second device receives the business consulting information sent by the first device, and obtains an information feature vector corresponding to the business consulting information;

[0012] The second device obtains the intent probability between the information feature vector and N candidate intents respectively; N is a positive integer;

[0013] The second device determines the target intent corresponding to the business consultation information from the N candidate intents based on the intent probability;

[0014] The second device obtains M target material data matching the target intent in the target service.

[0015] The information feature vector corresponding to the business consultation information is obtained, including:

[0016] The second device obtains consulting keywords in the business consulting information, performs vector conversion on the consulting keywords, and obtains word vectors corresponding to the consulting keywords;

[0017] The second device obtains the information feature vector corresponding to the consultation keyword based on the semantic information corresponding to the word vector.

[0018] The second device obtains the intent probabilities between the information feature vector and the N candidate intents, including:

[0019] The second device inputs the information feature vector into the neural network layer in the intent recognition model, and obtains a prediction vector corresponding to the information feature vector based on the weight matrix corresponding to the neural network layer;

[0020] The second device determines, in a classifier of the intent recognition model, intent probabilities between the prediction vector and the N candidate intents.

[0021] Among them, M target material data are displayed, including:

[0022] The second device obtains the material attribute characteristics corresponding to the M target material data, and obtains the user attribute characteristics corresponding to the target user who provides business consulting information;

[0023] The second device determines the recommendation evaluation values ​​corresponding to the M target material data respectively according to the material attribute characteristics, the user attribute characteristics and the information feature vector corresponding to the business consultation information;

[0024] The second device sorts the M recommended evaluation values, and displays the M target material data according to the sorted M recommended evaluation values.

[0025] Among them, the material attribute features include material semantic features and material type features;

[0026] The second device determines the recommendation evaluation values ​​corresponding to the M target material data respectively according to the material attribute characteristics, the user attribute characteristics, and the information feature vector corresponding to the business consultation information, including:

[0027] The second device inputs the information feature vector corresponding to the material semantic features, user attribute features, and business consultation information into the recommendation model, and determines the predicted evaluation values ​​corresponding to the M target material data respectively according to the recommendation model;

[0028] The second device obtains weight coefficients corresponding to the M target material data according to the material type characteristics;

[0029] The second device determines the recommended evaluation values ​​corresponding to the M target material data respectively according to the weight coefficient and the predicted evaluation value.

[0030] An embodiment of the present application provides a method for recommending service data, including:

[0031] The first device responds to an information input operation for a target business and sends the business consultation information determined by the information input operation to the second device, so that the second device determines consultation feedback information corresponding to the business consultation information from M target material data included in the target business; M is a positive integer;

[0032] The first device receives the consultation feedback information sent by the second device and displays the consultation feedback information.

[0033] The method further includes:

[0034] The first device obtains user behavior data for consultation feedback information and sends the user behavior data to the second device so that the second device updates the recommendation model based on the user behavior data; the recommendation model is used to determine the predicted evaluation values ​​corresponding to M target material data, and the predicted evaluation values ​​provide a basis for the second device to determine the consultation feedback information.

[0035] An embodiment of the present application provides a service data recommendation device, including:

[0036] A first sending module is configured to respond to an information input operation for a target business and send business consulting information determined by the information input operation to a second device;

[0037] The display module is used to obtain M target material data matching the business consulting information in the target business and display the M target material data; M is a positive integer;

[0038] A first determining module is configured to respond to a selection operation on the M target material data and determine, among the M target material data, the target material data determined by the selection operation as the consultation feedback information;

[0039] The second sending module is used to send the consultation feedback information to the first device.

[0040] The first determining module includes:

[0041] A first acquiring unit is configured to receive the business consulting information sent by the first device and acquire an information feature vector corresponding to the business consulting information;

[0042] The second acquisition unit is used to obtain the intention probabilities between the information feature vector and N candidate intentions respectively, where N is a positive integer;

[0043] A first determining unit is configured to determine a target intent corresponding to the business consulting information from N candidate intents based on the intent probability;

[0044] The third acquisition unit is configured to acquire M target material data matching the target intent in the target service.

[0045] The first acquisition unit is specifically configured to:

[0046] Obtain consulting keywords from business consulting information, perform vector conversion on the consulting keywords, and obtain word vectors corresponding to the consulting keywords;

[0047] According to the semantic information corresponding to the word vector, the information feature vector corresponding to the consulting keyword is obtained.

[0048] The second acquisition unit is specifically configured to:

[0049] Input the information feature vector into the neural network layer in the intent recognition model, and obtain the prediction vector corresponding to the information feature vector based on the weight matrix corresponding to the neural network layer;

[0050] In the classifier of the intent recognition model, the intent probability between the prediction vector and N candidate intents is determined.

[0051] The display module includes:

[0052] A fourth acquisition unit is configured to acquire material attribute features corresponding to the M target material data, and acquire user attribute features corresponding to the target user providing the business consulting information;

[0053] The second determining unit is configured to determine the recommendation evaluation values ​​corresponding to the M target material data respectively according to the material attribute characteristics, the user attribute characteristics and the information feature vector corresponding to the business consultation information;

[0054] The display unit is configured to sort the M recommended evaluation values ​​and display the M target material data according to the sorted M recommended evaluation values.

[0055] Among them, the material attribute features include material semantic features and material type features;

[0056] The second determining unit is specifically configured to:

[0057] Input the information feature vectors corresponding to the material semantic features, user attribute features, and business consultation information into the recommendation model, and determine the prediction evaluation values ​​corresponding to the M target material data respectively according to the recommendation model;

[0058] Obtain weight coefficients corresponding to M target material data respectively according to material type characteristics;

[0059] According to the weight coefficient and the predicted evaluation value, the recommended evaluation values ​​corresponding to the M target material data are determined.

[0060] An embodiment of the present application provides a service data recommendation device, including:

[0061] a second determining module, configured to respond to an information input operation for a target business and send the business consulting information determined by the information input operation to a second device, so that the second device determines consulting feedback information corresponding to the business consulting information from M target material data included in the target business; M is a positive integer;

[0062] The receiving module is used to receive the sent consultation feedback information and display the consultation feedback information.

[0063] The device further comprises:

[0064] The third sending module is used to obtain user behavior data for consultation feedback information and send the user behavior data to the second device so that the second device can update the recommendation model based on the user behavior data; the recommendation model is used to determine the predicted evaluation values ​​corresponding to M target material data, and the predicted evaluation values ​​provide a basis for the second device to determine the consultation feedback information.

[0065] In one aspect, the present application provides a computer device, comprising: a processor and a memory;

[0066] The memory is used to store a computer program, and the processor is used to call the computer program to perform the following steps:

[0067] In response to an information input operation for a target business, sending business consulting information determined by the information input operation to a second device;

[0068] Obtain M target material data matching the business consultation information in the target business, and display the M target material data; M is a positive integer;

[0069] In response to a selection operation on the M target material data, the target material data determined by the selection operation is determined as the consultation feedback information among the M target material data;

[0070] The consultation feedback information is sent to the first device.

[0071] In one aspect, an embodiment of the present application provides a computer-readable storage medium storing a computer program. When the computer program is executed by a processor, the computer program performs the following steps:

[0072] In response to an information input operation for a target business, sending business consulting information determined by the information input operation to a second device;

[0073] Obtain M target material data matching the business consultation information in the target business, and display the M target material data; M is a positive integer;

[0074] In response to a selection operation on the M target material data, the target material data determined by the selection operation is determined as the consultation feedback information among the M target material data;

[0075] The consultation feedback information is sent to the first device.

[0076] In one aspect, the present application provides a computer program product or computer program, which 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 described in the above aspect.

[0077] In an embodiment of the present application, the first device responds to an information input operation for a target business and sends the business consultation information determined by the information input operation to the second device. The second device obtains M target material data matching the business consultation information in the target business and displays the M target material data, where M is a positive integer. The second device responds to a selection operation for the M target material data and determines the target material data determined by the selection operation as consultation feedback information among the M target material data. The second device sends the consultation feedback information to the first device. It can be seen that for the business consultation information sent by the first device, the second device can preliminarily screen out the M target material data matching the business consultation information in the target business and display the M target material data. From the displayed M target material data, the consultation feedback information corresponding to the business consultation information can be determined, and the consultation feedback information is returned to the first device as the reply information of the business consultation information. This can reduce tedious manual operations and thus improve the efficiency of business data recommendation. The consultation feedback information determined from the preliminarily screened M target material data can improve the accuracy of business data recommendation. BRIEF DESCRIPTION OF THE DRAWINGS

[0078] 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.

