Message reply method, apparatus and device
By extracting keywords and performing semantic analysis on messages, and using a pre-defined model to process messages to obtain multiple keywords and semantic fields, the problem of low message response accuracy in existing technologies is solved, and higher response accuracy is achieved.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- MINSHENG BANKING CORP
- Filing Date
- 2023-04-25
- Publication Date
- 2026-07-21
AI Technical Summary
In existing technologies, message reply methods based on keyword matching suffer from low accuracy, especially due to inaccurate replies caused by semantic mismatch.
By extracting keywords and performing semantic analysis on messages, a pre-defined model is used to process the messages to obtain multiple keyword fields and semantic fields. The reply message is then determined by combining the keyword and semantic fields.
It improves the accuracy of message replies and avoids inaccurate replies caused by the same keywords but different meanings.
Smart Images
Figure CN116340496B_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of artificial intelligence technology, and in particular to a message reply method, apparatus, and device. Background Technology
[0002] In various business transaction scenarios (such as purchasing bonds, purchasing goods, etc.), artificial intelligence technology (such as customer service robots) can be used to automatically answer questions raised by users during business transactions, thereby reducing human input and improving work efficiency.
[0003] In related technologies, message replies can be performed as follows: After receiving a user's question message, the AI device can perform text extraction processing to obtain at least one keyword corresponding to the question message. Based on the at least one keyword corresponding to the question message, the device identifies the question with the highest similarity to the question message in the database. Then, it determines the corresponding reply and sends the reply to the client.
[0004] In the above process, the question text with the highest similarity to the question message is determined only by using at least one keyword corresponding to the question message. It's possible that the question text with the highest similarity to the question message may have a significant semantic difference from the question message. Therefore, the corresponding response text determined based on the question text may not match the question message, resulting in low accuracy of the message response. Summary of the Invention
[0005] This application provides a message reply method, apparatus, and device to solve the problem of low accuracy in message replies.
[0006] In a first aspect, embodiments of this application provide a message reply method, including:
[0007] Obtain a first message, the first message including object information of at least one first object;
[0008] The first message is filtered and keyword extraction is performed to obtain multiple keyword fields corresponding to the first message;
[0009] The first message is processed by a preset model to obtain multiple semantic fields corresponding to the first message;
[0010] Based on the multiple keyword fields and the multiple semantic fields, a reply message corresponding to the first message is determined, and the reply message is sent to a preset device.
[0011] Secondly, embodiments of this application provide a message reply device, the device comprising:
[0012] The acquisition module is used to acquire a first message, wherein the first message includes object information of at least one first object;
[0013] The first processing module is used to perform keyword extraction processing on the first message to obtain multiple keyword fields corresponding to the first message;
[0014] The second processing module is used to process the first message through a preset model to obtain multiple semantic fields corresponding to the first message;
[0015] The determining module is used to determine the reply message corresponding to the first message based on the plurality of keyword fields and the plurality of semantic fields, and to send the reply message to a preset device.
[0016] Thirdly, embodiments of this application provide a message reply device, including:
[0017] At least one processor; and
[0018] A memory communicatively connected to the at least one processor; wherein,
[0019] The memory stores instructions executable by the at least one processor, which, when executed by the at least one processor, enables the at least one processor to perform the method described in any of the first aspects.
[0020] Fourthly, embodiments of this application provide a non-transitory computer-readable storage medium storing computer instructions, wherein the computer instructions are used to cause the computer to perform the method described in any one of the first aspects.
[0021] Fifthly, embodiments of this application provide a computer program product, including a computer program that, when executed by a processor, implements the method described in any one of the first aspects.
[0022] The message reply method, apparatus, and device provided in this application embodiment acquire a first message sent by a user. Keyword extraction processing is performed on the first message to obtain multiple keyword fields corresponding to the first message; the first message is processed using a preset model to obtain multiple semantic fields corresponding to the first message; based on the multiple keyword fields and multiple semantic fields, a reply message corresponding to the first message is determined and sent to a preset device. In the above process, the fields of multiple keywords and semantic fields corresponding to the first message can be obtained through extraction processing and a preset model. The reply message corresponding to the first message is determined based on multiple keyword fields and multiple semantic fields, rather than determining the reply based on at least one keyword corresponding to the first message. This avoids inaccurate replies due to the same keywords but different semantics, thus improving the accuracy of message replies. Attached Figure Description
[0023] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.
[0024] Figure 1 A schematic diagram illustrating the application scenarios provided in the embodiments of this application;
[0025] Figure 2 A flowchart illustrating a message reply method provided in an embodiment of this application;
[0026] Figure 3 This is a schematic diagram illustrating the process of obtaining the first message provided in an embodiment of this application;
[0027] Figure 4 A flowchart illustrating another message reply method provided in an embodiment of this application;
[0028] Figure 5 This is a schematic diagram illustrating the message reply process provided in an embodiment of this application;
[0029] Figure 6 This is a schematic diagram of the structure of a message reply device provided in an embodiment of this application;
[0030] Figure 7 This is a schematic diagram of another message reply device provided in an embodiment of this application;
[0031] Figure 8 This is a schematic diagram of the structure of a message reply device provided in an embodiment of this application.
[0032] The accompanying drawings illustrate specific embodiments of this application, which will be described in more detail below. These drawings and descriptions are not intended to limit the scope of the concept in any way, but rather to illustrate the concept of this application to those skilled in the art through reference to particular embodiments. Detailed Implementation
[0033] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this application as detailed in the appended claims.
[0034] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.
