Intelligent matching model construction method, system and equipment applying deep reinforcement learning

Abstract

The invention discloses an intelligent matching model construction method, system and equipment applying deep reinforcement learning, and belongs to the technical field of deep learning and information processing. The method comprises the following steps: screening a matching transmission parameter, a sample transmission parameter and an operation transmission parameter through preset deep reinforcement learning by center equipment; and other matching devices train the configured intelligent matching model according to the screened matching transmission parameters, the screened sample transmission parameters and the screened operation transmission parameters, so that the intelligent matching model of each device performs candidate information matching in a scene of a plurality of users or a plurality of matching devices. According to the candidate information matching method and device, the operation, samples and parameters of the candidate information can be shared to other matching devices, the situation that other devices often need to re-train or migrate the model is avoided, the model is optimized by sharing the operation, the samples and the parameters, the efficiency is improved, and the accuracy in the candidate information matching process is further guaranteed.

Description

Technical Field

[0001] The present invention relates to the fields of deep learning technology and information processing technology, and particularly relates to a method, system and device for constructing an intelligent matching model applying deep reinforcement learning. Background Art

[0002] In the process of existing resume screening or candidate information screening, a matching model and a matching algorithm are often used to implement this process, so as to achieve a relatively high degree of fit in candidate information matching to ensure the efficiency in the candidate screening process;

[0003] However, in practical applications, the matching model or the matching algorithm often becomes different corresponding to the user or the configured device with continuous use by the user, that is, the model often reflects the personalized differences of the user or the user of the configured device in the candidate information matching process. For the same model, with different users and different configured devices, and with continuous use, the model continuously learns, and finally different models will also be obtained;

[0004] However, as is well known, since there are often more than one user or device in the candidate information matching process, and the existing model cannot be shared, that is, for the same batch of candidate information or the same position, in order to achieve fast matching, other devices often need to retrain or migrate the model to achieve the efficiency that can be achieved by the model for long-term use, resulting in low efficiency of the matching model. Summary of the Invention

[0005] To solve the problems of the prior art, embodiments of the present invention provide a method, system and device for constructing an intelligent matching model applying deep reinforcement learning, including:

[0006] On the one hand, a method for constructing an intelligent matching model applying deep reinforcement learning is provided. The method is applied to a distributed intelligent candidate information matching system, and the system includes multiple matching devices, each of which is configured with an intelligent matching model. The method includes:

[0007] After any one of the matching devices retrieves and matches candidate information through the intelligent matching model, obtain the matching information, sample information and operation information corresponding to the matching device;

[0008] Generate a matching transfer parameter according to the matching information;

[0009] Generate a sample transfer parameter according to the sample information;

[0010] Generate an operation transfer parameter according to the operation information;

[0011] Transmit the matching transfer parameter, the sample transfer parameter, and the operation transfer parameter to the central device respectively;

[0012] Upon receiving the matching transfer parameter, the sample transfer parameter, and the operation transfer parameter, the central device performs screening through a preset deep reinforcement learning;

[0013] When other matching devices are in an idle state, they train the configured intelligent matching model according to the screened matching transfer parameter, the screened sample transfer parameter, and the screened operation transfer parameter to obtain an updated intelligent matching model.

[0014] Optionally, the method further includes:

[0015] Construct the intelligent matching model through a wide neural network;

[0016] Train the intelligent matching model according to preset candidate information and corresponding matching information.

[0017] Optionally, after any one of the matching devices retrieves and matches candidate information through the intelligent matching model, obtaining the matching information, sample information, and operation information corresponding to the matching device includes:

[0018] After the user matches candidate information through keywords and descriptive text via the intelligent matching model, record the operations on the candidate information, where the operations at least include marking operations, deletion operations, and adding-to-candidate-list operations;

[0019] Generate the sample information according to the keywords, the descriptive text, and the matched candidate information;

[0020] Generate the matching information according to the first parameter matrix of the intelligent matching model;

[0021] Generate operation information corresponding to the keywords, the descriptive text, and the matched candidate information respectively according to the marking operation, the deletion operation, and the adding-to-candidate-list operation.

[0022] Optionally, the generating the matching transfer parameter according to the matching information includes:

[0023] Generate the corresponding matching transfer parameter according to the first parameter matrix, the keywords, and the descriptive text;

[0024] Transmit the matching transfer parameter to the central device.

