Message distribution methods, apparatus, computer equipment, storage media and products

Abstract

This application relates to a message distribution method, apparatus, computer equipment, storage medium, and computer program product, belonging to the field of computer technology, and can be used in the financial technology field or other related fields, improving message distribution efficiency. The method includes: receiving a message to be distributed sent by a message sender to a message distribution end; acquiring multiple message information from the message to be distributed, and removing interfering message information from the multiple message information, obtaining target message information based on the remaining message information after removal; the interfering message information is message information whose correlation with the message receiver is less than a preset threshold; inputting the target message information into a trained message sorting model, and having the message sorting model determine the target message receiver corresponding to the message to be distributed based on the target message information; and sending the message to be distributed to the target message receiver.

Description

Technical Field

[0001] This application relates to the field of computer technology, and in particular to a message distribution method, apparatus, computer equipment, storage medium, and computer program product. Background Technology

[0002] With the development of computer technology, to improve the efficiency of message exchange between financial institutions in different regions, their branches can join global message service providers to achieve information exchange with branches in different regions. To improve service efficiency, financial institutions can select only some of their branches to join the message service provider. For each branch's sub-branches, the branch can distribute multiple received messages to the corresponding sub-branches.

[0003] In related technologies, after a business outlet receives a message sent by a message agency, the business outlet can distribute the message containing one or more keywords to the corresponding sub-business outlet based on the keyword matching rule table maintained by the business personnel according to their experience.

[0004] However, the above method relies entirely on maintaining the keyword matching rule table. A large number of messages cannot be identified to their corresponding sub-business locations using this table, requiring manual sorting. Furthermore, business personnel need to spend a significant amount of time maintaining the table. Therefore, the relevant technology still suffers from low message distribution efficiency. Summary of the Invention

[0005] Therefore, it is necessary to provide a message distribution method, apparatus, computer equipment, computer-readable storage medium, and computer program product that can improve message distribution efficiency in response to the above-mentioned technical problems.

[0006] Firstly, this application provides a message distribution method, the method comprising:

[0007] Receive the message to be distributed sent by the message sender to the message distributor;

[0008] The system acquires multiple message information from the message to be distributed, removes interfering message information from the multiple message information, and obtains the target message information based on the remaining message information after removal; the interfering message information is message information whose correlation with the message receiving end is less than a preset threshold.

[0009] The target message information is input into the trained message sorting model, and the message sorting model determines the target message receiving end corresponding to the message to be distributed based on the target message information.

[0010] The message to be distributed is sent to the target message receiver.

[0011] In one embodiment, before removing interfering message information from the plurality of message information, the method further includes:

[0012] Obtain the service type of the message to be distributed and multiple service elements associated with the service type;

[0013] Identify the interfering service elements among the multiple service elements whose correlation with the message receiver is less than a preset threshold;

[0014] Based on the interference service elements, the interference message information among the multiple message information is determined.

[0015] In one embodiment, determining the interfering service elements among the plurality of service elements whose correlation with the message receiver is less than a preset threshold includes:

[0016] Obtain a pre-stored configuration table; the configuration table records multiple candidate service elements, and the correlation between each candidate service element and the message receiver is less than a preset threshold.

[0017] Based on the matching results between the multiple business elements and each candidate business element in the configuration table, the interfering business elements among the multiple business elements are determined.

[0018] In one embodiment, before removing interfering message information from the plurality of message information, the method further includes:

[0019] Obtain multiple message blocks of the message to be distributed; the multiple message blocks correspond to different topics;

[0020] Interference message blocks are identified among the plurality of message blocks, and the message information in the interference message blocks is identified as interference message information; the correlation between the subject of the interference message block and the message receiving end is less than a preset threshold.

[0021] In one embodiment, before the message to be distributed is sent from the message sending end to the message distribution end, the method further includes:

[0022] Receive messages uploaded by multiple message receivers;

[0023] Based on a preset, agreed-upon encryption key, the messages uploaded by the multiple message receiving terminals are encrypted to obtain an encrypted message set; the agreed-upon encryption key is an encryption key pre-agreed upon by the message distribution terminal and the message sending terminal.

[0024] The encrypted message set is sent to the message sending end to trigger the message sending end to return multiple messages to be distributed to the message distribution end based on the messages in the encrypted message set.

[0025] In one embodiment, the encryption of messages uploaded by the plurality of message receiving terminals based on a preset agreed encryption key to obtain an encrypted message set includes:

[0026] Based on the encryption key corresponding to each message receiver, the message uploaded by each message receiver is encrypted to obtain multiple encrypted messages;

[0027] Based on a preset, agreed-upon encryption key, the multiple encrypted messages are encrypted to obtain a set of encrypted messages.

[0028] In one embodiment, the message sorting model is trained through the following steps:

[0029] For each message uploaded by the multiple message receiving terminals, the interfering message information in each message is removed to obtain multiple sample messages;

[0030] Based on the message receiver corresponding to each message, determine the receiver tag for each sample message;

[0031] Based on the multiple sample messages and the receiver label of each sample message, a text classification model is trained, and the trained message sorting model is obtained when the training termination condition is met.