[0035] 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. Furthermore, the collection, use and processing of the relevant data must comply with relevant laws, regulations and standards, and corresponding operation entry points are provided for users to choose to authorize or refuse.
[0036] It should be noted that the voice processing method and apparatus of this application can be used in the field of artificial intelligence, or in any field other than artificial intelligence. The application field of the message reply method and apparatus of this application is not limited.
[0037] To facilitate understanding, the following will be combined with... Figure 1 The application scenarios applicable to the embodiments of this application will be described.
[0038] Figure 1 This is a schematic diagram illustrating an application scenario provided in an embodiment of this application. Please refer to [link / reference]. Figure 1 The system includes a terminal device 101 and a message reply device 102. The terminal device 101 can be a mobile phone, tablet, computer, etc., and has an application program installed. The message reply device 102 can be a server and has a database installed. Users can send messages to the message reply device 102 through the application program on the terminal device 101. After receiving a message from the terminal device 101, the message reply device 102 determines the corresponding reply through the database and sends the reply to the terminal device 101.
[0039] In related technologies, message replies can be performed as follows: After receiving a question message sent by a user, the message reply device can perform text extraction processing on the question message to obtain at least one keyword corresponding to the question message. Based on the at least one keyword corresponding to the question message, the question with the highest similarity to the question message is identified in the database. The corresponding reply is then determined and sent to the client. In the above process, only the question text with the highest similarity to the question message is identified using at least one keyword corresponding to the question message. It is possible that the question text with the highest similarity to the question message may have a significant semantic difference from the question message; therefore, the reply text determined based on the question text may not be the reply to the question message, resulting in low accuracy of message replies.
[0040] In this embodiment, a first message sent by a user is obtained. Keyword extraction processing is performed on the first message to obtain multiple keyword fields corresponding to the first message. The first message is then processed using a preset model to obtain multiple semantic fields corresponding to the first message. Based on the multiple keyword fields and multiple semantic fields, a reply message corresponding to the first message is determined and sent to a preset device. In the above process, the fields of multiple keywords and semantic fields corresponding to the first message are obtained through extraction processing and a preset model. The reply message corresponding to the first message is determined based on multiple keyword fields and multiple semantic fields, rather than determining the reply based on at least one keyword corresponding to the first message. This avoids inaccurate replies due to identical keywords but different semantics, thus improving the accuracy of message replies.
[0041] The method described in this application will now be illustrated through specific embodiments. It should be noted that the following embodiments may exist independently or in combination with each other; identical or similar content will not be repeated in different embodiments.
[0042] Figure 2 This is a flowchart illustrating a message reply method provided in an embodiment of this application. Please refer to... Figure 2 The method may include:
[0043] S201, Get the first message.
[0044] The execution entity in this application embodiment can be a message reply device or a message reply apparatus installed within a message reply device. The message reply apparatus can be implemented through software or a combination of software and hardware. The message reply device can be a server.
[0045] The first message includes object information for at least one first object.
[0046] Object information can include at least the object identifier, the corresponding data amount, and the interest rate.
[0047] For example, the first message can be shown in Table 1:
[0048] Table 1
[0049] Object identifier Data Amount interest rate SCP001 20,000 yuan 4.45％
[0050] Users can enter messages through the dialog box provided by the terminal device. The terminal device responds to the user's input and click operation by sending the message to the voice processing device.
[0051] Below, in conjunction with Figure 3 The process of obtaining the first message is explained. Figure 3 This is a schematic diagram illustrating the process of obtaining the first message as provided in an embodiment of this application. Please refer to... Figure 3 This includes interfaces 301 and 302. Interfaces 301 and 302 can be chat pages provided by the application on the terminal device. Referring to interface 301, the user inputs the first message shown in Table 1 above via the keyboard in the chat page provided by the application on the terminal device. After inputting the message, the user clicks the send button, and the terminal device responds to the user's input click by sending the first message to the message reply device. Referring to interface 302, the first message is displayed in the chat page provided by the application on the terminal device, indicating that the first message was sent successfully.
[0052] S202. Perform filtering and keyword extraction on the first message to obtain multiple keyword fields corresponding to the first message.
[0053] The first message can be filtered through keyword search and sensitive word detection to obtain the first message whose content is related to business transactions.
[0054] First messages unrelated to business transactions can be filtered out through keyword search. First messages including spam or useless information can be filtered out through sensitive word detection. First messages can also be filtered using multiple sensitive words stored in a sensitive word database.
[0055] The sensitive word library can be pre-configured according to business scenarios and stored in the preset storage space of the message reply device.
[0056] For example, suppose the first message is "I ate rice today". Before performing keyword extraction on the first message, keyword search determines that the first message is irrelevant to the business transaction. In this case, the first message is filtered out, and keyword extraction is no longer performed on it.
[0057] Regular expressions can be used to extract keywords from the first message, resulting in multiple keyword fields corresponding to the first message. These keyword fields can be in JSON format.
[0058] S203. Process the first message using a preset model to obtain multiple semantic fields corresponding to the first message.
[0059] The preset models can be named entity recognition models. The preset models include a pre-trained model, a first model, a second model, and a third model.
[0060] Multiple semantic fields corresponding to the first message can be obtained in the following ways: the first message is encoded by a pre-trained model to obtain the first code corresponding to the first message; the first code is processed by a first model to obtain at least one semantic feature corresponding to the first code, as well as the relationship between each semantic feature; the at least one semantic feature is classified by a second model to obtain the part-of-speech category to which the at least one semantic feature belongs; and the at least one semantic feature is processed by a third model to obtain at least one semantic field.