[0025] Optionally, the generating the sample transfer parameter according to the sample information includes:

[0026] Perform keyword recognition and extraction on the candidate information after the deletion operation, and generate a first sample transfer parameter based on the first extraction result, the keyword, and the description text;

[0027] Perform keyword recognition and extraction on the candidate information of the candidates who have undergone the deletion operation among the matched candidate information, and generate a second sample transfer parameter based on the second extraction result, the keyword, and the description text;

[0028] Merge the first sample transfer parameter and the second sample transfer parameter into the sample transfer parameter, and transmit it to the central device.

[0029] Optionally, generating an operation transfer parameter according to the operation information includes:

[0030] Obtain the final candidate information after performing the operation;

[0031] Train the intelligent matching model according to the final candidate information, the keyword, and the description text;

[0032] Generate an operation transfer parameter according to the second parameter matrix corresponding to the trained intelligent matching model, the keyword, and the description text;

[0033] Transmit the operation transfer parameter to the central device.

[0034] Optionally, when the central device receives the matching transfer parameter, the sample transfer parameter, and the operation transfer parameter, the screening through preset deep reinforcement learning includes:

[0035] Set the first parameter matrix and the second parameter matrix as the parameter matrices of the initial intelligent matching model respectively;

[0036] Use the sample transfer parameter and the matching transfer parameter as the input values of the training sample, and the operation transfer parameter, the keyword, and the description text as the expected output;

[0037] Output the input values to the initial intelligent matching model and the deep reinforcement learning algorithm respectively;

[0038] Evaluate the matching transfer parameter, the sample transfer parameter, and the operation transfer parameter according to the actual output value and the expected output value, and delete the matching transfer parameter, the sample transfer parameter, and the operation transfer parameter whose difference between the actual output value and the expected output value is greater than the preset value.

[0039] Optionally, training the configured intelligent matching model according to the filtered matching transfer parameters, filtered sample transfer parameters, and filtered operation transfer parameters to obtain an updated intelligent matching model includes:

[0040] Set the first parameter matrix as the first influence coefficient;

[0041] Set the second parameter matrix as the second influence coefficient;

[0042] Adjust the parameter matrix of the intelligent matching model according to the first influence coefficient and the second influence coefficient;

[0043] Use the sample transfer parameter and the matching transfer parameter as the input values of the training sample, and the operation transfer parameter, the keyword, and the description text as the expected output to train the intelligent matching model to obtain the updated intelligent matching model.

[0044] On the other hand, an intelligent matching model construction system applying deep reinforcement learning is provided. The system includes multiple matching devices, and each matching device is configured with an intelligent matching model. The method includes:

[0045] The matching device is used for:

[0046] After retrieving and matching candidate information through the intelligent matching model, obtain the matching information, sample information, and operation information corresponding to the matching device;

[0047] Generate matching transfer parameters according to the matching information;

[0048] Generate sample transfer parameters according to the sample information;

[0049] Generate operation transfer parameters according to the operation information;

[0050] Transmit the matching transfer parameter, the sample transfer parameter, and the operation transfer parameter to the central device respectively;

[0051] The central device is used to perform screening through preset deep reinforcement learning after receiving the matching transfer parameter, the sample transfer parameter, and the operation transfer parameter;

[0052] Other matching devices are used to train the configured intelligent matching model according to the filtered matching transfer parameters, filtered sample transfer parameters, and filtered operation transfer parameters to obtain an updated intelligent matching model when in an idle state.

[0053] On the other hand, an intelligent matching model construction device applying deep reinforcement learning is provided. The device includes:

[0054] An acquisition module, configured to retrieve and match candidate information through the intelligent matching model, and then acquire matching information, sample information, and operation information corresponding to the matching device;

[0055] A processing module, configured to generate matching transfer parameters according to the matching information;

[0056] The processing module is further configured to generate sample transfer parameters according to the sample information;

[0057] The processing module is further configured to generate operation transfer parameters according to the operation information;

[0058] A transmission module, configured to transmit the matching transfer parameters, the sample transfer parameters, and the operation transfer parameters to a central device respectively;

[0059] When in an idle state, train the configured intelligent matching model according to the filtered matching transfer parameters, filtered sample transfer parameters, and filtered operation transfer parameters to obtain an updated intelligent matching model.