[0032] Secondly, this application also provides a message distribution apparatus, the apparatus comprising:

[0033] The message receiving module is used to receive messages to be distributed sent by the message sending end to the message distribution end;

[0034] The interference message information removal module is used to acquire multiple message information from the message to be distributed, remove interference message information from the multiple message information, and obtain target message information based on the remaining message information after removal; the interference message information is message information whose correlation with the message receiving end is less than a preset threshold.

[0035] The sorting module is used to input the target message information into the trained message sorting model, and the message sorting model determines the target message receiving end corresponding to the message to be distributed based on the target message information.

[0036] The message forwarding module is used to send the message to be distributed to the target message receiver.

[0037] Thirdly, this application also provides a computer device. The computer device includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to perform the following steps:

[0038] Receive the message to be distributed sent by the message sender to the message distributor;

[0039] The system acquires multiple message information from the message to be distributed, removes interfering message information from the multiple message information, and obtains the target message information based on the remaining message information after removal; the interfering message information is message information whose correlation with the message receiving end is less than a preset threshold.

[0040] The target message information is input into the trained message sorting model, and the message sorting model determines the target message receiving end corresponding to the message to be distributed based on the target message information.

[0041] The message to be distributed is sent to the target message receiver.

[0042] Fourthly, this application also provides a computer-readable storage medium. The computer-readable storage medium stores a computer program thereon, which, when executed by a processor, performs the following steps:

[0043] Receive the message to be distributed sent by the message sender to the message distributor;

[0044] The system acquires multiple message information from the message to be distributed, removes interfering message information from the multiple message information, and obtains the target message information based on the remaining message information after removal; the interfering message information is message information whose correlation with the message receiving end is less than a preset threshold.

[0045] The target message information is input into the trained message sorting model, and the message sorting model determines the target message receiving end corresponding to the message to be distributed based on the target message information.

[0046] The message to be distributed is sent to the target message receiver.

[0047] Fifthly, this application also provides a computer program product. The computer program product includes a computer program that, when executed by a processor, performs the following steps:

[0048] Receive the message to be distributed sent by the message sender to the message distributor;

[0049] The system acquires multiple message information from the message to be distributed, removes interfering message information from the multiple message information, and obtains the target message information based on the remaining message information after removal; the interfering message information is message information whose correlation with the message receiving end is less than a preset threshold.

[0050] The target message information is input into the trained message sorting model, and the message sorting model determines the target message receiving end corresponding to the message to be distributed based on the target message information.

[0051] The message to be distributed is sent to the target message receiver.

[0052] The aforementioned message distribution method, apparatus, computer equipment, storage medium, and computer program product, after receiving a message to be distributed from a message sender, can acquire multiple message information from the message to be distributed, remove interfering message information from the multiple message information, and obtain target message information based on the remaining message information after removal. Interfering message information refers to message information whose correlation with the message receiver is less than a preset threshold. Then, the target message information can be input into a trained message sorting model, which determines the target message receiver corresponding to the message to be distributed based on the target message information, and then sends the message to be distributed to the target message receiver. In this application, after removing interfering message information from the messages to be distributed, the messages are input into the message sorting model and the model determines the corresponding target message receiver. On the one hand, the message sorting model can be used to analyze and determine the message receiver, avoiding manual sorting. On the other hand, by removing interfering information in the messages that is irrelevant to determining the receiver, the classification difficulty of the message sorting model is reduced, thereby effectively improving the message distribution efficiency. Attached Figure Description

[0053] Figure 1 This is a diagram illustrating the application environment of a message distribution method in one embodiment.

[0054] Figure 2 This is a flowchart illustrating a message distribution method in one embodiment;

[0055] Figure 3 This is a flowchart illustrating one step in determining interference message information in one embodiment;

[0056] Figure 4 This is a flowchart illustrating one step in determining interfering service elements in one embodiment;

[0057] Figure 5 This is a schematic diagram of the structure of a message processing device in one embodiment;

[0058] Figure 6 This is a structural block diagram of a message distribution device in one embodiment;

[0059] Figure 7 This is an internal structural diagram of a computer device in one embodiment. Detailed Implementation

[0060] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0061] It should be noted that the message distribution method and apparatus of this application can be used in the financial field for message distribution-related processing, and can also be used in message distribution-related processing in any field other than the financial field. The application fields of the message distribution method, apparatus, computer equipment, storage medium and computer program products of this application are not limited.

[0062] The message distribution method provided in this application embodiment can be applied to, for example, Figure 1 The application environment shown can include a message sender, a message distributor, and a message receiver. The message sender and the message distributor can communicate through a network, and the message distributor can also communicate with the message receiver through a network.

[0063] In this application, the message distribution end can receive the message to be distributed sent by the message sender to the message distribution end, and then obtain multiple message information in the message to be distributed, and remove the interfering message information from the multiple message information, and obtain the target message information based on the remaining message information after removal; the interfering message information is the message information whose correlation with the message receiver is less than a preset threshold; then the message distribution end can input the message information of the target message item into the trained message sorting model, and the message sorting model determines the target message receiver corresponding to the message to be distributed based on the message information, and the message distribution end can send the message to be distributed to the target message receiver.

[0064] The message sending end and message distribution end can be implemented using independent servers or a server cluster consisting of multiple servers. The message sending end and message distribution end can have a data storage system, which can store the data that the message distribution end needs to process. For example, the data storage system of the message distribution end can store messages to be forwarded to the message receiving end and messages to be distributed to the message receiving end. The data storage system can be integrated on the message distribution end or placed on the cloud or other network servers.