[0079] Figure 1 This is a schematic diagram of the architecture of a business data recommendation system provided by an embodiment of the present application;

[0080] Figure 2 This is a flowchart of a business data recommendation method provided by an embodiment of the present application;

[0081] Figure 3 is a schematic diagram of an intent recognition model provided in an embodiment of the present application;

[0082] Figure 4 This is a structural diagram of a perception layer in a neural network layer provided in an embodiment of the present application;

[0083] Figure 5 This is a schematic diagram of determining consultation feedback information corresponding to business consultation information provided by an embodiment of the present application;

[0084] Figure 6 This is a schematic diagram of an embodiment of the present application providing a method for recommending consultation feedback information on insurance services to a target user;

[0085] Figure 7 This is a schematic diagram of a method for determining consultation feedback information in a related technology provided in an embodiment of the present application;

[0086] Figure 8a This is a schematic diagram of determining target material data for display provided by an embodiment of the present application;

[0087] Figure 8b This is a schematic diagram of determining consultation feedback information provided by an embodiment of the present application;

[0088] Figure 8c This is a schematic diagram of determining consultation feedback information provided by an embodiment of the present application;

[0089] Figure 8d is a schematic diagram of target material data provided by an embodiment of the present application;

[0090] Figure 9 This is a schematic diagram of a business data recommendation method provided by an embodiment of the present application;

[0091] Figure 10 This is a schematic diagram of displaying consultation feedback information provided by an embodiment of the present application;

[0092] Figure 11 This is a structural diagram of a business data recommendation device provided in an embodiment of the present application.

[0093] Figure 12 This is a structural diagram of a business data recommendation device provided in an embodiment of the present application;

[0094] Figure 13 It is a structural diagram of a computer device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0095] 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.

[0096] See Figure 1 , Figure 1 This is a schematic diagram of the structure of a business data recommendation system provided by an embodiment of the present application. Figure 1 As shown, the business data recommendation system may include a server 10 and a user terminal cluster. The user terminal cluster may include one or more user terminals, and the number of user terminals is not limited here. Figure 1 As shown, it may specifically include user terminal 100a, user terminal 100b, user terminal 100c, ..., user terminal 100n. Figure 1 As shown, the user terminal 100a, the user terminal 100b, the user terminal 100c, ..., the user terminal 100n can respectively establish a network connection with the server 10, so that each user terminal can exchange data with the server 10 through the network connection.

[0097] Each user terminal in the user terminal cluster may include: smart phones, tablet computers, laptop computers, desktop computers, wearable devices, smart homes, head-mounted devices and other smart terminals with business data recommendations. Figure 1 Each user terminal in the user terminal cluster shown in FIG. 1 may be installed with a target application (ie, an application client). When the application client runs in each user terminal, it may be respectively connected to the above-mentioned Figure 1 The servers 10 shown here interact with each other.

[0098] Among them, such as Figure 1As shown, the server 10 can transmit business consultation information and consultation feedback information, and obtain M target material data matching the business consultation information in the target business. The server 10 can be an independent physical server, or a server cluster or distributed system composed of multiple physical servers. It can also be 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.

[0099] Among them, the first device and the second device in the embodiment of the present application may refer to any two user terminals among the user terminal 100a, the user terminal 100b, the user terminal 100c, ..., and the user terminal 100n. For ease of understanding, the first device in the embodiment of the present application may refer to Figure 1 The user terminal 100a shown, the second device may refer to Figure 1 The user terminal 100b shown. Both the user terminal 100a and the user terminal 100b can be integrated with a target application (i.e., an application client) having a business data recommendation function. In this case, the server 10 can be understood as the background server of the target application, and the user terminal 100a and the user terminal 100b can exchange data through the background server corresponding to the target application. For example, user A can enter business consulting information in the target application installed by the user terminal 100a. In this case, user A can be called the target user. The user terminal 100a can obtain the business consulting information entered by user A and send the business consulting information to the server 10. After the server 10 receives the business consulting information sent by the user terminal 100a, it can identify M target material data that matches the business consulting information in the target business corresponding to the business consulting information, and send the M target material data to the user terminal 100b, where M is a positive integer, such as M can take the value of 1, 2, 3, ... After the user terminal 100b receives the M target material data sent by the server 10, it can display the M target materials in the target application. Of course, after receiving the business advisory information, the server 10 can send the business advisory information to the user terminal 100b, so that the business advisory information can be displayed in the target application of the user terminal 100b. It is understandable that after receiving the business advisory information and sending it to the user terminal 100b, the server 10 can identify the M target material data that match the business advisory information. Therefore, for the user terminal 100b, the time of receiving the business advisory information may be earlier than the time of receiving the M target material data.

[0100] Furthermore, for the M target material data displayed in the target application of the user terminal 100b, the user holding the user terminal 100b (also known as the business personnel) can select the data that best matches the business consultation information from the M target material data as the consultation feedback information. The user terminal 100b can respond to the selection operation of the business personnel, obtain the consultation feedback information corresponding to the business consultation information, and transmit the consultation feedback information to the user terminal 100a through the server 10. The user terminal 100a displays the consultation feedback information to provide consultation and answer services for user A.

[0101] Optionally, user A and business personnel can communicate through different applications, that is, the target application installed by the user terminal 100a can be the first application, and the target application installed by the user terminal 100b can be the second application. At this time, the server 10 may include a background server corresponding to the first application and a background server corresponding to the second application (the user terminal 100a that installs the first application and the background server of the first application can be called the first device, and the user terminal 100b that installs the second application and the background server of the second application can be called the second device). For example, user A can enter business consultation information in the first application installed on user terminal 100a. User terminal 100a can obtain the business consultation information entered by user A and send the business consultation information to the background server of the first application. The background server of the first application can then transmit the business consultation information to the background server of the second application. After receiving the business consultation information, the background server of the second application can transmit the business consultation information to user terminal 100b and display the business consultation information in the second application of user terminal 100b. At the same time, the background server of the second application can identify M target material data that match the business consultation information in the target business corresponding to the consultation business information. The background server of the second application can send the M target material data to user terminal 100b. User terminal 100b can display the M target material data in the second application. The M target material data can be displayed in different areas of the same interface as the business consultation information. Business personnel can select the data that best matches the business consultation information from M target material data as consultation feedback information. The user terminal 100b can respond to the business personnel's selection operation, obtain the consultation feedback information corresponding to the business consultation information, and transmit the consultation feedback information to the background server of the second application. Then, through data transmission between the background server of the second application and the background server of the first application, the background server of the first application transmits the consultation feedback information to the user terminal 100a. The user terminal 100a displays the consultation feedback information to provide consultation and answer services for user A.

[0102] See Figure 2 , Figure 2This is a flow chart of a business data recommendation method provided by an embodiment of the present application. The business data recommendation method can be interactively executed by a first device and a second device, and the first device and the second device can be servers (such as the above Figure 1 Server 10 in the above), or user terminal (as mentioned above Figure 1 Any user terminal in a user terminal cluster), or a system consisting of a server and a user terminal, which is not limited in this application. Figure 2 As shown, the business data recommendation method may include steps S101-S104.

[0103] S101: A first device responds to an information input operation for a target business and sends business consultation information determined by the information input operation to a second device.

[0104] Specifically, in a business consultation scenario, when a user has doubts about a business in a target application, the user can input business consultation information for the above business on the interactive interface corresponding to the first device used. In the embodiment of the present application, for the convenience of description, the user who inputs the business consultation information can be referred to as the target user, and the business that the target user has doubts about can be referred to as the target business. The first device can respond to the target user's information input operation for the target business, obtain the business consultation information for the target business input by the target user, and send the business consultation information determined by the information input operation to the second device. Among them, the target application can refer to an application client for providing business data recommendation services, such as an e-commerce mall application, a communication application, an insurance application, and other applications that provide business consultation services; the target business can refer to a shopping business in an e-commerce mall, a package business in a communication application, an insurance business in an insurance application, etc. The embodiment of the present application does not limit the types of target applications and target businesses.

[0105] S102: The second device obtains M target material data matching the business consultation information in the target business, and displays the M target material data; M is a positive integer.

[0106] Specifically, after receiving the business consulting information for the target business sent by the first device, the second device can obtain M target material data matching the business consulting information in the target business and display the M target material data on the corresponding interactive interface of the second device.

[0107] Optionally, a specific method for the second device to obtain M target material data matching the business consultation information in the target business may include: the second device receives the business consultation information sent by the first device, and obtains an information feature vector corresponding to the business consultation information. The second device obtains the intent probability between the information feature vector corresponding to the business consultation information and N candidate intents, where N is a positive integer. The second device determines the target intent corresponding to the business consultation information from the N candidate intents based on the intent probability, and the second device obtains M target material data matching the target intent in the target business.