[0060] The present invention has at least the following beneficial effects:

[0061] By generating matching transfer parameters according to the matching information, generating sample transfer parameters according to the sample information, generating operation transfer parameters according to the operation information, the central device performs screening through preset deep reinforcement learning, and other matching devices, when in an idle state, train the configured intelligent matching model according to the filtered matching transfer parameters, filtered sample transfer parameters, and filtered operation transfer parameters to obtain an updated intelligent matching model, so that in a scenario of multiple users or multiple matching devices, after each device's intelligent matching model finishes matching candidate information, it can share its own operations, samples, and parameters with other matching devices, avoiding the situation that other devices often need to retrain or migrate the model. By sharing operations, samples, and parameters, the optimization of the model is realized, which not only improves the efficiency but also further ensures the accuracy in the process of candidate information matching. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0063] Figure 1 It is a schematic diagram of a candidate information matching system provided by an embodiment of the present invention;

[0064] Figure 2 Schematic diagram of the process of constructing an intelligent matching model applying deep reinforcement learning provided by an embodiment of the present invention;

[0065] Figure 3 Schematic diagram of the process of constructing an intelligent matching model applying deep reinforcement learning provided by an embodiment of the present invention;

[0066] Figure 4 Schematic diagram of the process of constructing an intelligent matching model applying deep reinforcement learning provided by an embodiment of the present invention;

[0067] Figure 5 Schematic diagram of a system for constructing an intelligent matching model applying deep reinforcement learning provided by an embodiment of the present invention. Detailed implementation manners

[0068] To make the objectives, technical solutions, and advantages of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0069] The candidate information described in the method of the embodiment of the present invention, in practical applications, is often information such as resumes used to describe the skills, resumes, and education backgrounds of candidates. Of course, other methods are also included to implement the above candidate information. The embodiment of the present invention does not limit the specific form of candidate information;

[0070] In addition, it should be noted that, as shown in Figure 1 When the method described in the embodiment of the present invention is mainly applied to a candidate information matching system, the configuration of the system at least meets the following conditions:

[0071] Candidate information is configured in a candidate information server, and the central device is at least configured with a large language model. The above candidate information server and the central device are deployed through the cloud to avoid the deployment of local computing resources and storage resources. The matching device is locally deployed, and an intelligent matching model is deployed on the matching device; the matching device can be a user's personal device, such as a computer, etc.

[0072] Among them, the user initiates an operation through the matching device, such as retrieving by keywords;

[0073] The intelligent matching model retrieves multiple pieces of matching candidate information in the candidate information server based on the keywords input by the user;

[0074] The candidate information server returns the obtained multiple candidate information to the matching device;

[0075] On this matching device, the user continues to perform operations on the multiple candidate information, such as marking operations, deletion operations, and adding to the candidate list operations.

[0076] Among them, this matching device can also be deployed remotely. That is, when a certain entity (such as a company) has matching devices deployed at multiple locations simultaneously, the central device and the candidate server can be shared through cloud services;

[0077] Refer to Figure 2 As shown, a method for constructing an intelligent matching model applying deep reinforcement learning is provided. The method is applied to a distributed candidate information intelligent matching system, and the system includes multiple matching devices, and each matching device is configured with an intelligent matching model. The method includes:

[0078] 101. After retrieving and matching candidate information through the intelligent matching model on any one matching device, obtain the matching information, sample information, and operation information corresponding to the matching device;

[0079] 102. Generate matching transfer parameters according to the matching information;

[0080] 103. Generate sample transfer parameters according to the sample information;

[0081] 104. Generate operation transfer parameters according to the operation information;

[0082] 105. Transmit the matching transfer parameters, sample transfer parameters, and operation transfer parameters to the central device respectively;

[0083] 106. After the central device receives the matching transfer parameters, sample transfer parameters, and operation transfer parameters, perform screening through preset deep reinforcement learning;

[0084] 107. When other matching devices are in an idle state, train the configured intelligent matching model according to the screened matching transfer parameters, screened sample transfer parameters, and screened operation transfer parameters to obtain an updated intelligent matching model.