[0065] The message receiver can be a terminal, which can be, but is not limited to, various personal computers, laptops, tablets, IoT devices, etc.; of course, the message receiver can also be implemented using a standalone server or a server cluster composed of multiple servers.

[0066] In an optional embodiment, the message distribution end can be the highest-level business outlet among multiple business outlets in the business area, and the message receiving end can be any business outlet in the business area other than the message distribution end; for example, if the business area is divided by province, the message distribution end can be a provincial business outlet, and the message receiving end can be any city-level business outlet and / or business outlets below the city level.

[0067] In one embodiment, such as Figure 2 As shown, a message distribution method is provided, which can be applied to... Figure 1 Taking the message distribution end in the middle as an example, the following steps can be included:

[0068] S201, Receive the message to be distributed sent by the message sender to the message distributor.

[0069] In practice, after receiving a message for a specified message receiver, the message receiver can send the message to the message distributor. The message distributor can then treat the message from the message sender as a message to be distributed. The message sender can send one or more messages (i.e., two or more) to the message receiver each time.

[0070] S202, obtain multiple message information from the message to be distributed, remove interfering message information from the multiple message information, and obtain the target message information based on the remaining message information after removal; the interfering message information is the message information whose correlation with the message receiver is less than a preset threshold.

[0071] As an example, the correlation between interference message information and message receivers can be understood as the degree of one-to-one correspondence between interference message information and message receivers. In practical applications, some message information can be used to identify the corresponding message receiver of a message to be distributed or to assist in determining the corresponding message receiver of a message to be distributed. The higher the correlation between message information and message receivers, the higher the correspondence between message information and message receivers can be. In some optional embodiments, when the preset threshold is zero, interference message information is irrelevant message information. There is no correspondence between irrelevant message information and message receivers, which can be understood as message information unrelated to determining a specific message receiver.

[0072] Specifically, after obtaining the message to be distributed, the message distribution end can parse the message to be distributed to obtain multiple message information in the message to be distributed. Then, it can identify the interfering message information in the multiple message information, remove the interfering message information in the multiple message information, and take the remaining message information after removal as the target message information.

[0073] S203, the target message information is input into the trained message sorting model, and the message sorting model determines the target message receiver corresponding to the message to be distributed based on the target message information.

[0074] Among them, the message sorting model can be used to determine the message receiving end.

[0075] In this step, the target message information of the message to be distributed can be input into the trained message sorting model. The message sorting model extracts features from the target message information and performs text classification based on the feature extraction results to determine the message receiving end corresponding to the message to be distributed, which is then used as the target message receiving end.

[0076] S204, send the message to be distributed to the target message receiver.

[0077] After the message distribution end determines the target message receiver of the message to be distributed, the message distribution end can send the message to be distributed to the target message receiver.

[0078] In this embodiment, after receiving a message to be distributed from a message sender, the message distribution end can obtain multiple message information from the message to be distributed, remove interfering message information from these multiple message information, and obtain the target message information based on the remaining message information after removal. Interfering message information refers to message information whose correlation with the message receiver is less than a preset threshold. Then, the target message information can be input into a trained message sorting model. The message sorting model determines the target message receiver corresponding to the message to be distributed based on the target message information, and then sends the message to be distributed to the target message receiver. In this application, by removing interfering message information from the message to be distributed and inputting it into the message sorting model, the model can determine the corresponding target message receiver. This approach not only allows for the analysis and determination of the message receiver using the message sorting model, avoiding manual sorting, but also reduces the classification difficulty of the message sorting model by removing interfering information irrelevant to the receiver, thereby effectively improving message distribution efficiency.

[0079] In one embodiment, such as Figure 3 As shown, before removing interfering message information from multiple message messages, the method may further include the following steps:

[0080] S301, obtain the service type of the message to be distributed and the multiple service elements associated with the service type.

[0081] As an example, a business element can also be called a message item (FIELD). It is a number of pieces of information involved in the processing of a business corresponding to a business type. Business elements can be associated with the business object (such as the message receiver), processing method, or processing result.

[0082] In practical applications, multiple messages can be messages for different business types. In one example, multiple business types of messages can be obtained based on the business type of the financial business, such as foreign exchange transactions, remittance transfers, etc. Accordingly, since the business of different business types is different, each business can have corresponding business elements, and the business elements associated with each business type can be stored in advance.

[0083] After receiving the message to be distributed, the service type of the message can be obtained. In one optional embodiment, the message receiver can determine the service type based on the service type identifier carried in the message to be distributed. For example, the message to be distributed may carry service type identifiers such as MT100 and MT201.

[0084] Furthermore, once the service type of the message to be distributed is determined, the service elements associated with the service type of the message to be distributed can be determined based on the pre-stored service elements associated with each service type.

[0085] S302, identify the interfering service elements among multiple service elements whose correlation with the message receiver is less than a preset threshold.

[0086] Among them, the degree of correlation between business elements and message receivers can be understood as the degree to which business elements correspond one-to-one with message receivers.

[0087] Specifically, among the multiple business elements of the message to be distributed, there may be information related to the message receiving end, or information related to the processing method or processing result. Among the information related to the processing method or processing result, there may be information with low or no correlation to the message receiving end. Based on this, after determining the multiple associated business elements of the message to be distributed, interfering business elements with a correlation to the message receiving end less than a preset threshold can be identified from among these multiple business elements.