[0061] The pre-trained model can be a BERT-based Chinese model, the first model can be a BERT model, the second model can be a BiGRU model, and the third model can be a conditional random field model.
[0062] The preset model can extract and identify all keywords with the same or similar semantics, and obtain the semantic fields corresponding to all keywords with the same or similar semantics.
[0063] For example, by processing the first message shown in Table 1 above using a preset model, we obtain multiple semantic fields corresponding to the first message, namely SCP001; JE2W; L4.45%.
[0064] S204. Based on multiple keyword fields and multiple semantic fields, determine the reply message corresponding to the first message, and send the reply message to the preset device.
[0065] The response message corresponding to the first message can be determined based on multiple keyword fields and multiple semantic fields as follows: Identify at least one keyword field corresponding to each first object from among the multiple keyword fields, and at least one semantic field corresponding to each first object from among the multiple semantic fields; merge the at least one keyword field and at least one semantic field corresponding to each first object to obtain the target field corresponding to each first object; obtain the current state of each first object, including untraded state, in-transaction state, or transaction completed state; and determine the response message corresponding to the first message based on the target field corresponding to each first object and the current state of each first object.
[0066] When merging at least one keyword field and at least one semantic field corresponding to each first object, for keyword fields and semantic fields with the same semantics, only the semantic field corresponding to the same semantics is retained. All other different keyword fields and semantic fields are determined as the target fields.
[0067] For example, based on the first message shown in Table 1 above, the first message is filtered and keywords are extracted to obtain three keyword fields corresponding to the first message. Among these three keyword fields, the three keyword fields corresponding to the first object are determined to be: object identifier SCP001; amount 2W; interest rate 4.45%. Based on the example above, the multiple semantic fields corresponding to the first message are SCP001; JE2W; L4.45%. Among these semantic fields, the semantic fields corresponding to the first object are determined to be SCP001; JE2W; L4.45%. Since at least one keyword field corresponding to the first object has a semantic field with the same semantic meaning, when merging at least one keyword field and at least one semantic field corresponding to the first object, the target fields include SCP001; JE2W; L4.45%. The current state of the first object is obtained. If the current state of the first object is untraded, the reply message corresponding to the first message can be determined based on the target fields corresponding to the first object and the current state of each first object. The reply message is used to indicate whether the transaction was successful and can include the object's online information.
[0068] The message reply method provided in this application embodiment obtains a first message. Keyword extraction processing is performed on the first message to obtain multiple keyword fields corresponding to the first message. The first message is then processed using a preset model to obtain multiple semantic fields corresponding to the first message. Based on the multiple keyword fields and multiple semantic fields, a reply message corresponding to the first message is determined and sent to a preset device. In the above process, the fields of multiple keywords and semantic fields corresponding to the first message are obtained through extraction processing and a preset model. The reply message corresponding to the first message is determined based on multiple keyword fields and multiple semantic fields, rather than determining the reply based on at least one keyword corresponding to the first message. This avoids inaccurate replies due to identical keywords but different semantics, thus improving the accuracy of message replies.
[0069] Based on any of the above embodiments, the following, in conjunction with Figure 4 The detailed process of replying to messages is explained.
[0070] Figure 4 This is a flowchart illustrating another message reply method provided in an embodiment of this application. Please refer to... Figure 4 The method includes:
[0071] S401, Get the first message.
[0072] Before receiving first information, businesses need to authenticate users so that they can conduct business transactions with the business.
[0073] User identity can be verified in the following way: receive an authentication request sent by a preset device, the authentication request including a user identifier; perform verification processing on the authentication request according to the user identifier, determine the verification result corresponding to the authentication request, the verification result is verification passed or verification failed; if the verification result is verification passed, send object information of multiple objects to the preset device.
[0074] The default devices can be mobile phones, tablets, computers, etc. Users can send authentication requests to the terminal devices used by enterprise staff through the application provided by the default device. The enterprise's instant messaging (Instant Messaging) listener can receive the authentication request and send it to the data collector. The data collector then sends the authentication request to the enterprise's business system, enabling the business system to authenticate the user based on the authentication request.
[0075] The object information should include at least the total amount, interest rate, issuance period, and object identifier of the object.
[0076] When the verification result is successful, the business system determines that the user can conduct business transactions. At this time, the business system can periodically send object information for multiple objects to the user's preset devices, based on the settings of the enterprise staff. When the object information is updated, the updated object information can also be sent to the user's preset devices in real time, so that the user can be informed of the object information in a timely manner and conduct corresponding business transactions based on the object information.
[0077] The method for sending object information of multiple objects to a preset device can be preset and stored in the preset storage space of the message reply device, so that the message reply device sends the object information of multiple objects to the preset device according to the preset sending method.
[0078] S402. Perform filtering and keyword extraction on the first message to obtain multiple keyword fields corresponding to the first message.
[0079] It should be noted that the execution process of S402 can be found in S202, and will not be repeated here.
[0080] S403. Process the first message using a preset model to obtain multiple semantic fields corresponding to the first message.
[0081] Before processing the first message using the preset model, the preset model can be trained.
[0082] The preset model can be trained as follows: Obtain a training set, which includes multiple second messages and multiple semantic fields corresponding to each second message; divide the training set to obtain a first training set and a second training set; obtain an initial model; train the i-th intermediate model through the second messages in the first training set for the i-th iteration to obtain the (i+1)-th intermediate model, where i takes the values 1, 2, 3, ..., until i is greater than or equal to N, and then determine the i-th intermediate model as the preset model, where N is the preset number of iterations and is an integer greater than 1. The first intermediate model is the initial model.