[0085] Optionally, the method further includes:

[0086] Construct an intelligent matching model through a wide neural network;

[0087] Train the intelligent matching model according to the preset candidate information and the corresponding matching information.

[0088] The above construction process can be specifically:

[0089] Through feature extraction, the training text and input text of the intelligent matching model are constructed. In the feature extraction of text data, the bag-of-words model, TF-IDF, and word embedding can be used;

[0090] After the model is constructed, the constructed BLS is flat in structure, where the original input is converted into random features in the "feature nodes", and then the width is expanded in the "enhanced nodes".

[0091] In BLS, the input data is first converted into random features through some feature mappings, and then connected to the enhanced nodes through a non-linear activation function to realize the training of the model; among them, the feature mapping is the result output by the large language model of the central device; that is, the keywords, description text, and candidate information input by the user are input into the large language model, and the large language model obtains the features corresponding to the keywords, description text, and candidate information input by the user through recognition and feature extraction, and maps the features to the input layer of the intelligent matching model;

[0092] The random features (nodes) are connected to the output layer together with the output of the enhanced layer, and the weights of the output layer are determined by the fast pseudo-inverse of the system equation or the iterative gradient descent training algorithm;

[0093] Set the training data set and n feature mappings φ i , i = 1, 2,..., n, then the i-th mapped feature is:

[0094] Z i = φ i (XW ei + β ei ), i = 1, 2,..., n

[0095] where the weight W ei and the bias term β ei are randomly generated matrices.

[0096] Z n = [Z 1 , Z 2 ,..., Z n represents a set of n groups of feature nodes; Z n is connected to the enhanced node layer.

[0097] Preferably, it can also be through:

[0098] H j = ξ j (Z n W hj + β hj ), j = 1, 2,..., m represents the output of the j-th group of enhanced nodes, where ξj is a non - linear activation function. In addition, the model uses H m = [H 1 , H 1 ,..., H m to represent the output of the enhancement layer.

[0099] Finally, the output Y of BLS is in the following form:

[0100] Y = [Z 1 , Z 2 ,..., Z n , H 1 , H 2 ,..., H m W m

[0101] where W m is the weight connecting the feature node layer and the enhancement node layer to the output layer, and can be calculated by the pseudo - inverse [Z n , H m + .

[0102] After the training is completed, in order to enhance the generalization ability of the model and prevent overfitting, regularization techniques will be adopted. Regularization limits the complexity of model parameters by adding additional terms (such as L1 or L2 penalty terms) to the loss function, thereby reducing the risk of the model over - fitting to the training data.

[0103] Optionally, as shown in Figure 3 , after any matching device retrieves and matches candidate information through the intelligent matching model, the matching information, sample information, and operation information corresponding to the matching device are obtained, including:

[0104] 201. After the user matches candidate information through keywords and descriptive text via the intelligent matching model, record the operations on the candidate information. The operations include at least marking operations, deletion operations, and adding to the candidate list operations;

[0105] After recording the operations on the candidate information, return the multiple candidate information pointed to by the operations to the central device;

[0106] If marking operations and adding to the candidate list operations are performed on candidate information A, then return the text marked by the user, the flag of adding to the candidate list, and the identifier of candidate information A to the central device;

[0107] 202. Generate sample information based on keywords, descriptive text, and the matched candidate information;

[0108] Specifically, this process can be: ​

[0109] Add the keywords, descriptive text, and identifiers of the matching candidate information to a data packet, which is the sample information;

[0110] 203. Generate matching information according to the first parameter matrix of the intelligent matching model;

[0111] Specifically, after a user performs a single search using keywords and descriptive text, multiple candidate information is obtained;

[0112] The user indicates the search results through marking operations, deletion operations, and adding to the candidate list operations;

[0113] That is, when the user performs a deletion operation on a certain candidate information, it indicates that this candidate information is not the expected result of the keywords and descriptive text input by the user;

[0114] When the user performs a marking operation and an adding to the candidate list operation on a certain candidate information, it indicates that this candidate information is the expected result of the keywords and descriptive text input by the user;

[0115] Use the user's keywords and descriptive text as the expected output, and use the multiple candidate information when the user finally performs the marking operation and the adding to the candidate list operation as the input to train the intelligent matching model to obtain the first parameter matrix of the intelligent matching model. Use this first parameter matrix, keywords, descriptive text, and identifiers of multiple candidate information as the matching information;

[0116] 204. Generate operation information corresponding to the keywords, descriptive text, and matching candidate information respectively according to the marking operation, deletion operation, and adding to the candidate list operation.