[0088] S303, based on the interference service elements, determine the interference message information among multiple message messages.

[0089] After identifying the interfering business elements, the interfering message information among multiple message information in the message to be distributed is determined based on these elements. Specifically, message information under the interfering business elements can be identified as interfering message information. For example, for message items FIELD 32 and FIELD 71 in the MT100 message, FIELD 32 records information such as the value date, currency, and amount, while FIELD 71 records the cost burden and has a weak correlation with the message receiver. For message items FIELD19, FIELD 30, and FIELD 32 in the MT201 message, FIELD19 records the total position, FIELD 30 records the value date, and FIELD 32 records information such as the value date, currency, and amount.

[0090] In this embodiment, by combining the two dimensions of the business type and business elements of the message to be distributed, interference message information with weak correlation to the message receiver can be quickly filtered out from multiple message information of the message to be distributed. This helps to quickly remove content that is not related to message sorting in the message to be distributed and improve the sorting efficiency of the message sorting model.

[0091] In one embodiment, such as Figure 4 As shown, S302 identifies interfering service elements among multiple service elements whose correlation with the message receiver is less than a preset threshold, which may include the following steps:

[0092] S401, retrieve the pre-stored configuration table; the configuration table records multiple candidate service elements, and the correlation between each candidate service element and the message receiver is less than a preset threshold.

[0093] In practice, a configuration table can be pre-stored, which records multiple candidate business elements. The candidate business elements are those whose correlation with the message receiver is less than a preset threshold.

[0094] In one embodiment, the multiple candidate business elements recorded in the configuration table can be set and maintained by relevant business personnel. Of course, in other embodiments, they can also be determined based on the sorting results of the message sorting model. For example, for messages that are sorted incorrectly by the message sorting model, the common business elements in multiple messages that are sorted incorrectly can be determined, and candidate business elements can be further identified based on the common business elements.

[0095] S402, based on the matching results of multiple business elements with each candidate business element in the configuration table, determines the interfering business elements among the multiple business elements.

[0096] After obtaining the configuration table, the multiple business elements of the message to be distributed can be matched one by one with the candidate business elements in the configuration table. If the current business element matches any candidate business element in the configuration table, the current business element can be identified as an interfering business element. Otherwise, the current business element can be ignored, and the next business element can be used as the current business element to continue matching with the candidate business elements in the configuration table until all business elements in the message to be distributed have been processed. In this way, the interfering business elements among the multiple business elements can be identified.

[0097] In this embodiment, by matching the pre-maintained and stored configuration table with the various service elements of the message to be distributed, the interfering service elements among the multiple service elements can be quickly identified.

[0098] In one embodiment, before removing interfering message information from the plurality of message information, the method may further include the following steps:

[0099] Obtain multiple message blocks of the message to be distributed; identify the interfering message blocks among the multiple message blocks, and determine the message information in the interfering message blocks as interfering message information.

[0100] Among them, multiple message blocks correspond to different topics; the correlation between the topic of the interfering message block and the message receiver is less than a preset threshold. The correlation between the topic and the message receiver can be understood as the degree to which the content under the topic corresponds one-to-one with the message receiver.

[0101] In practice, the message information in the message to be distributed can be divided into different message blocks, and each message block can record content corresponding to different topics. For example, the message block corresponding to the message header can record the port number corresponding to the source port of the message sender; some message blocks can record the business elements of the message, that is, the topic of this part of the message is the various business elements of the message.

[0102] After obtaining the message to be distributed, multiple message blocks of the message to be distributed can be obtained, and the topic of each message block can be determined. For example, the topic of the message block can be determined by a pre-trained text recognition model, or the order of each topic in the message can be determined in advance, and then the topic corresponding to each message block can be obtained according to the order of each message block in the message.

[0103] After determining the subject of each message block, interfering message blocks can be identified from multiple message blocks, and each message information in the interfering message block can be identified as interfering message information. For example, the message block BLOCK 5 (tail) of the MT100 message and the MT201 message can be identified as interfering message information.

[0104] In this embodiment, interference message information in a message can be quickly determined from the perspective of message block.

[0105] In one embodiment, before receiving the message to be distributed sent by the message sender to the message distributor, the following steps may also be included:

[0106] Receive messages uploaded by multiple message receivers; encrypt the messages uploaded by multiple message receivers based on a preset agreed encryption key to obtain an encrypted message set; send the encrypted message set to the message sender to trigger the message receivers to return multiple messages to be distributed to the message distributor based on the messages in the encrypted message set.

[0107] The agreed encryption key is an encryption key pre-agreed upon by the message distribution end and the message sending end. The message sending end may have an agreed decryption key corresponding to the agreed encryption key.

[0108] In practical applications, a message distribution terminal can receive messages uploaded by multiple message receiving terminals. In one embodiment, when a message receiving terminal is a business outlet in a business area, the message receiving terminal can receive messages uploaded by other business outlets (i.e., multiple message receiving terminals) belonging to the same business area.

[0109] After receiving multiple messages from multiple message receivers, the message distributor can obtain a preset agreed-upon encryption key, use the agreed-upon encryption key to encrypt the messages uploaded by the multiple message receivers, obtain an encrypted message set, and send the encrypted message set to the message sender.