[0108] Specifically, after receiving the business consultation information sent by the first device, the second device can obtain an information feature vector corresponding to the business consultation information. For example, the information feature vector can be converted to a feature vector using a feature vector conversion model to obtain an information feature vector corresponding to the business consultation information, which is composed of floating-point numbers. After obtaining the information feature vector corresponding to the business consultation information, the second device can obtain the intent probabilities between the information feature vectors and each of the N candidate intents, thereby determining the intent information corresponding to the business consultation information and determining the target intent of the target user for the target business. After obtaining the intent probabilities between the business consultation information and each of the N candidate intents, the second device can determine the target intent corresponding to the business consultation information from the N candidate intents based on the intent probabilities between the business consultation information and each of the candidate intents. Because each person's speaking style and word choice have multiple meanings, the business consultation information input by the target user for the target business may belong to intent A or intent B. Therefore, the intent probabilities between the business consultation information and each of the N candidate intents can be obtained, and based on the intent probabilities, the target intent corresponding to the business consultation information can be determined from the N candidate intents. For example, the intent probabilities corresponding to each candidate intent are ranked, and the candidate intents with the top three intent probabilities are determined as the target intent corresponding to the business consultation information. After obtaining the target intent, the second device may obtain M target material data matching the target intent in the target business. The M material data are used to answer the business consultation information of the target user for the target business.

[0109] Optionally, the specific method for the second device to obtain the information feature vector corresponding to the business consulting information may include: the second device obtains the consulting keywords in the business consulting information, performs vector conversion on the consulting keywords, obtains the word vector corresponding to the consulting keywords, and the second device obtains the information feature vector corresponding to the consulting keywords based on the semantic information corresponding to the word vector.

[0110] Specifically, the second device can obtain consulting keywords from the business consulting information. These consulting keywords can refer to keywords, key phrases, or key sentences in the business consulting information. Specifically, the second device can extract keywords from the target user's business consulting information regarding the target business and parse the target user's business consulting information regarding the target business. For example, in the insurance business, the target user's business consulting information regarding the target business can include information such as "What does accident insurance cover?" or "What one-year critical illness insurance policies can I purchase?" When obtaining keywords from the consulting information, candidate keywords can be extracted from the business consulting information and then replaced with standard keywords pre-stored in the second device to obtain consulting keywords corresponding to the business consulting information. For example, if the target user inputs business consulting information regarding insurance, such as "How do I purchase critical illness insurance?", the second device, upon receiving the business consulting information, can extract candidate keywords from the business consulting information, such as "purchase," "critical illness insurance," "operation," and so on. Based on the candidate keywords corresponding to the business consulting information, standard keywords matching the candidate keywords are retrieved from a keyword database as consulting keywords. For example, "What do I need to do to purchase critical illness insurance?" The corresponding consultation keyword is "How to purchase critical illness insurance?"

[0111] After obtaining the consulting keywords corresponding to the business consulting information, the second device can perform vector conversion on the consulting keywords to obtain word vectors corresponding to the consulting keywords, and obtain the information feature vectors corresponding to the consulting keywords based on the semantic information corresponding to the word vectors. The consulting keywords can be vector converted using a vector conversion model to obtain word vectors corresponding to the consulting keywords. Specifically, the vector conversion model can refer to a BERT model (Bidirectional Encoder Representations from Transformer, language representation model). The purpose of the BERT model is to obtain vectors of each word / character in the input text after integrating the semantic information of the entire text. After obtaining the consulting keywords in the business consulting information, the consulting keywords are input into the BERT model. The BERT model can convert the consulting keywords into one-dimensional word vectors by querying the word vector table. The word vector table is the word vector corresponding to the feature information of each word collected in advance. At the same time, when there are multiple consulting keywords, the BERT model will also obtain the global semantic vectors corresponding to each consulting keyword and merge them with the word vector corresponding to each consulting keyword. In addition, since the semantic information carried by words / characters appearing in different positions is different, the BERT model will also obtain the position vectors corresponding to the consulting keywords, that is, the BERT model can obtain the position vectors corresponding to the consulting keywords in different positions. For example, "I love you" and "You love me" are both composed of the three characters "I," "love," and "you," but the different positions of the characters represent different semantic information. The BERT model uses the word vectors, global semantic vectors, and position vectors corresponding to the consultation keywords to obtain the sentence feature vectors corresponding to the consultation keywords.

[0112] Among them, the BERT model in the embodiment of the present application is a vector conversion model obtained by fine-tuning the basic BERT model, that is, the current corpus corresponding to the target business (that is, the business corpus corresponding to the target business) is added to the basic model for fine-tuning. Specifically, the specific method of obtaining the BERT model in the embodiment of the present application may include: obtaining an initial vector conversion model (that is, a basic BERT model), training sample data, and a label vector corresponding to the training sample data. The label information of the training sample refers to the label vector for the target business. For example, in the insurance business, the corresponding label information may refer to a label vector with corpus information corresponding to insurance businesses such as purchase of critical illness insurance and surrender of medical insurance added. The training sample data is vector-converted using the initial vector conversion model to obtain the predicted feature vector corresponding to the training sample data. The loss value corresponding to the initial vector conversion model is determined based on the predicted feature vector and the label vector. The initial vector conversion model is adjusted based on the loss value to obtain the vector conversion model in the embodiment of the present application.

[0113] Among them, it is possible to verify whether the loss value corresponding to the initial vector conversion model meets the convergence condition based on the loss value corresponding to the initial vector conversion model. If the loss value does not meet the convergence condition, the loss degree to which the loss value of the initial vector conversion model belongs is determined, and the parameters in the initial vector conversion model are adjusted according to the loss degree to which the loss value belongs. Among them, the convergence condition may refer to whether the loss value of the initial vector conversion model is less than a preset loss threshold, or whether the number of iterations of the initial vector conversion model reaches a preset number of iterations. If the loss value of the initial vector conversion model is less than the preset loss threshold, it is determined that the initial vector conversion model meets the convergence condition, and the initial vector conversion model with a loss value less than the preset loss threshold is determined as the vector conversion model in the embodiment of the present application. If the loss value of the initial vector conversion model is greater than or equal to the preset loss threshold, it is determined that the initial vector conversion model does not meet the convergence condition, and the initial vector conversion model continues to be trained; or, if the number of iterations of the initial vector conversion model is greater than the preset number of iterations, it is determined that the initial vector conversion model meets the convergence condition, and the initial vector conversion model with the number of iterations greater than the preset number of iterations is determined to be the vector conversion model in the embodiment of the present application; if the number of iterations of the initial vector conversion model is less than or equal to the preset number of iterations, it is determined that the initial vector conversion model does not meet the convergence condition, and the initial vector conversion model continues to be trained.

[0114] Among them, the loss function corresponding to the initial vector conversion model can be expressed by the following formula (1).

[0115] loss total =λloss base +(1-λ)loss insure (1)

[0116] Among them, loss base Represents the loss of the initial vector conversion model (i.e., the basic BERT model), loss insure represents the loss of the current corpus environment, and λ represents a hyperparameter.

[0117] Optionally, the second device may obtain the intent probabilities between the information feature vector and each of the N candidate intents by inputting the information feature vector into a neural network layer in the intent recognition model and obtaining a prediction vector corresponding to the information feature vector based on a weight matrix corresponding to the neural network layer. The second device may determine the intent probabilities between the information feature vector and each of the N candidate intents in a classifier of the intent recognition model.

[0118] Among them, after the second device obtains the information feature vector corresponding to the business consulting information, it can input the information feature vector into the neural network layer in the intent recognition model, and obtain the prediction vector corresponding to the information feature vector according to the corresponding weight matrix in the neural network layer. The neural network layer in the intent recognition model can simulate the human brain, perform intent prediction on the information feature vector corresponding to the business consulting information, and obtain a prediction vector corresponding to the information feature vector. After the second device obtains the prediction vector corresponding to the information feature vector, it can determine the intent probability between the prediction vector and N candidate intentions in the classifier of the intent recognition model.

[0119] Among them, N candidate intentions are associated with the target business. For example, in the insurance business, the N candidate intentions may refer to candidate intentions associated with the insurance business, such as "critical illness insurance purchase", "auto insurance cancellation", "accident insurance claim", "medical insurance consultation", etc. After the intention recognition model is trained, the initial intention recognition model, intent training sample data, and labeled intent information corresponding to the intent training sample data can be obtained. The initial intention recognition model is used to perform intent recognition on the intent training sample data to obtain the predicted intent probability corresponding to the intent training sample data. The loss value of the initial intention recognition model is determined based on the predicted intent probability and the labeled intent information, and the network parameters in the initial intention recognition model are adjusted based on the loss value corresponding to the initial intention recognition model. When the initial intention recognition model meets the convergence condition, the initial intention recognition model that meets the convergence condition is used to determine the intent recognition model.

[0120] like Figure 3 As shown, Figure 3 is a schematic diagram of an intent recognition model provided in an embodiment of the present application, such as Figure 3 As shown, after obtaining the information feature vector corresponding to the business consultation information, the information feature vector can be input into the neural network layer in the intent recognition model. The number of neural network layers in the intent recognition model can be multiple layers. Each neural network layer includes a perception layer. The neural network layer in the intent recognition model refers to an artificial neural network (ANN), which abstracts the human brain neural network from the perspective of information processing to establish a simple model. Different networks are formed according to different connection methods. The neural network layer is composed of a large number of nodes (or neurons) connected to each other. The information feature vector corresponding to the business consultation information is subjected to multiple nonlinear combinations through the perception layer in the neural network layer to obtain a prediction vector corresponding to the information feature vector. After obtaining the prediction vector corresponding to the information feature vector, the prediction vector corresponding to the information feature vector is classified by the classifier in the intent recognition model to obtain the intent probability between the information feature vector and N candidate intents.