[0117] Optionally, generate matching transfer parameters according to the matching information, including:

[0118] Generate corresponding matching transfer parameters according to the first parameter matrix, keywords, and descriptive text;

[0119] Transmit the matching transfer parameters to the central device.

[0120] Optionally, generate sample transfer parameters according to the sample information, including:

[0121] Perform keyword recognition and extraction on the candidate information after the deletion operation, and generate the first sample transfer parameter according to the first extraction result, keywords, and descriptive text;

[0122] Perform keyword recognition and extraction on the candidate information that has undergone the deletion operation among the matching candidate information, and generate the second sample transfer parameter according to the second extraction result, keywords, and descriptive text;

[0123] Combine the first sample transfer parameter and the second sample transfer parameter into a sample transfer parameter and transmit it to the central device.

[0124] Optionally, generate an operation transfer parameter according to the operation information, including:

[0125] Obtain the final candidate information after performing the operation;

[0126] Train the intelligent matching model according to the final candidate information, keywords, and description text;

[0127] Generate an operation transfer parameter according to the second parameter matrix corresponding to the trained intelligent matching model, keywords, and description text;

[0128] Transmit the operation transfer parameter to the central device.

[0129] Optionally, after receiving the matching transfer parameter, sample transfer parameter, and operation transfer parameter, the central device performs screening through preset deep reinforcement learning, including:

[0130] Set the first parameter matrix and the second parameter matrix as the parameter matrices of the initial intelligent matching model respectively;

[0131] Use the sample transfer parameter and the matching transfer parameter as the input values of the training sample, and the operation transfer parameter, keywords, and description text as the expected output;

[0132] Output the input values to the initial intelligent matching model and the deep reinforcement learning algorithm respectively;

[0133] Evaluate the matching transfer parameter, sample transfer parameter, and operation transfer parameter according to the actual output value and the expected output value, and delete the matching transfer parameter, sample transfer parameter, and operation transfer parameter whose difference between the actual output value and the expected output value is greater than the preset value.

[0134] Optionally, train the configured intelligent matching model according to the screened matching transfer parameter, screened sample transfer parameter, and screened operation transfer parameter to obtain an updated intelligent matching model, including:

[0135] Set the first parameter matrix as the first influence coefficient;

[0136] Set the second parameter matrix as the second influence coefficient;

[0137] Adjust the parameter matrix of the intelligent matching model according to the first influence coefficient and the second influence coefficient;

[0138] Taking the sample transfer parameter and the matching transfer parameter as the input values of the training sample, and the operation transfer parameter, keyword, and description text as the expected output, train the intelligent matching model to obtain an updated intelligent matching model.

[0139] Referring to Figure 4 As shown, an intelligent matching model construction system applying deep reinforcement learning is provided. The system includes multiple matching devices, and each matching device is configured with an intelligent matching model. The method includes:

[0140] The matching device is used for:

[0141] After retrieving and matching candidate information through the intelligent matching model, obtain the matching information, sample information, and operation information corresponding to the matching device;

[0142] Generate a matching transfer parameter according to the matching information;

[0143] Generate a sample transfer parameter according to the sample information;

[0144] Generate an operation transfer parameter according to the operation information;

[0145] Transmit the matching transfer parameter, sample transfer parameter, and operation transfer parameter to the central device respectively;

[0146] The central device is used to screen through preset deep reinforcement learning after receiving the matching transfer parameter, sample transfer parameter, and operation transfer parameter;

[0147] Other matching devices are used to train the configured intelligent matching model according to the filtered matching transfer parameter, filtered sample transfer parameter, and filtered operation transfer parameter when in an idle state to obtain an updated intelligent matching model.

[0148] Optionally, the matching device is used for:

[0149] Construct an intelligent matching model through a wide neural network;

[0150] Train the intelligent matching model according to the preset candidate information and the corresponding matching information.