[0110] Upon receiving the encrypted message set, the message sender can decrypt the encrypted message set using a pre-agreed decryption key. Then, the message sender can respond to the messages in the message set, returning multiple messages to be distributed to the message distributor, which then sends the corresponding messages to the appropriate message receiver. In an optional embodiment, the message distributor and message receiver can use the DES algorithm to encrypt the message set obtained by packaging multiple messages.

[0111] In this embodiment, the message distribution end uses an agreed-upon encryption key to encrypt messages uploaded from multiple message receiving ends, which ensures the security of the packaged message set when it is transmitted to the message sending end.

[0112] In one embodiment, encrypting messages uploaded by multiple message receivers based on a preset, agreed-upon encryption key to obtain an encrypted message set may include the following steps:

[0113] Based on the encryption key corresponding to each message receiver, the message uploaded by each message receiver is encrypted to obtain multiple encrypted messages; based on the preset agreed encryption key, the multiple encrypted messages are encrypted to obtain a set of encrypted messages.

[0114] In a specific implementation, the message distribution end and each message receiving end can pre-determine corresponding encryption and decryption keys. For each message uploaded by a message receiving end, the message distribution end can encrypt the message based on the encryption key corresponding to that message receiving end, obtaining an encrypted message. This encrypted message can be decrypted by a device holding the decryption key (such as the message receiving end or other authorized devices), thereby preventing the leakage of message content. In an optional embodiment, the message uploaded by each message receiving end can be encrypted using the RSA (RSA algorithm) encryption algorithm.

[0115] After receiving multiple encrypted messages, the message distribution end can encrypt the multiple encrypted messages based on a preset agreed encryption key to obtain a set of encrypted messages.

[0116] In this embodiment, encrypting a single message using the encryption key corresponding to the message receiver ensures the security of the message itself. Furthermore, encrypting a file containing multiple messages using an agreed-upon encryption key ensures the security of the entire message set, achieving excellent results in both encryption speed and system security.

[0117] In one embodiment, the message sorting model can be trained through the following steps:

[0118] S1: For each message uploaded by multiple message receivers, remove the interference message information in each message to obtain multiple sample messages.

[0119] In practice, after the message distribution end receives messages uploaded by multiple message receiving ends, it can remove interference message information from each uploaded message and determine each message with removed interference message information as a sample message, thereby obtaining multiple sample messages.

[0120] S2, based on the message receiver corresponding to each message, determine the receiver tag for each sample message.

[0121] Since the message distribution end can clearly identify the message receiving end that sent each message after receiving each message uploaded by the message receiving end, and thus determine the specific source of each message, the message receiving end corresponding to each message can be identified as the receiving end tag.

[0122] S3. Based on multiple sample messages and the receiver label of each sample message in the multiple sample messages, a text classification model is trained, and the trained message sorting model is obtained when the training termination condition is met.

[0123] After obtaining multiple sample messages and the receiver label for each sample message, the text classification model to be trained can be trained. In an optional embodiment, the text classification model can employ the Fasttext algorithm. Fasttext is a fast text classification algorithm whose model architecture is similar to CBOWT in word2vec, except that Fasttext predicts labels (e.g., it can predict the specific message receiver) while CBOWT predicts intermediate words.

[0124] In the Fasttext algorithm, x1, x2, ..., x N-1 ,x N The input is fed into the hidden layer of the Fasttext model, and the prediction result is obtained based on the output of the hidden layer, where x1, x2, ..., x N-1 ,x N This represents an n-gram vector in a text, where each feature is the average of the word vectors. In practical applications, after receiving a message with interference removed, the FastText algorithm can use a language model-based n-gram algorithm to slide a window of size N through the text content in byte order, ultimately forming a sequence of byte segments with a window size of N. The advantages of using this method include: 1. Generating better word vectors for rare words in the message; 2. Constructing word vectors from character-level n-grams even if the words in the message do not appear in the training corpus; 3. Allowing the model to learn partial information about local word order.

[0125] In some optional embodiments, Fasttext uses a hierarchical softmax function, which acts as an activation function in the output layer of the neural network to normalize the values ​​of the output layer. For example, for the current message to be distributed, if the prediction result for line A is z1 = 3, the prediction result for line B is z2 = 1, and the prediction result for line C is z3 = -3, then softmax is used for normalization. Specifically, firstly, the prediction results can be converted into non-negative numbers. For the three prediction results mentioned above, the following calculations can be performed to obtain the corresponding converted results X1, X2, and X3:

[0126] X1 = exp(z1) = exp(3) = 20

[0127] X² = exp(z²) = exp(1) = 2.7

[0128] X3 = exp(z3) = exp(-3) = 0.05

[0129] Then, the sum of the probabilities of all predicted outcomes can be made equal to 1, that is:

[0130] Y1 = 20 / (20 + 2.7 + 0.05) ≈ 0.88

[0131] Y2 = 2.7 / (20 + 2.7 + 0.05) ≈ 0.12

[0132] Y3 = 0.05 / (20 + 2.7 + 0.05) ≈ 0

[0133] Therefore, based on Y1, Y2, and Y3, we can obtain the probability that the currently input message belongs to line A, line B, or line C.