[0121] like Figure 4 As shown, Figure 4 This is a structural diagram of the perception layer in a neural network layer provided by an embodiment of the present application, such as Figure 4 As shown, Figure 4 Shows a perceptron neuron in a neural network layer, which is the basic unit of the perceptron in a neural network. Figure 4 As shown in Figure 1, the perceptual neuron can receive the input information feature vector x, and perform nonlinear combination through the perceptual layer in multiple neural network layers to obtain the prediction vector corresponding to the information feature vector. The perceptual neuron combines the information feature vector q with the bias coefficient (i.e., external deviation) b and the weight coefficient ω. j Perform nonlinear combination and then pass the activation function to get the output result y. Among them, the information feature vector q and the bias coefficient b, as well as the weight coefficient ω j The nonlinear combination can be expressed by formula (2).

[0122] C=qω j +b (2)

[0123] Among them, C in formula (2) refers to the information feature vector x, the bias coefficient B, and the weight coefficient ω j The value obtained by nonlinear combination, q refers to the information eigenvector, ω j refers to the weight coefficient of the jth perception layer, b refers to the bias coefficient, and j refers to the perception layer.

[0124] Get the information feature vector q, bias coefficient b, and weight coefficient ω j After the value C is obtained by nonlinear combination, the output value of the j-th perception layer neural unit can be obtained through the activation function. The function expression corresponding to the activation function can be expressed by formula (3).

[0125] y=a(C) (3)

[0126] Where y in formula (3) refers to the output value of the j-th perception layer neural unit, and a refers to the activation function.

[0127] In this way, the output value y corresponding to the perception layer in each neural network is obtained through the method described above, and finally the output value x of the multi-layer neural network (that is, the prediction vector corresponding to the information feature vector) is obtained based on the output value y corresponding to the perception layer in each neural network.

[0128] Among them, in the classifier of the intent recognition model, the following formula (4) can be used to determine the intent probability between the prediction vector and N candidate intents.

[0129]

[0130] Where i in formula (4) refers to the intention, x refers to the prediction vector corresponding to the information feature vector, and W i refers to the classification weight coefficient of intention i, c refers to the total intention, p(y i |x) represents the probability that the information feature vector belongs to intention i.

[0131] Among them, after the second device obtains the intention probability between the information feature vector and N candidate intentions, it can determine the target intention from the N candidate intentions based on the intention probability. The number of the target intentions can be 1 or more, and the embodiment of the present application does not limit this. For example, the candidate intention corresponding to the largest intention probability can be determined as the target intention, or the intention probabilities can be sorted in descending order, and the candidate intentions corresponding to the top three intention probabilities can be determined as the target intention.

[0132] Among them, after the second device obtains the target intent corresponding to the business consulting information, it can obtain M target material data corresponding to the target intent from the Q candidate material data in the material index library associated with the target business according to the target intent, where Q is a positive integer. Specifically, the second device can pre-establish a material index library, add an intent label to each of the Q candidate material data, and establish a corresponding relationship between the candidate material data and the candidate intent. Among them, the candidate material data can be in the form of text material data, video material data, and image material data, etc. For example, in the insurance business, the material data corresponding to the detailed process of purchasing critical illness insurance can be added with the "critical illness insurance purchase" intent label. After establishing the material index library corresponding to the candidate intent and the candidate material data, the M target material data corresponding to the target intent can be determined from the material index library according to the target intent. Among them, considering the actual performance impact, the number of M target material data recalled from the material index library can be controlled at around 50-100. Among them, the candidate material data in the material index library can refer to pre-collected standardized material data with normativeness and professionalism. In this way, the normativeness and professionalism of business data recommendations can be improved, the service image and service level of the target business can be improved, and the service quality can be improved.

[0133] Optionally, the second device may display the M target material data in a specific manner including: obtaining material attribute characteristics corresponding to each of the M target material data, and obtaining target user attribute characteristics corresponding to the target user providing the business consulting information. The second device determines, based on the material attribute characteristics, the target user attribute characteristics, and the information feature vector corresponding to the business consulting information, a recommended evaluation value corresponding to each of the M target material data; the second device sorts the M recommended evaluation values; and displays the M target material data based on the sorted M recommended evaluation values.

[0134] Specifically, after the second device obtains the M target material data corresponding to the target intent from the material index library, it can obtain the material attribute features corresponding to the M target material data, as well as the user attribute features corresponding to the target user providing business consulting information. The second device can determine the recommended evaluation values ​​corresponding to the M target material data based on the material attribute features, target user attribute features, and information feature vectors corresponding to the business consulting information corresponding to each target material data in the M target material data, that is, combine the material attribute features, user attribute features, and information feature vectors to score each target material data in the M target material data. After obtaining the recommended evaluation values ​​corresponding to the M candidate target material data, the second device can sort the M recommended evaluation values ​​and display the M target material data according to the sorted M recommended evaluation values. For example, after sorting the M recommended evaluation values ​​in descending order, the target material data corresponding to the recommended evaluation values ​​sorted in front can be displayed.

[0135] Optionally, the material attribute features include material semantic features and material type features. A specific method for the second device to determine the recommended evaluation values ​​corresponding to the M target material data based on the material attribute features, user attribute features, and information feature vectors corresponding to the business consultation information may include: the second device inputting the material semantic features, user attribute features, and information feature vectors corresponding to the business consultation information into a recommendation model, and determining the predicted evaluation values ​​corresponding to the M target material data based on the recommendation model. The second device obtains weight coefficients corresponding to the M target material data based on the material type features, and determines the recommended evaluation values ​​corresponding to the M target material data based on the weight coefficients and the predicted evaluation values.

[0136] Specifically, the second device can input the material semantic features, user attribute features and information feature vectors corresponding to the business consulting information into the recommendation model, and determine the predicted evaluation values ​​corresponding to the M target material data according to the recommendation model. The material semantic features corresponding to each target material data refer to the semantic features corresponding to each target material data, such as the keywords in each target material data or the target product corresponding to each target material data, etc. For example, in the insurance business, the material semantic features of the target material data corresponding to how to buy critical illness insurance can refer to the target product "critical illness insurance" or the keywords "purchase, critical illness insurance, process", etc. The user attribute features corresponding to the target user refer to the target user's age, gender, behavioral preferences, etc., and the material semantic features, user attribute features and information feature vectors are combined to perform a recommendation evaluation on each target material data to obtain the recommendation evaluation value corresponding to each target material data, that is, to determine the degree of association between each target material data and the target user, and to determine the target user's interest in each target material data.

[0137] The recommendation model in the embodiment of the present application may refer to an FM-FTRL (Factor Machine-Follow the Regularized Leader, an online machine learning algorithm model) recommendation model. The FM-FTRL recommendation model is a FM model (Factor Machine, a factor decomposition model that can be used to learn the interactive hidden relationships between features). The FM-FTRL recommendation model can be used to learn the cross-hidden relationships between the semantic features of the material, the user attribute features, and the information feature vectors corresponding to the business consulting information to obtain the recommendation evaluation value corresponding to each target material data. The function expression of the FM-FTRL recommendation model can be expressed as follows: Formula (5).

[0138]

[0139] Among them, w in formula (5) is the parameter of input feature, H i It refers to the sum of the input feature vectors (i.e., the total vector obtained by splicing the semantic features of the learning material, the user attribute features, and the information feature vectors corresponding to the business consulting information), V refers to the latent vector, each feature vector corresponds to a latent vector, M refers to the total number of features, and h refers to each feature vector (such as any one of the semantic features of the learning material, the user attribute features, or the information feature vectors corresponding to the business consulting information). <w,H i > refers to linear impact, which can obtain independent feature information corresponding to each eigenvector. Refers to linear regression, which can obtain combined feature information corresponding to multiple feature vectors. Among them, the second device can also collect feedback information from target users on target material data to update and iterate the FM-FTRL recommendation model in this application, so as to improve the real-time response capability of the FM-FTRL recommendation model, and at the same time meet the needs of real-time increase in material / cold start exposure, etc. The predicted evaluation value corresponding to each target material data is obtained through the FM-FTRL recommendation model, that is, each target material data is scored, so as to provide reference information for subsequent recommendation consultation feedback information to the target user.