[0151] Optionally, the matching device is used for:

[0152] After the user matches candidate information through keywords and description text through the intelligent matching model, record the operations on the candidate information, and the operations at least include marking operations, deletion operations, and adding to the candidate list operations;

[0153] Generate sample information according to the keywords, description text, and matched candidate information;

[0154] Generate matching information according to the first parameter matrix of the intelligent matching model;

[0155] Generate operation information corresponding to the keyword, description text, and matching candidate information respectively according to the marking operation, deletion operation, and adding to the candidate list operation.

[0156] Optionally, the matching device is used for:

[0157] Generate corresponding matching transfer parameters according to the first parameter matrix, keyword, and description text;

[0158] Transmit the matching transfer parameters to the central device.

[0159] Optionally, the matching device is used for:

[0160] Perform keyword recognition and extraction on the candidate information after the deletion operation, and generate the first sample transfer parameters according to the first extraction result, keyword, and description text;

[0161] Perform keyword recognition and extraction on the candidate information of the candidates who have performed the deletion operation among the matching candidate information, and generate the second sample transfer parameters according to the second extraction result, keyword, and description text;

[0162] Merge the first sample transfer parameters and the second sample transfer parameters into sample transfer parameters and transmit them to the central device.

[0163] Optionally, the matching device is used for:

[0164] Obtain the final candidate information after performing the operation;

[0165] Train the intelligent matching model according to the final candidate information, keyword, and description text;

[0166] Generate operation transfer parameters according to the second parameter matrix corresponding to the trained intelligent matching model, keyword, and description text;

[0167] Transmit the operation transfer parameters to the central device.

[0168] Optionally, the central device is used for:

[0169] Set the first parameter matrix and the second parameter matrix as the parameter matrices of the initial intelligent matching model respectively;

[0170] Use the sample transfer parameters and the matching transfer parameters as the input values of the training samples, and the operation transfer parameters, keyword, and description text as the expected outputs;

[0171] Output the input values to the initial intelligent matching model and the deep reinforcement learning algorithm respectively;

[0172] According to the actual output value and the expected output value, evaluate the matching transfer parameters, sample transfer parameters, and operation transfer parameters, and delete the matching transfer parameters, sample transfer parameters, and operation transfer parameters whose difference between the actual output value and the expected output value is greater than the preset value.

[0173] Optionally, the matching device is used for:

[0174] Set the first parameter matrix as the first influence coefficient;

[0175] Set the second parameter matrix as the second influence coefficient;

[0176] According to the first influence coefficient and the second influence coefficient, adjust the parameter matrix of the intelligent matching model;

[0177] Use the sample transfer parameters and matching transfer parameters as the input values of the training samples, and the operation transfer parameters, keywords, and description text as the expected outputs to train the intelligent matching model to obtain an updated intelligent matching model.

[0178] Refer to Figure 5 As shown, an intelligent matching model construction device applying deep reinforcement learning is provided. The device includes:

[0179] An acquisition module, configured to retrieve and match candidate information through the intelligent matching model, and then acquire the matching information, sample information, and operation information corresponding to the matching device;

[0180] A processing module, configured to generate matching transfer parameters according to the matching information;

[0181] The processing module is further configured to generate sample transfer parameters according to the sample information;

[0182] The processing module is further configured to generate operation transfer parameters according to the operation information;

[0183] A transmission module, configured to transmit the matching transfer parameters, sample transfer parameters, and operation transfer parameters to the central device respectively;

[0184] The processing module is further configured to, when in an idle state, train the configured intelligent matching model according to the filtered matching transfer parameters, filtered sample transfer parameters, and filtered operation transfer parameters to obtain an updated intelligent matching model.

[0185] Optionally, the processing module is used for:

[0186] Construct an intelligent matching model through a width neural network;

[0187] Train the intelligent matching model according to the preset candidate information and the corresponding matching information.

[0188] Optionally, after any matching device retrieves and matches candidate information through an intelligent matching model, the matching information, sample information, and operation information corresponding to the matching device are obtained, including:

[0189] After the user matches candidate information through keywords and descriptive text using the intelligent matching model, record the operations on the candidate information. The operations include at least marking operations, deletion operations, and adding to the candidate list operations;

[0190] Generate sample information based on the keywords, descriptive text, and the matched candidate information;

[0191] Generate matching information based on the first parameter matrix of the intelligent matching model;

[0192] Generate operation information corresponding to the keywords, descriptive text, and the matched candidate information respectively based on the marking operation, deletion operation, and adding to the candidate list operation.