[0083] The (i+1)th intermediate model can be obtained as follows: process each second message using the i-th intermediate model to obtain at least one predicted semantic field corresponding to each second message; determine the loss function based on the at least one predicted semantic field corresponding to each second message and at least one semantic feature corresponding to it in the first training set; update the model parameters of the i-th intermediate model based on the loss function to obtain the (i+1)-th intermediate model. The model parameters include the area of a single training block, the learning rate, and the proportion of ignored neurons.
[0084] The loss function may include mean squared error (MSE) and mean absolute error (MAE). The coefficients of MSE and MAE can be adjusted during training, and this application does not impose any restrictions on this.
[0085] After training is complete, the preset model can be validated using a second training set. If the loss value obtained during training is less than or equal to the preset threshold, the preset model can be used directly. If the loss value obtained during training is greater than the preset threshold, the model parameters can be adjusted and the preset model updated according to the above method to make the loss value less than or equal to the preset threshold.
[0086] S404. Determine at least one keyword field corresponding to each first object among multiple keyword fields, and determine at least one semantic field corresponding to each first object among multiple semantic fields.
[0087] For example, the first message can be shown in Table 2:
[0088] Table 2
[0089] First object Object identifier Data Amount interest rate Object 1 SCP001 40,000 yuan 4.45％ Object 2 SCP003 10,000 yuan 6.5％
[0090] The first message shown in Table 2 is then filtered and its keywords extracted, resulting in multiple keyword fields corresponding to the first message: object identifier SCP001; amount 4W; interest rate 4.45%; SCP003; amount 1W; interest rate 6.5%. The first message in Table 2 is processed using a preset model, resulting in multiple semantic fields corresponding to the first message: SCP001; JE4W; L4.45%; SCP003; JE1W; L6.5%. Among these keyword fields, at least one keyword field corresponding to object 1 is determined to be SCP001; amount 4W; interest rate 4.45%. Similarly, among these semantic fields, at least one semantic field corresponding to object 2 is determined to be SCP003; amount 1W; interest rate 6.5%. Finally, among these semantic fields, at least one semantic field corresponding to object 2 is determined to be SCP003; JE1W; L6.5%.
[0091] S405. Merge at least one keyword field and at least one semantic field corresponding to each first object to obtain the target field corresponding to each first object.
[0092] By merging at least one keyword field and at least one semantic field corresponding to each first object, we can avoid the situation where important information is missed in the fields obtained by processing with a single method, and ensure that each keyword in the first message has a corresponding field.
[0093] After determining the target field, the target field and the identifier of the preset device that sent the target field can be stored in the preset storage space of the message reply device.
[0094] For example, based on at least one keyword field and at least one semantic field corresponding to object 1 as shown in the example above, and at least one keyword field and at least one semantic field corresponding to object 2, the target fields corresponding to each first object can be specifically determined as shown in Table 3:
[0095] Table 3
[0096] First object target field Object 1 SCP001; JE4W; L4.45% Object 2 SCP003; JE1W; L6.5%
[0097] S406. Get the current state of each first object.
[0098] The current state of the first object can be retrieved from the database of the message reply device. When the current state of the first object changes, the current state of the first object stored in the database needs to be updated.
[0099] The current status includes untraded, in-trade, or trade completed.
[0100] For example, based on the first object shown in Table 3 above, the current state of each first object can be obtained as shown in Table 4:
[0101] Table 4
[0102] First object Current status Object 1 Transaction completed status Object 2 Untraded status
[0103] S407. For any first object, determine whether the current state of the first object is a transaction completed state.
[0104] If so, execute S408.
[0105] If not, execute S412.
[0106] S408. Determine that the first reply message of the first object indicates that the status of the first object is the transaction completion status.
[0107] When the status of the first object is determined to be "transaction completed," transaction information for the first object can be generated based on multiple semantic fields from historical time periods, enabling internal staff to compile and analyze the data. The transaction information for the first object includes at least multiple user identifiers, as well as the target amount and interest rate for each user.
[0108] For example, based on the current status of each first object shown in Table 4 above, the status of object 1 can be determined to be a transaction completed state. Therefore, the first reply message of object 1 can be determined as indicating that object 1's status is a transaction completed state. The user can determine from the first reply message of object 1 that object 1 is currently unable to proceed with the transaction.
[0109] S409. Obtain the associated object of the first object.
[0110] The associated object is an object that has a relationship or similarity with the first object, and the current state of the associated object is either untraded or in the process of being traded.
[0111] Multiple relationships between objects can be pre-defined and stored in the message reply device's preset storage space. When the status of the first object is determined to be "transaction completed," related objects of the first object whose transaction is not yet completed can be recommended to the user.
[0112] For example, based on the example of object 1 above, the status of object 1 is determined to be "transaction completed". At this point, the associated objects of object 1 can be retrieved. The specific object information of the associated objects is shown in Table 5:
[0113] Table 5
[0114] Related objects Object identifier Data Amount interest rate Object 3 SCP002 1 million yuan 5.5％ Object 5 SCP005 2.2 million yuan 3.0％
[0115] S410. Based on the target amount, determine the recommended transaction amount for the related parties.
[0116] Based on the target amount, a range of suggested transaction amounts for related parties can be determined. Based on the related parties' information, the suggested transaction amount for each related party is determined within this range.
[0117] For example, based on the related objects of Object 1 shown in Table 5 above, the suggested transaction amount for the related objects can be determined to be between 30,000 and 60,000 yuan, depending on the target amount. The suggested transaction amount for Object 3 can be 50,000 yuan, and the suggested transaction amount for Object 5 can be 30,000 yuan.