[0140] The second device may also obtain weight coefficients corresponding to the M target material data based on the material type characteristics corresponding to each target material data. That is, the weight coefficient corresponding to each candidate material data may be determined in advance based on the material type characteristics corresponding to each candidate material data. The material type characteristics corresponding to the material data may refer to the importance or exposure rate of the product corresponding to the material data, etc. For example, the weight coefficient of the material data corresponding to more important products may be set higher, and the weight coefficient of the material data corresponding to products that require exposure may be set higher, etc. If the candidate material data K is the material data corresponding to a new product that is currently being promoted, the weight coefficient of the candidate material data K may be set higher.

[0141] After the second device obtains the weight coefficients and predicted evaluation values ​​corresponding to the M target material data, it can obtain the recommended evaluation value corresponding to each target material data in the M target material data based on the weight coefficient and predicted evaluation value corresponding to each target material data. If the weight coefficient corresponding to the target material data S is e, and the predicted evaluation value corresponding to the target material data S is z, then the recommended evaluation value corresponding to the target material data S can be the product between the weight coefficient and the predicted evaluation value, that is, e*z. After the second device obtains the recommended evaluation value corresponding to each target material data, it can sort the M recommended evaluation values ​​to obtain the sorted M recommended evaluation values, and display the M target material data based on the sorted M recommended evaluation values. For example, if the M target material data are sorted based on the sorted M recommended evaluation values, the M recommended evaluation values ​​can be sorted in descending order, or in ascending order, and so on.

[0142] After obtaining the predicted evaluation value corresponding to each target material data, the second device can sort the predicted evaluation value corresponding to each target material data to obtain M sorted predicted evaluation values. The M target material data can then be sorted based on the sorted M predicted evaluation values ​​to obtain M sorted target material data. Specifically, the target material data for each target intent can be sorted. After obtaining the sorted target material data for each target intent, the sorted target material data can be adjusted, such as by obtaining the material type features of each target material data to adjust the sorted target material data. For example, the material type features can include video material type, text material type, and image material type. A certain amount of target material data is retained for the video material type, a certain amount of target material data is retained for the text material type, and a certain amount of target material data is retained for the image material type. That is, a certain amount of target material data is retained for each material type to meet the potential different needs of the target user while saving storage space. Retaining a certain amount of target material data for each material type can determine the target material data to be retained based on the predicted evaluation value corresponding to each target material data, and the retained target material features are displayed in the interactive interface corresponding to the second device. By using certain configuration rules to optimize the configuration of the M target material data, it can be made to meet the needs of the target business as much as possible. This can make the M target material data more organized, facilitate the subsequent provision of consulting feedback information to the target users, and improve the efficiency of business data recommendation.

[0143] For example, in the insurance business, if the target intention is to purchase critical illness insurance, after recalling 60 target material data from the material index library based on the target intention of purchasing critical illness insurance, the recommendation model can be used to obtain the predicted evaluation values ​​corresponding to the 60 target material data. Then, the material type features corresponding to the 60 target material data are obtained, and the 60 target material data are classified, that is, the 60 target material data are divided into video material data, text material data or picture material data. According to the predicted evaluation values ​​corresponding to the 60 target material data, the target material data under each material type are sorted. The 60 target material data are sorted. For example, if there are 20 target material data under the video material type, the 20 target material data under the video material type are sorted according to the predicted evaluation values ​​corresponding to the 20 target material data, and then the target material data sorted in the front can be retained. Among them, the M target material data can be displayed on the interactive interface corresponding to the second device.

[0144] S103: The second device responds to the selection operation on the M target material data, and determines the target material data determined by the selection operation as the consultation feedback information among the M target material data.

[0145] Specifically, after the M target material data are displayed on the interactive interface corresponding to the second device, the business personnel can select the consulting feedback information to be sent to the target user in the interactive interface displaying the M target material data in the second device. The business personnel refers to the staff corresponding to the target business. The second device can respond to the business personnel's selection operation for the M target material data, and determine the target material data determined by the selection operation as the consulting feedback information among the M target material data. Since the second device directly determines the feedback information for feedback to the target user from the M target material data, the error rate is high and it is relatively mechanical, and it cannot provide the target user with more accurate consulting foul information. Therefore, this solution can display the M target material data on the display interface (i.e., interactive interface) of the second device where the business personnel is located, and the business personnel decides the consulting feedback information that is finally sent to the target user, which can improve the accuracy of business data recommendations and bring a better consulting experience and better solution to user problems to the target user.

[0146] S104: The second device sends the consultation feedback information to the first device.

[0147] Specifically, after obtaining the consultation feedback information, the second device can send the consultation feedback information to the first device where the target user is located. After receiving the consultation feedback information sent by the second device, the first device displays the consultation feedback information on the display interface corresponding to the first device to answer the business consultation information corresponding to the target user.

[0148] like Figure 5 As shown, Figure 5 This is a schematic diagram of determining consultation feedback information corresponding to business consultation information provided by an embodiment of the present application, such as Figure 5As shown, after a target user enters business consultation information for a target business on the display interface corresponding to the first device, the first device transmits the business consultation information for the target business to the second device. After receiving the business consultation information from the first device, the second device performs keyword extraction 50a on the business consultation information to extract the consultation keywords from the business consultation information. The second device performs feature vector conversion 50b on the consultation keywords in the business consultation information to obtain information feature vectors corresponding to the consultation keywords. After obtaining the information feature vectors corresponding to the consultation keywords, the second device performs intent recognition 50c on the business consultation information based on the information feature vectors to obtain the target intent corresponding to the business consultation information. Based on the target intent corresponding to the business consultation information, the second device retrieves materials from a material index library 50d to obtain M target material data corresponding to the target intent. Based on the user attribute characteristics of the target user, the material attribute characteristics corresponding to each target material data, and the information feature vector corresponding to the business consultation information, the second device can obtain predicted evaluation values ​​corresponding to each of the M target material data. Based on the predicted evaluation values ​​corresponding to each of the M target material data, the second device performs material sorting 50e on the M target material data to obtain sorted M target material data. The second device can adjust the M target materials based on a certain configuration strategy (e.g., determining a weight coefficient based on the material type characteristics corresponding to the target material data, and then obtaining a recommended evaluation value based on the weight coefficient and the predicted evaluation value). The adjusted M target material data is displayed on a display interface corresponding to the second device, and the business personnel can determine the consultation feedback information from the displayed M target material data.

[0149] Among them, the embodiment of the present application can also be applied when the target user triggers the confirmation operation for the manual reply, and the first device where the target user is located can connect to the communication channel between the second device, thereby connecting the communication channel between the target user and the business personnel. After the first device receives the business consultation information of the target user for the target business, the business consultation information is sent to the second device. After the second device receives the business consultation information sent by the first device, it determines the consultation feedback information corresponding to the business consultation information, and returns the consultation feedback information to the first device. After the first device receives the consultation feedback information sent by the second device, it displays the consultation feedback information on the display interface corresponding to the first device, giving the target user a better consultation experience and better solving user problems. Of course, the embodiment of the present application can also be applied when the target user inputs the business consultation information for the target business, the first device directly sends the business consultation information to the second device, and receives the consultation feedback information corresponding to the business consultation information returned by the second device, and displays the consultation feedback information.

[0150] The embodiments of the present application can be applied to insurance business, such as Figure 6 As shown, Figure 6 This is a schematic diagram of an embodiment of the present application providing a method for recommending consultation feedback information on insurance business to a target user. Figure 6 As shown, if the target user wants to obtain information about the operational process of paying and renewing critical illness insurance in the insurance business, the target user can enter business consultation information for critical illness insurance in the insurance business on the corresponding interactive page of the first device 60a (such as a mobile phone). After the first device 60a receives the critical illness insurance consultation information of the target user for critical illness insurance, it sends the critical illness insurance consultation information to the second device 60b, and the second device 60b can send the critical illness insurance consultation information to the server 60c corresponding to the second device 60b. After the server 60c corresponding to the second device 60b receives the critical illness insurance consultation information of the target user for critical illness insurance, the server 60c can obtain the information feature vector corresponding to the critical illness insurance consultation information, determine the target intent corresponding to the critical illness insurance consultation information based on the information feature vector, and recall M target material data associated with the target intent from the material index library. After the server 60c sends the M target material data corresponding to the critical illness insurance consultation information to the second device 60b, the second device 60b displays the M target material data on the corresponding interactive interface. The business personnel can determine critical illness insurance consultation feedback information to be fed back to the target user from the M target material data. The second device 60b responds to the business personnel's selection operation on the M target material data and determines the target material data determined by the selection operation as the critical illness insurance consultation feedback information from the M target material data. The second device 60b returns the critical illness insurance consultation feedback information to the first device 60a. After receiving the critical illness insurance consultation information returned by the second device 60b, the first device 60a displays the critical illness insurance consultation information on the corresponding display interface of the first device 60a, and the target user can view it on the corresponding page.