[0193] Optionally, the processing module is used for:

[0194] Generate corresponding matching transfer parameters based on the first parameter matrix, keywords, and descriptive text;

[0195] Transmit the matching transfer parameters to the central device.

[0196] Optionally, the processing module is used for:

[0197] Perform keyword recognition and extraction on the candidate information after the deletion operation, and generate first sample transfer parameters based on the first extraction result, keywords, and descriptive text;

[0198] Perform keyword recognition and extraction on the candidate information that has the deletion operation among the matched candidate information, and generate second sample transfer parameters based on the second extraction result, keywords, and descriptive text;

[0199] Merge the first sample transfer parameters and the second sample transfer parameters into sample transfer parameters and transmit them to the central device.

[0200] Optionally, the processing module is used for:

[0201] Obtain the final candidate information after performing the operations;

[0202] Train the intelligent matching model based on the final candidate information, keywords, and descriptive text;

[0203] Generate operation transfer parameters based on the second parameter matrix corresponding to the trained intelligent matching model, keywords, and descriptive text;

[0204] Transmit the operation transfer parameters to the central device.

[0205] Optionally, the processing module is configured to:

[0206] Set the first parameter matrix as the first influence coefficient;

[0207] Set the second parameter matrix as the second influence coefficient;

[0208] Adjust the parameter matrix of the intelligent matching model according to the first influence coefficient and the second influence coefficient;

[0209] Use the sample transfer parameter and the matching transfer parameter as the input values of the training sample, and the operation transfer parameter, the keyword and the description text as the expected output to train the intelligent matching model to obtain an updated intelligent matching model.

[0210] The above several specific embodiments can be combined with each other, and the same or similar concepts or processes may not be repeated in some embodiments.

[0211] The technical features of the above embodiments can be combined arbitrarily (as long as there is no contradiction in the combination of these technical features). For the sake of brevity of description, not all possible combinations of the technical features in the above embodiments are described; these embodiments not explicitly written out should also be considered to be within the scope described in this specification.

[0212] In the above, the present invention has been described in a relatively specific and detailed manner through general descriptions and specific embodiments. It should be noted that, without departing from the concept of the present invention, it is obvious that several deformations and improvements can still be made to these specific embodiments, and these all belong to the protection scope of this application. Therefore, the protection scope of the patent of this application should be subject to the appended claims.

[0213] The above is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. A method for constructing an intelligent matching model using deep reinforcement learning, characterized in that: The method is applied to a distributed candidate information intelligent matching system, the system comprising a plurality of matching devices, each of which is configured with an intelligent matching model, and the method comprises: After any matching device retrieves and matches candidate information through the intelligent matching model, the matching information, sample information and operation information corresponding to the matching device are obtained; Generate matching transfer parameters according to the matching information; Generate sample transfer parameters according to the sample information; Generate operation transfer parameters according to the operation information; Transmitting the matching transfer parameter, the sample transfer parameter and the operation transfer parameter to a central device respectively; After receiving the matching transmission parameter, the sample transmission parameter and the operation transmission parameter, the central device performs screening through preset deep reinforcement learning; When other matching devices are in an idle state, they train the configured intelligent matching model according to the screened matching transfer parameters, the screened sample transfer parameters, and the screened operation transfer parameters to obtain an updated intelligent matching model.

2. The method according to claim 1, characterized in that The method further comprises: Constructing the intelligent matching model through a wide neural network; The intelligent matching model is trained according to the preset candidate information and the corresponding matching information.

3. The method according to claim 2, characterized in that After any one of the matching devices retrieves and matches the candidate information through the intelligent matching model, obtaining the matching information, sample information and operation information corresponding to the matching device includes: After the user matches the candidate information through the keywords and description texts through the intelligent matching model, the operation on the candidate information is recorded, and the operation at least includes a marking operation, a deleting operation, and an adding operation to a candidate list; generating the sample information according to the keywords, the description text and the matching candidate information; generating the matching information according to the first parameter matrix of the intelligent matching model; According to the marking operation, the deleting operation and the adding to candidate list operation, operation information corresponding to the keyword, the description text and the matching candidate information is generated.