[0118] S411. Generate a prompt message based on the identifier of the associated object and the suggested transaction amount of the associated object, and send the prompt message to the preset device.
[0119] The prompt message is used to indicate whether to recommend a similar object to the first object for business transactions.
[0120] For example, based on the above example, we can determine that the associated objects of object 1 are object 3 and object 5. Based on the identifiers of the associated objects and the suggested transaction amounts for those objects, the generated prompt information can be shown in Table 6:
[0121] Table 6
[0122] Related objects Object identifier Suggested transaction amount interest rate Object 3 SCP002 50,000 yuan 5.5％ Object 5 SCP005 30,000 yuan 3.0％
[0123] Different message templates can be set according to different business scenarios. After matching the generated prompt information with the corresponding template, it is sent to the preset device.
[0124] For example, after matching the corresponding template according to the prompts shown in Table 6, the message text sent to the preset device could be: "Dear User A, Hello! Since Object 1 is currently in the 'transaction completed' state, the transaction cannot be completed at this time. We recommend Object 3 and Object 5, which are similar to Object 1. Specific information: SCP002 Object 3, suggested transaction amount 50,000 yuan, interest rate 5.5%; SCP005 Object 5, suggested transaction amount 30,000 yuan, interest rate 3.0%."
[0125] After S411, execute S417.
[0126] S412. Obtain multiple semantic fields corresponding to the first object within the preset time period.
[0127] If the first object is in an untraded or transacting state, the transaction information of the first object within the historical period can be determined based on multiple semantic fields corresponding to the first object within the historical period.
[0128] For example, suppose the preset time period is from 10:00 AM on April 19, 2023 to the current time. Based on the current status of each first object shown in Table 4 above, the status of object 2 can be determined to be a transaction completed state. The specific steps for obtaining the multiple semantic fields corresponding to the first object within the preset time period are shown in Table 7:
[0129] Table 7
[0130]
[0131]
[0132] S413. Determine the data amount corresponding to each semantic field and the target amount corresponding to the target field.
[0133] For example, based on the multiple semantic fields corresponding to the first object shown in Table 7 above, the specific data amount corresponding to each semantic field can be determined as shown in Table 8:
[0134] Table 8
[0135] user Data Amount User 1 100,000 yuan User 2 50,000 yuan User 4 80,000 yuan User 6 30,000 yuan User 7 60,000 yuan
[0136] Based on the target field of object 2 shown in Table 3 above, the target amount corresponding to the target field can be determined to be 10,000 yuan.
[0137] S414. Determine the first data amount corresponding to the first object based on multiple data amounts and the target amount.
[0138] The sum of multiple data amounts and the target amount can be determined as the first data amount corresponding to the first object.
[0139] For example, based on the multiple data amounts and target amounts of object 2 shown in the example above, the first data amount corresponding to object 2 is determined to be 10+5+2+8+3+6=340,000 yuan.
[0140] S415. Get the total amount corresponding to the first object.
[0141] The total amount of the first object can be stored in the preset storage space of the message reply device.
[0142] For example, the total amount corresponding to object 2 is 1 million yuan, which is retrieved from the preset storage space of the message reply device.
[0143] S416. Determine the first response message for the first object based on the first data amount and the total amount.
[0144] The first response message for the first object can be determined based on the first data amount and the total amount as follows: determine whether the first data amount is less than or equal to the total amount; if so, generate the first response message for the first object based on the identifier of the first object and the target amount, and the first response message is used to indicate that the business transaction of the first object is successful; if not, determine that the first response message for the first object indicates that the business transaction of the first object has failed.
[0145] When it is determined that the first object can conduct a business transaction, if the user conducting the transaction is the first user, a business information table corresponding to the first object is generated and stored in the preset storage space of the message reply device. The business information corresponding to the first user is stored in the business information table corresponding to the first object. If the user conducting the transaction is not the first user, the business information table of the first object is updated, that is, the business information corresponding to the current user is stored in the business information table of the first object.
[0146] For example, according to the example above, the first data amount corresponding to object 2 is 340,000 yuan, and the total amount corresponding to object 2 is 1,000,000 yuan. It can be determined that the first data amount is less than the total amount. Therefore, based on the identifier of the first object and the target amount, a first reply message is generated for the first object. The first reply message is used to indicate that the business transaction of the first object was successful. The specific details of the first reply message are shown in Table 9:
[0147] Table 9
[0148] First object Object identifier Target Amount Object 2 SCP003 10,000 yuan
[0149] S417. Determine the reply message corresponding to the first message based on the first reply message corresponding to each first object.
[0150] The first reply message corresponding to each first object can be matched with the corresponding message reply template to generate the reply message corresponding to the first message.
[0151] For example, based on the first reply message of object 1 and the first reply message of object 2 shown in the example above, the first reply message of object 1 is matched with template 1, and the first reply message of object 2 is matched with template 2 to generate the reply message corresponding to the first message.
[0152] S418. Send a reply message to the preset device.
[0153] The message reply device can send reply messages to preset devices through the collector.
[0154] For example, suppose terminal device A sends the first message shown in Table 2 above to the message reply device. Based on the first reply message of object 1 and the first reply message of object 2 shown in the example above, a reply message corresponding to the first message is generated and sent to terminal device A.