[0151] like Figure 7 As shown, Figure 7 This is a schematic diagram of a method for determining consultation feedback information in a related technology provided in an embodiment of the present application. Figure 7 As shown in the related art, after the business personnel corresponding to the target business (such as customer service personnel or business stewards) receive the business consultation information of the target user for the target business, the business personnel need to query the material data associated with the business consultation information from the terminal device where the business personnel is located according to the business consultation information sent by the target user, or take screenshots of the relevant material data from some web pages or some small programs and send them to the target user. Figure 7As shown, in the related art, after the business personnel receive the business consultation information sent by the target user, there is no reference material data provided to the business personnel on the terminal display interface corresponding to the business personnel. The business personnel need to find the consultation feedback information corresponding to the business consultation information by themselves, which will result in non-standard consultation feedback information. For the same business consultation information, the consultation feedback information found by each business personnel may be different, resulting in problems with the professionalism and uniformity of the business service, and may even affect the service image of the target business. At the same time, the business personnel searching for consultation feedback information by themselves will cause the target user to wait too long, resulting in a poor experience for the target user. The business personnel are also unable to take into account more user consultations, resulting in low manpower efficiency, and are prone to sending errors and misoperation, resulting in low accuracy and efficiency of business data recommendations.

[0152] like Figure 8a As shown, Figure 8a This is a schematic diagram of a target material data to be displayed provided by an embodiment of the present application. After the second device receives the insurance business consultation information "How to renew the payment" sent by the target user for the insurance business, the second device obtains the information feature vector corresponding to the insurance business consultation information "How to renew the payment". The second device determines the target intent corresponding to the insurance business consultation information "How to renew the payment" based on the information feature vector, and recalls M target material data associated with the target intent from the material index library, and displays the M target material data on the corresponding display interface. Figure 8a As shown, the business personnel can see the insurance business consultation information "How to pay the renewal fee" sent by the target user on the display interface corresponding to the second device. Figure 8a As shown, the right end of the display interface corresponding to the second device displays M target material data corresponding to the insurance business consultation information, that is, M answers to the consultation information of "How to pay the renewal fee". The business personnel can determine the insurance consultation feedback information to be sent to the target user from the M target material data displayed on the second device. Figure 8a As shown, when displaying M target material data, they can be displayed according to the material type corresponding to the target material data, such as displaying text type target material data first, then displaying picture type target material data, etc., to facilitate business personnel to select.

[0153] like Figure 8b As shown, Figure 8b This is a schematic diagram of determining consultation feedback information provided by an embodiment of the present application. Figure 8bAs shown, the business personnel can determine the insurance consultation feedback information to be sent to the target user in the target material data of text type. After the business personnel determine the target material data to be sent to the target user, they can directly click the "Send" touch key, and the second device can determine the corresponding target material data as consultation feedback information and send the consultation feedback information to the first device where the target user is located. Figure 8b As shown, the business personnel can also click the "Copy" touch key, paste it in the corresponding sending window and then send it. In this way, the business personnel can adjust the copied target material data, and can also send the copied target material data to other users, which improves the convenience of the business personnel's operation and improves work efficiency.

[0154] like Figure 8c As shown, Figure 8c This is a schematic diagram of determining consultation feedback information provided by an embodiment of the present application. Figure 8c As shown, the business personnel can determine the target material data of the picture type to be sent to the target user in the corresponding picture template (ie, the target material data in the picture type). Similarly, the business personnel can click "Send" to send it directly, or copy it first and then send it.

[0155] like Figure 8d As shown, Figure 8d is a schematic diagram of target material data provided in an embodiment of the present application, such as Figure 8d For example, in the insurance business, for the insurance consultation information "How to renew the premium" sent to the target user, the corresponding target material data can be the specific operation process of renewal payment. For example, the first step is to select the insurance type; the second step is to select the specific insurance; the third step is to select the coverage amount; the fourth step is to select the payment period, etc.

[0156] Among them, the embodiments of the present application can also be applied to medical consultation services. For example, when the target user wants to consult medical insurance information or consultation procedures for the dental department and other medical consultation information, the corresponding medical consultation information can be entered in the corresponding page through a mobile phone. After the terminal device corresponding to the business personnel receives the medical consultation information, it can determine the M target material data corresponding to the medical consultation information, and the corresponding business personnel can select the target material data to be sent to the target user from the M target material data. The terminal device corresponding to the business personnel responds to the business personnel's determination operation for the M target material data, and after determining the medical feedback information, the medical feedback information is sent to the terminal device where the target user is located. The terminal device where the target user is located displays the medical feedback information and answers the medical feedback information corresponding to the target user.

[0157] In an embodiment of the present application, the first device responds to an information input operation for a target business and sends the business consultation information determined by the information input operation to the second device. The second device obtains M target material data matching the business consultation information in the target business and displays the M target material data, where M is a positive integer. The second device responds to a selection operation for the M target material data and determines the target material data determined by the selection operation as consultation feedback information among the M target material data. The second device sends the consultation feedback information to the first device. It can be seen that for the business consultation information sent by the first device, the first and second devices can preliminarily screen out M target material data matching the business consultation information in the target business and display the M target material data. From the displayed M target material data, the consultation feedback information corresponding to the business consultation information can be determined, and the consultation feedback information is returned to the first device as the reply information of the business consultation information. This can reduce tedious manual operations and thus improve the efficiency of business data recommendation. The consultation feedback information determined from the preliminarily screened M target material data can improve the accuracy of business data recommendation. At the same time, this application can also improve the convenience and per capita productivity of business personnel, as well as improve the standardization and professionalism of business data recommendations, and bring users a better consulting experience and better solutions to user problems, thereby improving the service experience corresponding to the target business.

[0158] like Figure 9 As shown, Figure 9 Schematic diagram of a business data recommendation method provided by an embodiment of the present application, which can be executed by a computer device, which can be a server (such as the above Figure 1 Server 10 in the above), or target user terminal (as mentioned above Figure 1 Any target user terminal in the target user terminal cluster), or a system consisting of a server and a target user terminal, which is not limited in this application. Figure 9 The steps of the business data recommendation method include S201-202.

[0159] S201, the first device responds to the information input operation for the target business and sends the business consultation information determined by the information input operation to the second device, so that the second device determines the consultation feedback information corresponding to the business consultation information in the M target material data included in the target business; M is a positive integer.

[0160] Specifically, if the target user wants to consult about the target business, the target user can input business consultation information for the target business in the interactive page corresponding to the first device. The first device can respond to the target user's information input operation for the target business, and send the business consultation information determined by the information input operation and the business consultation letter for the target business to the second device. After the second device receives the business consultation information sent by the first device, it can obtain the information feature vector corresponding to the business consultation information, determine the target intent corresponding to the business consultation information based on the information feature vector, and recall M target material data associated with the target intent from the material index library corresponding to the target business, and determine the consultation feedback information corresponding to the business consultation information in the M target material data. The specific content of step S201 of the embodiment of the present application can be found in the above. Figure 2 The contents described in the embodiments of this application will not be repeated here.

[0161] S202: The first device receives the consultation feedback information sent by the second device and displays the consultation feedback information.

[0162] Specifically, after the second device determines the consultation feedback information corresponding to the business consultation information, it returns the consultation feedback information to the first device. The first device can receive the consultation feedback information sent by the second device and display the consultation feedback information on the corresponding interactive interface of the first device so that the target user can view it and provide answers to the target user.

[0163] like Figure 10 As shown, Figure 10 This is a schematic diagram of displaying consultation feedback information provided by an embodiment of the present application. Figure 10 As shown, Figure 10 The interactive interface shown is the interactive interface of the target user end. The target user can enter business consultation information for the target business in the interactive interface corresponding to the first device. For example, in the insurance business, when the target user wants to inquire about the specific operation process of insurance renewal payment, he can find the customer service interactive interface corresponding to the insurance business in the first device corresponding to the target user (such as a mobile phone) and enter the business consultation information for the insurance business "How to renew the payment". After the first device corresponding to the target user obtains the "How to renew the payment" input by the target user, it can send the "How to renew the payment" to the second device corresponding to the insurance business. After receiving the "How to renew the payment", the second device determines the consultation feedback information corresponding to the "How to renew the payment", such as the renewal payment operation process in text type, or the renewal payment process in picture type, etc. The consultation feedback information corresponding to "How to renew the payment" is returned to the first device, and the first device displays the consultation feedback information on the corresponding interactive interface.

[0164] Optionally, the first device may also obtain user behavior data regarding the consultation feedback information and send the user behavior data to the second device, so that the second device can update the recommendation model based on the user behavior data. The recommendation model is used to determine the predicted evaluation values ​​corresponding to the M target material data, and the predicted evaluation values ​​provide a basis for the second device to determine the consultation feedback information.