[0134] In some embodiments, to optimize the efficiency of standard softmax in normalization, the complexity can be reduced from N to logN by constructing a Huffman tree using hierarchical softmax:

[0135] In the Huffman tree model, leaf nodes do not have directly output vectors for their words, while non-leaf nodes do. A Huffman tree is constructed using Huffman coding, assigning vectors to non-leaf nodes. Therefore, the probability of the target word W can be expressed as:

[0136] p(w)=∏σ(sign(w,j)·θ n(w,j) T h)

[0137] Where θn(w,j) is the vector representation of the non-leaf node n(w,j); h is the output value of the hidden layer, calculated from the vectors of the input words; and sign(x,j) is a special function definition. Simultaneously, the sum of the probabilities of all words is 1, that is:

[0138]

[0139] In some cases, as the corpus increases, memory requirements will also increase, affecting the model's processing speed. The following solutions can be used to address this problem: 1. Filter out words that appear infrequently; 2. Use hash storage; 3. Change the word granularity to word granularity.

[0140] During the training of the text classification model, increasing the number of training epochs from 5 to 30 can increase the learning rate h (i.e., increase the step size of gradient descent), making it closer to the optimal point with a limited number of iterations. In addition, 2-gram features can be added to help the model capture the correlation between words in the message and better extract classification rules for model classification.

[0141] In one embodiment, the loss parameter in the train_supervised method can be set to modify the loss calculation method (equivalent to the structure of the output layer). In one example, it can be a softmax layer structure, for example, it can be set as: model = fasttext.train_supervised(input = "jsp.train", lr = 1.0, epoch = 30, wordNgrams = 2, loss = 'hs'). Furthermore, model parameters may be correlated, and different datasets require different hyperparameters, making manual tuning and finding hyperparameters difficult. To address this, the `autotuneValidationFile` parameter of FastText can be used for automatic hyperparameter tuning. The `autotuneValidationFile` parameter specifies the path to the validation dataset, 'jsp.valid'. It will use a random search method on the validation set to determine the potentially optimal hyperparameters. When determining the optimal hyperparameters, the `autotuneDuration` parameter controls the random search time, which can be extended or shortened depending on different needs. In some examples, a random search time of 600 seconds can be selected, i.e., set to: `model = fasttext.train_supervised(input = 'jsp.train', autotuneValidationFile = 'jsp.valid', autotuneDuration = 600)`.

[0142] In addition, for multi-label multi-classification problems, since multiple receiver labels are set, and softmax generally only maximizes one label, it is also possible to choose to use an independent binary classifier as the output layer structure for each receiver label. The corresponding loss calculation method is 'ova' to represent one vs all. This change representation of the output layer can train multiple binary classification models simultaneously under the same corpus.

[0143] In this embodiment, by utilizing messages uploaded by multiple message receivers, a large amount of corpus can be quickly obtained for training the message sorting model, providing a foundation for improving message sorting efficiency in the future.

[0144] To enable those skilled in the art to better understand the above steps, the following example illustrates the embodiments of this application, but it should be understood that the embodiments of this application are not limited thereto.

[0145] Figure 5 A schematic diagram of a message processing device is shown. The device is deployed at the message receiving end and may include: a message sending task module, a message control module, a message sending module, a message sorting module, and an incoming message task module.

[0146] The message sending task module can obtain messages to be distributed from the financial institution's business system (a message sending end) and store them in a preset format as message text for sending. In addition, the message sending task module can also process messages to be printed. Preferably, but not limited to, a communication connection is established with the financial institution's business system via a message queue (MQ) to achieve message exchange. Specifically, a communication connection can be established between the local message queue (MQ) corresponding to the financial institution's business system and the local queue manager in the message distribution end's local MQ, as well as between the local queue manager in the message distribution end's local MQ and the target queue manager in the target MQ corresponding to the SWIFT terminal system. Based on the MQ configuration information, the module obtains SWIFT messages to be received from the receive queue of the specified local queue manager.

[0147] The message control module can standardize messages uploaded from financial institution business systems according to the requirements of SWIFT terminals (such as PCC) (e.g., message length is 512 bytes, and the file extension is ABI). It also preprocesses the messages uploaded from financial institution business systems to generate training data for training message sorting models. The message control module can monitor each stage of message processing for querying and back up message files for critical stages. Furthermore, the message control module can split files sent from PCC, perform business requirement conversion on incoming messages from PCC, and send the converted messages to the financial institution business systems, recording received message acknowledgments (ACK, NAK, UNK).

[0148] The message sending module can package multiple messages into a single file and send it to the PCC. In one optional embodiment, two encryption technologies can be combined to achieve encrypted data transmission: RSA algorithm is used for data signing and authentication, encrypting individual messages to ensure their security; while DES algorithm is used for data transmission between PCCs, encrypting the packaged file containing multiple messages uploaded by various financial institution business systems to ensure the security of the entire file package. This achieves good results in both encryption speed and system security. Additionally, it can delete generated but not yet sent messages; query and count sent and deleted messages; and perform an UNRELEASE operation on messages (an operation that allows manual intervention to intercept messages before they are sent).

[0149] Incoming Task Module: Deploys a watch (monitoring) mechanism to enable file interaction on the agreed storage service and obtain incoming files from the PCC system. In an optional embodiment, two encryption technologies can be combined to achieve encrypted data transmission. The RSA algorithm is used for data signing and authentication, while the DES algorithm is used to encrypt the transmitted data, thus achieving good results in both encryption speed and system security.