[0155] This application provides a message reply method according to an embodiment, which obtains a first message. The first message is filtered and keyword extracted to obtain multiple keyword fields corresponding to the first message. The first message is processed using a preset model to obtain multiple semantic fields corresponding to the first message. At least one keyword field corresponding to each first object is determined from the multiple keyword fields, and at least one semantic field corresponding to each first object is determined from the multiple semantic fields. The at least one keyword field and at least one semantic field corresponding to each first object are merged to obtain a target field corresponding to each first object. The current state of each first object is obtained. For any first object, a first reply message corresponding to the first object is determined based on the target field and the current state of the first object. A reply message corresponding to the first message is determined based on the first reply message corresponding to each first object. In the above process, multiple keyword fields and semantic fields corresponding to the first message can be obtained through extraction processing and a preset model. The reply message corresponding to the first message is determined based on multiple keyword fields and multiple semantic fields, rather than determining the reply message corresponding to the first message based on at least one keyword. This avoids inaccurate replies due to the same keyword but different semantics, thus improving the accuracy of message replies.
[0156] Based on any of the above embodiments, the following, in conjunction with Figure 5 The process of replying to messages will be illustrated with examples.
[0157] Figure 5 This is a schematic diagram illustrating the message reply process provided in an embodiment of this application. Please refer to [link / reference]. Figure 5 This includes a terminal device 501 and a message reply device 502. The terminal device 501 can be a mobile phone, tablet computer, computer, etc., and contains an application program. The message reply device 502 can be a server, and contains a database and a preset model.
[0158] When the message reply device 502 determines that the user's authentication result through the terminal device 501 is successful, it sends object information of multiple objects to the terminal device 501, as shown in Table 10:
[0159] Table 10
[0160] object Object identifier lump sum interest rate Object 1 SBP001 1.5 million yuan 5.5％ Object 2 SBP002 2 million yuan 3.0％ Object 3 SBP003 1 million yuan 4.5％ Object 4 SBP004 500,000 yuan 6.0％ Object 5 SBP005 3 million yuan 5.0％
[0161] After the user determines the object information of the multiple objects shown in Table 10 through the application of the terminal device 501, they can enter a first message in the chat page provided by the application of the terminal device 501. The first message can be shown in Table 11 as follows:
[0162] Table 11
[0163]
[0164]
[0165] In response to a user's input click, terminal device 501 sends the first message shown in Table 11 to message reply device 502. After receiving the first message from terminal device 501, message reply device 502 performs filtering and keyword extraction on the first message to obtain multiple keyword fields corresponding to the first message. It then processes the first message using a preset model to obtain multiple semantic fields corresponding to the first message. Message reply device 502 determines at least one keyword field corresponding to each first object from the multiple keyword fields, and at least one semantic field corresponding to each first object from the multiple semantic fields. Message reply device 502 merges the at least one keyword field and at least one semantic field corresponding to each first object to obtain the target field corresponding to each first object. The specific target fields corresponding to each first object are shown in Table 12.
[0166] Table 12
[0167] First object target field Object 3 SBP003; JE8W; L4.5% Object 4 SBP004; JE5W; L6.0%
[0168] The message reply device 502 retrieves the current state of each first object from the database, specifically as follows:
[0169] As shown in Table 13:
[0170] Table 13
[0171] First object Current status Object 3 Transaction completed Object 4 Untraded status
[0172] Message reply device 502 determines that the status of object 3 is "transaction completed" according to Table 13, and therefore determines that the first reply message for object 3 indicates that the status of object 3 is "transaction completed". At this time, message reply device 502 obtains the associated object of object 3 and determines the suggested transaction amount for the associated object based on the target amount. Message reply device 502 generates a prompt message based on the identifier of the associated object and the suggested transaction amount, as shown in Table 14.
[0173] Table 14
[0174] Related objects Object identifier Suggested transaction amount interest rate Object 6 SBP006 50,000 yuan 5.5％
[0175] After matching the prompt information shown in Table 14 with the corresponding template, the message reply device 502 generates a matching prompt and sends the matching prompt to the terminal device 501.
[0176] Message reply device 502 determines that the status of object 4 is "not traded" based on Table 13, and then retrieves multiple semantic fields corresponding to object 4 within the preset time period of 10:00 on 2023 / 04 / 19 to the current time. The specific semantic fields corresponding to object 4 are shown in Table 15.
[0177] Table 15
[0178] user target field User 1 SBP004; JE10W; L6.0% User 3 SBP004; JE10W; L6.0% User 4 SBP004; JE15W; L6.0% User 7 SBP004; JE5W; L6.0% User 8 SBP004; JE8W; L6.0%
[0179] The message reply device 502 determines the data amount corresponding to each semantic field based on the multiple semantic fields corresponding to object 4 shown in Table 15, as shown in Table 16:
[0180] Table 8
[0181] user Data Amount User 1 100,000 yuan User 3 100,000 yuan User 4 150,000 yuan User 7 50,000 yuan User 8 80,000 yuan
[0182] Based on the target field of object 4 shown in Table 12, message reply device 502 determines the target amount corresponding to the target field to be 50,000 yuan. Based on the multiple data amounts and the target amount of object 4, message reply device 502 determines the first data amount corresponding to object 4 to be 10 + 10 + 15 + 5 + 8 + 5 = 530,000 yuan. Message reply device 502 determines the total amount corresponding to object 4 in the database to be 500,000 yuan. Since message reply device 502 determines that the first data amount is greater than the total amount, it determines that the first reply message for the first object indicates that the business transaction of the first object has failed. Message reply device 502 matches the first reply message of object 4 with the corresponding template, generates the reply message corresponding to the first message, and sends the reply message to terminal device A.