[0165] Specifically, after the first device displays the consultation feedback information on the corresponding interactive interface, it can also obtain the user behavior data of the target user for the consultation feedback information. The user behavior data can be: within a preset time period, the target user viewed the consultation feedback information, or the target user did not view the consultation feedback information. After the first device obtains the user behavior data corresponding to the target user, it can send the user behavior data to the second device. After the second device receives the user behavior data for the consultation feedback information sent by the first device, it can set feedback label information for the target material data corresponding to the consultation feedback information according to the user behavior data. If the user behavior data shows that the user has viewed the consultation feedback information, the feedback label information of the target material data corresponding to the consultation feedback information can be set to "1". If the user behavior data shows that the user has not viewed the consultation feedback information, the feedback label information of the target material data corresponding to the consultation feedback information can be set to "0". Updating the recommendation model based on the target material data corresponding to the consultation feedback information and the corresponding feedback label information can provide the material recommendation capability of the recommendation model, as well as improve the real-time response capability of the recommendation model and accelerate the iteration efficiency of the recommendation model.

[0166] In an embodiment of the present application, the consultation feedback information sent by the second device is received by the first device, and the consultation feedback information is displayed, so that the second device can determine the consultation feedback information corresponding to the business consultation information among the M target material data contained in the target business. The first device can receive the consultation feedback information sent by the second device and display the consultation feedback information. The accuracy and efficiency of business data recommendations can be improved by sending the consultation feedback information to the second device through the first device so that the second device can determine the consultation feedback information corresponding to the business consultation information among the M target material data contained in the target business. At the same time, the embodiment of the present application can send the user behavior data for the consultation feedback information to the second device through the first device, so that the second device updates the recommendation model based on the user behavior data, thereby improving the accuracy of the recommendation model in the second device, and then improving the accuracy of the business data recommendation.

[0167] See Figure 11 , Figure 11This is a schematic diagram of the structure of a business data recommendation device provided in an embodiment of the present application. The business data recommendation device 1 can be a computer program (including program code) running on a computer device, for example, the business data recommendation device 1 is an application software; the device 1 can be used to execute the corresponding steps in the business data recommendation method provided in an embodiment of the present application. Figure 11 As shown, the service data recommendation device 1 may include: a first determination module 11 , a display module 12 , a second determination module 13 , and a sending module 14 .

[0168] The first sending module 11 is configured to respond to an information input operation for a target business and send the business consulting information determined by the information input operation to the second device;

[0169] The display module 12 is used to obtain M target material data matching the business consulting information in the target business and display the M target material data; M is a positive integer;

[0170] The first determining module 13 is configured to respond to a selection operation on the M target material data and determine the target material data determined by the selection operation as the consultation feedback information among the M target material data;

[0171] The second sending module 14 is configured to send the consultation feedback information to the first device.

[0172] The first determining module 13 includes:

[0173] The first acquiring unit 1301 is configured to receive the business consulting information sent by the first device and acquire an information feature vector corresponding to the business consulting information;

[0174] The second acquisition unit 1302 is used to obtain the intention probabilities between the information feature vector and N candidate intentions respectively, where N is a positive integer;

[0175] The first determining unit 1303 is configured to determine a target intent corresponding to the business consulting information from N candidate intents based on the intent probability;

[0176] The third acquisition unit 1304 is configured to acquire M target material data matching the target intent in the target service.

[0177] The first acquiring unit 1301 is specifically configured to:

[0178] Obtain consulting keywords from business consulting information, perform vector conversion on the consulting keywords, and obtain word vectors corresponding to the consulting keywords;

[0179] According to the semantic information corresponding to the word vector, the information feature vector corresponding to the consulting keyword is obtained.

[0180] The second acquiring unit 1302 is specifically configured to:

[0181] Input the information feature vector into the neural network layer in the intent recognition model, and obtain the prediction vector corresponding to the information feature vector based on the weight matrix corresponding to the neural network layer;

[0182] In the classifier of the intent recognition model, the intent probability between the prediction vector and N candidate intents is determined.

[0183] The display module 12 includes:

[0184] The fourth acquisition unit 1201 is used to acquire material attribute features corresponding to the M target material data, and acquire user attribute features corresponding to the target user providing business consulting information;

[0185] The second determining unit 1202 is configured to determine the recommendation evaluation values ​​corresponding to the M target material data according to the material attribute characteristics, the user attribute characteristics, and the information feature vector corresponding to the business consultation information;

[0186] The display unit 1203 is configured to sort the M recommended evaluation values ​​and display the M target material data according to the sorted M recommended evaluation values.

[0187] Among them, the material attribute features include material semantic features and material type features;

[0188] The second determining unit 1202 is specifically configured to:

[0189] Input the information feature vectors corresponding to the material semantic features, user attribute features, and business consultation information into the recommendation model, and determine the prediction evaluation values ​​corresponding to the M target material data respectively according to the recommendation model;

[0190] Obtain weight coefficients corresponding to M target material data respectively according to material type characteristics;

[0191] According to the weight coefficient and the predicted evaluation value, the recommended evaluation values ​​corresponding to the M target material data are determined.

[0192] According to one embodiment of the present application, Figure 2 The steps involved in the business data recommendation method shown can be represented by Figure 11 The business data recommendation device 2 shown in FIG. Figure 2 The step S101 shown in FIG. Figure 11 The first sending module 11 is executed, Figure 2 The step S102 shown in FIG. Figure 11 The display module 12 is executed; Figure 2 The step S103 shown in FIG. Figure 11The first determination module 13 is executed; Figure 2 The step S104 shown in FIG. Figure 11 The second sending module 14 is executed.

[0193] According to one embodiment of the present application, Figure 11 The various modules in the business data recommendation device 1 shown can be individually or all combined into one or several units to constitute, or one (or some) of the units can be further divided into multiple functionally smaller sub-units to achieve the same operation without affecting the realization of the technical effects of the embodiments of the present application. The above modules are divided based on logical functions. In actual applications, the functions of a module can also be implemented by multiple units, or the functions of multiple modules can be implemented by one unit. In other embodiments of the present application, the test device may also include other units. In actual applications, these functions can also be implemented with the assistance of other units, and can be implemented by the collaboration of multiple units.

[0194] In an embodiment of the present application, the first device responds to the information input operation for the target business, and sends the business consultation information determined by the information input operation to the second device. The second device obtains M target material data that match the business consultation information in the target business, and displays the M target material data, where M is a positive integer. Obtaining M target material data that match the business consultation information in the target business can provide reference information for business personnel to determine the consultation feedback information in the M target material data, and can improve the standardization and efficiency of business data recommendations. The second device responds to the selection operation for the M target material data, and determines the target material data determined by the selection operation as the consultation feedback information in the M target material data, and the second device sends the consultation feedback information to the first device. It can be seen that for the business consultation information sent by the first device, the first and second devices can preliminarily screen out M target material data that match the business consultation information from the target business, and display the M target material data. The consultation feedback information corresponding to the business consultation information can be determined from the displayed M target material data, and the consultation feedback information is returned to the first device as the reply information of the business consultation information. This can reduce tedious manual operations, thereby improving the efficiency of business data recommendations. The consultation feedback information determined from the preliminarily screened M target material data can improve the accuracy of business data recommendations. At the same time, this application can also improve the convenience and per capita productivity of business personnel, as well as improve the standardization and professionalism of business data recommendations, and bring users a better consultation experience and better solutions to user problems, thereby improving the service experience corresponding to the target business.

[0195] See Figure 12 , Figure 12This is a structural diagram of a business data recommendation device provided in an embodiment of the present application. The business data recommendation device 2 can be a computer program (including program code) running on a computer device, for example, the business data recommendation device 2 is an application software; the device 2 can be used to execute the corresponding steps in the business data recommendation method provided in the embodiment of the present application. Figure 12 As shown, the service data recommendation device 2 may include: a third determination module 21 , a receiving module 22 , and a returning module 23 .

[0196] The second determining module 21 is configured to respond to an information input operation for a target business and send the business consulting information determined by the information input operation to the second device, so that the second device determines consulting feedback information corresponding to the business consulting information from M target material data included in the target business; M is a positive integer;

[0197] The receiving module 22 is configured to receive the consultation feedback information sent by the second device and display the consultation feedback information.

[0198] The device 2 further comprises:

[0199] The third sending module 23 is used to obtain user behavior data for consultation feedback information and send the user behavior data to the second device so that the second device can update the recommendation model based on the user behavior data; the recommendation model is used to determine the predicted evaluation values ​​corresponding to M target material data, and the predicted evaluation values ​​provide a basis for the second device to determine the consultation feedback information.

[0200] According to one embodiment of the present application, Figure 9 The steps involved in the business data recommendation method shown can be represented by Figure 12 The business data recommendation device 2 shown in FIG. Figure 9 The step S201 shown in FIG. Figure 12 The second determination module 21 is executed, Figure 9 The step S202 shown in FIG. Figure 12 The receiving module 22 in the process executes the above steps.