[0150] The message sorting module: It splits incoming message files from the PCC system one by one, removes interference information from the messages according to predefined rules, completes preprocessing, and then uses a sorting model to match and determine the target message receiver for each message. After matching, messages belonging to various branches are sent to the financial institution's business system via MQ. In addition, it can re-sort incorrectly assigned messages.

[0151] This message processing equipment can improve the automation level of message processing, reduce risks, increase efficiency, enhance processing capacity, and improve sorting accuracy.

[0152] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0153] Based on the same inventive concept, this application also provides a message distribution apparatus for implementing the message distribution method described above. The solution provided by this apparatus is similar to the implementation described in the above method; therefore, the specific limitations in one or more message distribution apparatus embodiments provided below can be found in the limitations of the message distribution method described above, and will not be repeated here.

[0154] In one embodiment, such as Figure 6 As shown, a message distribution device is provided, comprising:

[0155] The message receiving module 601 is used to receive messages to be distributed sent by the message sending end to the message distributing end;

[0156] The interference message information removal module 602 is used to acquire multiple message information from the message to be distributed, remove interference message information from the multiple message information, and obtain target message information based on the remaining message information after removal; the interference message information is message information whose correlation with the message receiving end is less than a preset threshold.

[0157] The sorting module 603 is used to input the target message information into the trained message sorting model, and the message sorting model determines the target message receiving end corresponding to the message to be distributed based on the target message information.

[0158] The message forwarding module 604 is used to send the message to be distributed to the target message receiver.

[0159] In one embodiment, the interference message information removal module 602 is further configured to:

[0160] Obtain the service type of the message to be distributed and multiple service elements associated with the service type;

[0161] Identify the interfering service elements among the multiple service elements whose correlation with the message receiver is less than a preset threshold;

[0162] Based on the interference service elements, the interference message information among the multiple message information is determined.

[0163] In one embodiment, the interference message information removal module 602 is specifically used for:

[0164] Obtain a pre-stored configuration table; the configuration table records multiple candidate service elements, and the correlation between each candidate service element and the message receiver is less than a preset threshold.

[0165] Based on the matching results between the multiple business elements and each candidate business element in the configuration table, the interfering business elements among the multiple business elements are determined.

[0166] In one embodiment, the interference message information removal module 602 is further configured to:

[0167] Obtain multiple message blocks of the message to be distributed; the multiple message blocks correspond to different topics;

[0168] Interference message blocks are identified among the plurality of message blocks, and the message information in the interference message blocks is identified as interference message information; the correlation between the subject of the interference message block and the message receiving end is less than a preset threshold.

[0169] In one embodiment, the apparatus further includes:

[0170] The upload message acquisition module is used to receive messages uploaded by multiple message receiving terminals;

[0171] An encryption module is used to encrypt messages uploaded by the multiple message receiving terminals based on a preset agreed encryption key, to obtain an encrypted message set; the agreed encryption key is an encryption key pre-agreed between the message distribution terminal and the message sending terminal;

[0172] An encrypted message sending module is used to send the encrypted message set to the message sending end, so as to trigger the message sending end to return multiple messages to be distributed to the message distribution end according to the messages in the encrypted message set.

[0173] In one embodiment, the encryption module is used for:

[0174] Based on the encryption key corresponding to each message receiver, the message uploaded by each message receiver is encrypted to obtain multiple encrypted messages;

[0175] Based on a preset, agreed-upon encryption key, the multiple encrypted messages are encrypted to obtain a set of encrypted messages.

[0176] In one embodiment, the message sorting model is trained through the following steps:

[0177] For each message uploaded by the multiple message receiving terminals, the interfering message information in each message is removed to obtain multiple sample messages;

[0178] Based on the message receiver corresponding to each message, determine the receiver tag for each sample message;

[0179] Based on the multiple sample messages and the receiver label of each sample message, a text classification model is trained, and the trained message sorting model is obtained when the training termination condition is met.

[0180] Each module in the aforementioned message distribution device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in the processor of a computer device in hardware form or independent of it, or stored in the memory of the computer device in software form, so that the processor can call and execute the operations corresponding to each module.

[0181] In one embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 7 As shown, this computer device includes a processor, memory, input / output interfaces (I / O), and a communication interface. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface is also connected to the system bus via the I / O interfaces. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and a database. The internal memory provides the environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The database stores message data. The I / O interfaces are used for exchanging information between the processor and external devices. The communication interface is used for communicating with external terminals via a network connection. When the computer program is executed by the processor, it implements a message distribution method.

[0182] Those skilled in the art will understand that Figure 7 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0183] In one embodiment, a computer device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to perform the following steps:

[0184] Receive the message to be distributed sent by the message sender to the message distributor;

[0185] The system acquires multiple message information from the message to be distributed, removes interfering message information from the multiple message information, and obtains the target message information based on the remaining message information after removal; the interfering message information is message information whose correlation with the message receiving end is less than a preset threshold.

[0186] The target message information is input into the trained message sorting model, and the message sorting model determines the target message receiving end corresponding to the message to be distributed based on the target message information.

[0187] The message to be distributed is sent to the target message receiver.

[0188] In one embodiment, the processor also performs the steps described in the other embodiments when executing the computer program.