[0183] This application provides a message reply method according to an embodiment, which obtains a first message. The first message is filtered and keyword extracted to obtain multiple keyword fields corresponding to the first message. The first message is processed using a preset model to obtain multiple semantic fields corresponding to the first message. At least one keyword field corresponding to each first object is determined from the multiple keyword fields, and at least one semantic field corresponding to each first object is determined from the multiple semantic fields. The at least one keyword field and at least one semantic field corresponding to each first object are merged to obtain a target field corresponding to each first object. The current state of each first object is obtained. For any first object, a first reply message corresponding to the first object is determined based on the target field and the current state of the first object. A reply message corresponding to the first message is determined based on the first reply message corresponding to each first object. In the above process, multiple keyword fields and semantic fields corresponding to the first message can be obtained through extraction processing and a preset model. The reply message corresponding to the first message is determined based on multiple keyword fields and multiple semantic fields, rather than determining the reply message corresponding to the first message based on at least one keyword. This avoids inaccurate replies due to the same keyword but different semantics, thus improving the accuracy of message replies.
[0184] Figure 6 This is a schematic diagram of a message reply device provided in an embodiment of this application. Please refer to... Figure 6 The message reply device 10 may include:
[0185] Acquisition module 11 is used to acquire a first message, the first message including object information of at least one first object;
[0186] The first processing module 12 is used to perform keyword extraction processing on the first message to obtain multiple keyword fields corresponding to the first message;
[0187] The second processing module 13 is used to process the first message through a preset model to obtain multiple semantic fields corresponding to the first message;
[0188] The determining module 14 is used to determine the reply message corresponding to the first message based on the plurality of keyword fields and the plurality of semantic fields, and to send the reply message to a preset device.
[0189] In one possible implementation, the determining module 14 is specifically used for:
[0190] Determine at least one keyword field corresponding to each first object from among the plurality of keyword fields, and determine at least one semantic field corresponding to each first object from among the plurality of semantic fields;
[0191] The at least one keyword field and the at least one semantic field corresponding to each first object are merged to obtain the target field corresponding to each first object;
[0192] Obtain the current state of each first object, including the current state of no transaction, transaction in progress, or transaction completed.
[0193] Based on the target field corresponding to each first object and the current state of each first object, determine the reply message corresponding to the first message.
[0194] In one possible implementation, the determining module 14 is specifically used for:
[0195] For any first object, determine the first reply message corresponding to the first object based on the target field corresponding to the first object and the current state of the first object;
[0196] Based on the first reply message corresponding to each first object, determine the reply message corresponding to the first message.
[0197] In one possible implementation, the determining module 14 is specifically used for:
[0198] If the current state of the first object is a transaction completed state, then the first reply message of the first object is determined to indicate that the state of the first object is the transaction completed state;
[0199] If the current state of the first object corresponding to the target field is the non-transactional state or the transaction state, then multiple semantic fields corresponding to the first object within a preset time period are obtained, and the first reply message of the first object is determined based on the multiple semantic fields and the target field.
[0200] In one possible implementation, the determining module 14 is specifically used for:
[0201] Determine the data amount corresponding to each semantic field and the target amount corresponding to the target field;
[0202] Based on multiple data amounts and the target amount, determine the first data amount corresponding to the first object;
[0203] Get the total amount corresponding to the first object;
[0204] The first response message for the first object is determined based on the first data amount and the total amount.
[0205] In one possible implementation, the determining module 14 is specifically used for:
[0206] Determine whether the first data amount is less than or equal to the total amount;
[0207] If so, a first reply message is generated for the first object based on the identifier of the first object and the target amount. The first reply message is used to indicate that the business transaction of the first object is successful.
[0208] If not, then the first response message from the first object is determined to indicate that the business transaction of the first object has failed.
[0209] In one possible implementation, the second processing module 13 is specifically used for:
[0210] The first message is encoded using the pre-trained model to obtain the first code corresponding to the first message;
[0211] The first encoding is processed by the first model to obtain at least one semantic feature corresponding to the first encoding, as well as the association between each semantic feature;
[0212] The at least one semantic feature is classified using the second model to obtain the part-of-speech category to which the at least one semantic feature belongs;
[0213] The at least one semantic feature is processed by the third model to obtain the at least one semantic field.
[0214] The message reply device provided in this application embodiment can execute the technical solution shown in the above method embodiment. Its implementation principle and beneficial effects are similar, and will not be described again here.
[0215] Figure 7 This is a schematic diagram of another message reply device provided in an embodiment of this application. Figure 6 Based on the illustrated embodiments, please refer to Figure 7The message reply device 10 also includes a generation module 15 and a sending module 16.
[0216] The generation module 15 is used for:
[0217] Obtain the associated object of the first object. The associated object is an object that has an association relationship or similar relationship with the first object. The current state of the associated object is either the untraded state or the transacting state.
[0218] Based on the target amount, determine the suggested transaction amount for the related parties;
[0219] Based on the identifier of the associated object and the suggested transaction amount of the associated object, a prompt message is generated and sent to the preset device. The prompt message is used to instruct the user to recommend an object similar to the first object for business transactions.
[0220] The sending module 16 is used for:
[0221] Receive an authentication request sent by the preset device, the authentication request including a user identifier;
[0222] The authentication request is verified based on the user identifier to determine the verification result corresponding to the authentication request, wherein the verification result is either successful or unsuccessful.
[0223] If the verification result is that the verification is successful, the object information of multiple objects is sent to the preset device.
[0224] The message reply device provided in this application embodiment can execute the technical solution shown in the above method embodiment. Its implementation principle and beneficial effects are similar, and will not be described again here.