[0201] According to one embodiment of the present application, Figure 12The various modules in the business data recommendation device 2 shown can be individually or all combined into one or several units to constitute, or one (or some) of the units can be further divided into multiple functionally smaller sub-units to achieve the same operation without affecting the realization of the technical effects of the embodiments of the present application. The above modules are divided based on logical functions. In actual applications, the functions of a module can also be implemented by multiple units, or the functions of multiple modules can be implemented by one unit. In other embodiments of the present application, the test device may also include other units. In actual applications, these functions can also be implemented with the assistance of other units, and can be implemented by the collaboration of multiple units.

[0202] According to one embodiment of the present application, the program can be executed by running on a general computer device such as a computer including a central processing unit (CPU), a random access memory (RAM), a read-only memory (ROM) and other processing elements and storage elements. Figure 2 or Figure 9 A computer program (including program code) for each step involved in the corresponding method shown in Figure 11 The service data recommendation device 1 shown in Figure 12 The business data recommendation device 2 shown in the figure implements the business data recommendation method of the embodiment of the present application. The above computer program can be recorded on a computer-readable recording medium, for example, and loaded into the above computing device through the computer-readable recording medium and run therein.

[0203] In an embodiment of the present application, the consultation feedback information sent by the second device is received by the first device, and the consultation feedback information is displayed, so that the second device can determine the consultation feedback information corresponding to the business consultation information among the M target material data contained in the target business. The first device can receive the consultation feedback information sent by the second device and display the consultation feedback information. The accuracy and efficiency of business data recommendations can be improved by sending the consultation feedback information to the second device through the first device so that the second device can determine the consultation feedback information corresponding to the business consultation information among the M target material data contained in the target business. At the same time, the embodiment of the present application can send the user behavior data for the consultation feedback information to the second device through the first device, so that the second device updates the recommendation model based on the user behavior data, thereby improving the accuracy of the recommendation model in the second device, and then improving the accuracy of the business data recommendation.

[0204] See Figure 13 , Figure 13 This is a schematic diagram of the structure of a computer device provided in an embodiment of the present application. Figure 13As shown, the above-mentioned computer device 1000 may include: a processor 1001, a network interface 1004 and a memory 1005. In addition, the above-mentioned computer device 1000 may also include: a target user interface 1003, and at least one communication bus 1002. The communication bus 1002 is used to realize the connection and communication between these components. The target user interface 1003 may include a display screen (Display), a keyboard (Keyboard), and the target user interface 1003 may optionally include a standard wired interface and a wireless interface. The network interface 1004 may optionally include a standard wired interface and 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 memory. The memory 1005 may optionally also be at least one storage device located away from the aforementioned processor 1001. As Figure 13 As shown, the memory 1005 as a computer-readable storage medium may include an operating system, a network communication module, a target user interface module, and a device control application.

[0205] exist Figure 13 In the computer device 1000 shown, the network interface 1004 can provide network communication functions; the target user interface 1003 is mainly used to provide an input interface for the target user; and the processor 1001 can be used to call the device control application stored in the memory 1005 to achieve:

[0206] Optionally, the processor 1001 may be configured to call a device control application stored in the memory 1005 to implement:

[0207] In response to an information input operation for a target business, sending business consulting information determined by the information input operation to a second device;

[0208] Obtain M target material data matching the business consultation information in the target business, and display the M target material data; M is a positive integer;

[0209] In response to a selection operation on the M target material data, the target material data determined by the selection operation is determined as the consultation feedback information among the M target material data;

[0210] The consultation feedback information is sent to the first device.

[0211] Optionally, the processor 1001 may be configured to call a device control application stored in the memory 1005 to implement:

[0212] In response to an information input operation for a target business, the business consultation information determined by the information input operation is sent to the second device, so that the second device determines consultation feedback information corresponding to the business consultation information from M target material data included in the target business; M is a positive integer;

[0213] Receive the consultation feedback information sent by the second device, and display the consultation feedback information.

[0214] It should be understood that the computer device 1000 described in the embodiment of the present application can execute the above Figure 2 or Figure 9 The description of the business data recommendation method in the corresponding embodiment can also be performed as described above. Figure 11 The corresponding business data recommendation device 1 and Figure 12 The description of the corresponding service data recommendation device 2 will not be repeated here. In addition, the description of the beneficial effects of adopting the same method will not be repeated here either.

[0215] According to one aspect of the present application, a computer program product or computer program is provided, the computer program product or computer program including computer instructions, the computer instructions being stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer device can execute the aforementioned Figure 2 or Figure 9 The description of the service data recommendation 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 here either.

[0216] As an example, the above program instructions may be deployed on a computer device for execution, or deployed on multiple computer devices located at one location for execution, or executed on multiple computer devices distributed at multiple locations and interconnected by a communication network. Multiple computer devices distributed at multiple locations and interconnected by a communication network may constitute a blockchain network.

[0217] Those skilled in the art will appreciate that all or part of the processes in the above-described method embodiments can be implemented by instructing related hardware through a computer program. The above-described program can be stored in a computer-readable storage medium. When executed, the program can include the processes in the above-described method embodiments. The above-described storage medium can be a magnetic disk, an optical disk, a read-only memory (ROM), or a random access memory (RAM).

[0218] 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 business data recommendation method, characterized in that: include: The first device responds to the information input operation for the target business and sends the business consultation information determined by the information input operation to the second device; The second device obtains M target material data matching the business consultation information from the target business; M is a positive integer; The second device obtains material attribute features corresponding to the M target material data respectively, and obtains user attribute features corresponding to the target user who provides the business consulting information; the material attribute features include material semantic features and material type features; The second device inputs the material semantic features, the user attribute features, and the information feature vector corresponding to the business consultation information into a recommendation model, and determines the predicted evaluation values ​​corresponding to the M target material data respectively according to the recommendation model; The second device obtains weight coefficients corresponding to the M target material data respectively according to the material type characteristics; The second device determines, according to the weight coefficient and the predicted evaluation value, the recommended evaluation values ​​corresponding to the M target material data respectively; The second device sorts the M recommended evaluation values, and displays the M target material data according to the sorted M recommended evaluation values; The second device responds to the selection operation on the M target material data and determines, among the M target material data, the target material data determined by the selection operation as the consultation feedback information; The second device sends the consultation feedback information to the first device.

2. The method according to claim 1, characterized in that The second device obtains M target material data matching the business consultation information from the target business, including: The second device receives the business consulting information sent by the first device, and obtains an information feature vector corresponding to the business consulting information; The second device obtains the intention probabilities between the information feature vector and N candidate intentions respectively, where N is a positive integer; The second device determines, from the N candidate intentions according to the intention probability, a target intention corresponding to the business consultation information; The second device obtains M target material data matching the target intent in the target service.

3. The method according to claim 2, characterized in that The obtaining of the information feature vector corresponding to the business consulting information includes: The second device obtains consulting keywords in the business consulting information, performs vector conversion on the consulting keywords, and obtains word vectors corresponding to the consulting keywords; The second device obtains the information feature vector corresponding to the consultation keyword based on the semantic information corresponding to the word vector.

4. The method according to claim 2, characterized in that The second device obtains the intent probabilities between the information feature vector and the N candidate intents, including: The second device inputs the information feature vector into a neural network layer in an intent recognition model, and obtains a prediction vector corresponding to the information feature vector according to a weight matrix corresponding to the neural network layer; The second device determines, in a classifier of the intent recognition model, intent probabilities between the prediction vector and N candidate intents respectively.

5. A business data recommendation method, characterized in that: include: The first device responds to an information input operation for a target business and sends the business consultation information determined by the information input operation to the second device, so that the second device determines consultation feedback information corresponding to the business consultation information from the M target material data included in the target business; M is a positive integer; The first device receives the consultation feedback information sent by the second device, and displays the consultation feedback information; Among them, the display form of the M target material data in the second device is determined by sorting the recommended evaluation values ​​corresponding to the M target material data respectively; the recommended evaluation values ​​corresponding to the M target material data respectively are determined based on the weight coefficients and predicted evaluation values ​​corresponding to the M target material data respectively, the weight coefficients are obtained based on the material type features in the material attribute features corresponding to the M target material data respectively, and the predicted evaluation values ​​are predicted by the recommendation model based on the material semantic features in the material attribute features, the user attribute features corresponding to the target user providing the business consulting information, and the information feature vector corresponding to the business consulting information.

6. The method according to claim 5, characterized in that Also includes: The first device obtains user behavior data for the consultation feedback information and sends the user behavior data to the second device so that the second device updates the recommendation model based on the user behavior data; the recommendation model is used to determine the predicted evaluation values ​​corresponding to the M target material data, and the predicted evaluation values ​​provide a basis for the second device to determine the consultation feedback information.

7. A computer device, characterized in that: include: processor and memory; The memory stores a computer program. When the computer program is executed by the processor, the processor performs the method according to any one of claims 1 to 6.

8. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a computer program, which is suitable for being loaded by a processor and executing the method according to any one of claims 1 to 6.

9. A computer program product, characterized in that The computer program product comprises computer instructions, and when the computer instructions are executed by a processor, the method according to any one of claims 1 to 6 is implemented.

Images