[0189] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, the computer program performing the following steps when executed by a processor:

[0190] Receive the message to be distributed sent by the message sender to the message distributor;

[0191] The system acquires multiple message information from the message to be distributed, removes interfering message information from the multiple message information, and obtains the target message information based on the remaining message information after removal; the interfering message information is message information whose correlation with the message receiving end is less than a preset threshold.

[0192] The target message information is input into the trained message sorting model, and the message sorting model determines the target message receiving end corresponding to the message to be distributed based on the target message information.

[0193] The message to be distributed is sent to the target message receiver.

[0194] In one embodiment, the computer program, when executed by a processor, also implements the steps described in the other embodiments above.

[0195] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, performs the following steps:

[0196] Receive the message to be distributed sent by the message sender to the message distributor;

[0197] The system acquires multiple message information from the message to be distributed, removes interfering message information from the multiple message information, and obtains the target message information based on the remaining message information after removal; the interfering message information is message information whose correlation with the message receiving end is less than a preset threshold.

[0198] The target message information is input into the trained message sorting model, and the message sorting model determines the target message receiving end corresponding to the message to be distributed based on the target message information.

[0199] The message to be distributed is sent to the target message receiver.

[0200] In one embodiment, the computer program, when executed by a processor, also implements the steps described in the other embodiments above.

[0201] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data shall comply with the relevant laws, regulations and standards of the relevant countries and regions.

[0202] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.

[0203] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0204] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A message distribution method, characterized in that, The method includes: Receive the message to be distributed sent by the message sender to the message distributor; The system acquires multiple message information from the message to be distributed, removes interfering message information from the multiple message information, and obtains the target message information based on the remaining message information after removal; the interfering message information is message information whose correlation with the message receiving end is less than a preset threshold. The target message information is input into the trained message sorting model, and the message sorting model determines the target message receiving end corresponding to the message to be distributed based on the target message information. The message to be distributed is sent to the target message receiver; Before removing interfering message information from the plurality of message information, the method further includes: Obtain the service type of the message to be distributed and multiple service elements associated with the service type; Obtain a pre-stored configuration table; the configuration table records multiple candidate service elements, and the correlation between each candidate service element and the message receiver is less than a preset threshold. Based on the matching results between the multiple business elements and each candidate business element in the configuration table, the interfering business elements among the multiple business elements are determined. Based on the interference service elements, the interference message information among the multiple message information is determined.

2. The method according to claim 1, characterized in that, Before removing interfering message information from the plurality of message information, the method further includes: Obtain multiple message blocks of the message to be distributed; the multiple message blocks correspond to different topics; Interference message blocks are identified among the plurality of message blocks, and the message information in the interference message blocks is identified as interference message information; the correlation between the subject of the interference message block and the message receiving end is less than a preset threshold.

3. The method according to any one of claims 1-2, characterized in that, Before the message to be distributed is sent from the message receiving end to the message distribution end, the following is also included: Receive messages uploaded by multiple message receivers; Based on a preset, agreed-upon encryption key, the messages uploaded by the multiple message receiving terminals are encrypted to obtain an encrypted message set; the agreed-upon encryption key is an encryption key pre-agreed upon by the message distribution terminal and the message sending terminal. The encrypted message set is sent to the message sending end to trigger the message sending end to return multiple messages to be distributed to the message distribution end based on the messages in the encrypted message set.

4. The method according to claim 3, characterized in that, The method involves encrypting the messages uploaded by the multiple message receiving terminals based on a preset, agreed-upon encryption key, resulting in a set of encrypted messages, including: Based on the encryption key corresponding to each message receiver, the message uploaded by each message receiver is encrypted to obtain multiple encrypted messages; Based on a preset, agreed-upon encryption key, the multiple encrypted messages are encrypted to obtain a set of encrypted messages.

5. The method according to claim 3, characterized in that, The message sorting model is trained through the following steps: For each message uploaded by the multiple message receiving terminals, the interfering message information in each message is removed to obtain multiple sample messages; Based on the message receiver corresponding to each message, determine the receiver tag for each sample message; Based on the multiple sample messages and the receiver label of each sample message, a text classification model is trained, and the trained message sorting model is obtained when the training termination condition is met.

6. A message distribution device, characterized in that, The device includes: The message receiving module is used to receive messages to be distributed sent by the message sending end to the message distribution end; The interference message information removal module is used to acquire multiple message information from the message to be distributed, remove interference message information from the multiple message information, and obtain target message information based on the remaining message information after removal; the interference message information is message information whose correlation with the message receiving end is less than a preset threshold. The sorting module is used to input the target message information into the trained message sorting model, and the message sorting model determines the target message receiving end corresponding to the message to be distributed based on the target message information. The message forwarding module is used to send the message to be distributed to the target message receiver; The interference message information removal module is further configured to obtain the service type of the message to be distributed and multiple service elements associated with the service type before removing the interference message information from the multiple message information; Obtain a pre-stored configuration table; the configuration table records multiple candidate service elements, and the correlation between each candidate service element and the message receiver is less than a preset threshold. Based on the matching results between the multiple business elements and each candidate business element in the configuration table, the interfering business elements among the multiple business elements are determined. Based on the interference service elements, the interference message information among the multiple message information is determined.

7. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 5.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 5.

9. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 5.

Images