4. The method according to claim 3, characterized in that Generating matching transfer parameters according to the matching information includes: Generate corresponding matching transfer parameters according to the first parameter matrix, the keywords and the description text; The matching transfer parameters are transmitted to the central device.

5. The method according to claim 3, characterized in that: Generating sample transfer parameters according to the sample information includes: Perform keyword recognition and extraction on the candidate information after the deletion operation, and generate a first sample transfer parameter according to the first extraction result, the keyword and the description text; Perform keyword recognition and extraction on the candidate information for which the deletion operation is performed in the matched candidate information, and generate a second sample transfer parameter according to the second extraction result, the keyword and the description text; The first sample transfer parameter and the second sample transfer parameter are combined into the sample transfer parameter, and transmitted to the central device.

6. The method according to claim 3, characterized in that: Generating the operation transfer parameter according to the operation information includes: Get the final candidate information after the operation is performed; Training the intelligent matching model according to the final candidate information, the keywords and the description text; Generate operation transfer parameters according to the second parameter matrix corresponding to the trained intelligent matching model, the keywords and the description text; The operation transfer parameter is transmitted to the central device.

7. The method according to claim 6, characterized in that After receiving the matching transmission parameter, the sample transmission parameter and the operation transmission parameter, the central device performs screening through preset deep reinforcement learning, including: Setting the first parameter matrix and the second parameter matrix as parameter matrices of an initial intelligent matching model respectively; The sample transfer parameter and the matching transfer parameter are used as input values ​​of the training sample, and the operation transfer parameter, the keyword and the description text are used as expected outputs; Outputting the input values ​​to the initial intelligent matching model and the deep reinforcement learning algorithm respectively; The matching transfer parameters, the sample transfer parameters and the operation transfer parameters are evaluated according to the actual output value and the expected output value, and the matching transfer parameters, the sample transfer parameters and the operation transfer parameters whose difference between the actual output value and the expected output value is greater than a preset value are deleted.

8. The method according to claim 7, characterized in that The step of training the configured intelligent matching model according to the screened matching transfer parameters, the screened sample transfer parameters, and the screened operation transfer parameters to obtain an updated intelligent matching model includes: Setting the first parameter matrix to a first influence coefficient; Setting the second parameter matrix to a second influence coefficient; Adjusting the parameter matrix of the intelligent matching model according to the first influence coefficient and the second influence coefficient; The sample transfer parameter and the matching transfer parameter are used as input values ​​of the training sample, and the operation transfer parameter, the keyword and the description text are used as expected outputs to train the intelligent matching model to obtain the updated intelligent matching model.

9. An intelligent matching model construction system using deep reinforcement learning, characterized in that: The system includes a plurality of matching devices, each of which is configured with an intelligent matching model. The method includes: The matching device is used for: After retrieving and matching the candidate information through the intelligent matching model, the matching information, sample information and operation information corresponding to the matching device are obtained; Generate matching transfer parameters according to the matching information; Generate sample transfer parameters according to the sample information; Generate operation transfer parameters according to the operation information; Transmitting the matching transfer parameter, the sample transfer parameter and the operation transfer parameter to a central device respectively; The central device is used to perform screening through preset deep reinforcement learning after receiving the matching transmission parameter, the sample transmission parameter and the operation transmission parameter; The other matching devices are used to train the configured intelligent matching model according to the screened matching transfer parameters, the screened sample transfer parameters and the screened operation transfer parameters when in an idle state to obtain an updated intelligent matching model.

10. An intelligent matching model construction device using deep reinforcement learning, characterized in that: The device comprises: An acquisition module, configured to retrieve and match candidate information through the intelligent matching model, and then acquire matching information, sample information, and operation information corresponding to the matching device; A processing module, used for generating matching transfer parameters according to the matching information; The processing module is also used to generate sample transfer parameters according to the sample information; The processing module is also used to generate operation transfer parameters according to the operation information; A transmission module, used for transmitting the matching transfer parameter, the sample transfer parameter and the operation transfer parameter to a central device respectively; When in an idle state, the configured intelligent matching model is trained according to the screened matching transfer parameters, the screened sample transfer parameters, and the screened operation transfer parameters to obtain an updated intelligent matching model.

Images