[0225] Figure 8 This is a schematic diagram of the structure of a message reply device provided in an embodiment of this application. Please refer to... Figure 8 The message reply device 20 may include a memory 21 and a processor 22. Exemplarily, the memory 21 and the processor 22 are interconnected via a bus 23.
[0226] Memory 21 is used to store program instructions;
[0227] The processor 22 is used to execute the program instructions stored in the memory, so that the message reply device 20 performs the method shown in the above method embodiment.
[0228] The message reply device provided in this application embodiment can execute the technical solution shown in the above method embodiment. Its implementation principle and beneficial effects are similar, and will not be described again here.
[0229] This application provides a computer-readable storage medium storing computer-executable instructions, which are used to implement the above-described method when executed by a processor.
[0230] This application embodiment may also provide a computer program product, including a computer program that, when executed by a processor, can implement the above-described method.
[0231] All or part of the steps in the above-described method embodiments can be implemented by hardware related to program instructions. The aforementioned program can be stored in a readable memory. When the program is executed, it performs the steps of the above-described method embodiments; and the aforementioned memory (storage medium) includes: read-only memory (ROM), random access memory (RAM), flash memory, hard disk, solid-state drive, magnetic tape, floppy disk, optical disk, and any combination thereof.
[0232] This application describes embodiments with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It should be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processing unit of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processing unit of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.
[0233] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.
[0234] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.
[0235] Obviously, those skilled in the art can make various modifications and variations to the embodiments of this application without departing from the spirit and scope of this application. Therefore, if these modifications and variations to the embodiments of this application fall within the scope of the claims of this application and their equivalents, this application also intends to include these modifications and variations.
Claims
1. A message reply method, characterized in that, include: Obtain a first message, the first message including object information of at least one first object; The first message is filtered and keyword extraction is performed to obtain multiple keyword fields corresponding to the first message; The first message is processed by a preset model to obtain multiple semantic fields corresponding to the first message; Based on the multiple keyword fields and the multiple semantic fields, determine the reply message corresponding to the first message, and send the reply message to the preset device; The step of determining the reply message corresponding to the first message based on the plurality of keyword fields and the plurality of semantic fields includes: Determine at least one keyword field corresponding to each first object from among the plurality of keyword fields, and determine at least one semantic field corresponding to each first object from among the plurality of semantic fields; merge the at least one keyword field and the at least one semantic field corresponding to each first object to obtain the target field corresponding to each first object; obtain the current state of each first object, the current state including non-transaction state, transaction state, or transaction completed state; for any first object, determine the first reply message corresponding to the first object based on the target field corresponding to the first object and the current state of the first object; determine the reply message corresponding to the first message based on the first reply message corresponding to each first object.
2. The method according to claim 1, characterized in that, Based on the target field corresponding to the first object and the current state of the first object, the first reply message corresponding to the first object is determined, including: If the current state of the first object is a transaction completed state, then the first reply message of the first object is determined to indicate that the state of the first object is the transaction completed state; If the current state of the first object corresponding to the target field is the non-transactional state or the transaction state, then multiple semantic fields corresponding to the first object within a preset time period are obtained, and the first reply message of the first object is determined based on the multiple semantic fields and the target field.
3. The method according to claim 2, characterized in that, Based on the plurality of semantic fields and the target field, the first response message of the first object is determined, including: Determine the data amount corresponding to each semantic field and the target amount corresponding to the target field; Based on multiple data amounts and the target amount, determine the first data amount corresponding to the first object; Get the total amount corresponding to the first object; The first response message for the first object is determined based on the first data amount and the total amount.
4. The method according to claim 3, characterized in that, Based on the first data amount and the total amount, the first response message of the first object is determined, including: Determine whether the first data amount is less than or equal to the total amount; If so, a first reply message is generated for the first object based on the identifier of the first object and the target amount. The first reply message is used to indicate that the business transaction of the first object is successful. If not, then the first response message from the first object is determined to indicate that the business transaction of the first object has failed.
5. The method according to claim 4, characterized in that, After determining that the first response message from the first object indicates that the business transaction of the first object has failed, the method further includes: Obtain the associated object of the first object. The associated object is an object that has an association relationship or similar relationship with the first object. The current state of the associated object is either the untraded state or the transacting state. Based on the target amount, determine the suggested transaction amount for the related parties; Based on the identifier of the associated object and the suggested transaction amount of the associated object, a prompt message is generated and sent to the preset device. The prompt message is used to instruct the user to recommend an object similar to the first object for business transactions.
6. The method according to claim 1, characterized in that, The preset model includes a pre-trained model, a first model, a second model, and a third model; The first message is processed using a preset model to obtain at least one semantic field corresponding to the first message, including: The first message is encoded using the pre-trained model to obtain the first code corresponding to the first message; The first encoding is processed by the first model to obtain at least one semantic feature corresponding to the first encoding, as well as the association between each semantic feature; The at least one semantic feature is classified using the second model to obtain the part-of-speech category to which the at least one semantic feature belongs; The at least one semantic feature is processed by the third model to obtain the at least one semantic field.
7. The method according to any one of claims 1-6, characterized in that, Before receiving the first message, including: Receive an authentication request sent by the preset device, the authentication request including a user identifier; The authentication request is verified based on the user identifier to determine the verification result corresponding to the authentication request, wherein the verification result is either successful or unsuccessful. If the verification result is that the verification is successful, the object information of multiple objects is sent to the preset device.
8. A message reply device, characterized in that, The device includes: The acquisition module is used to acquire a first message, wherein the first message includes object information of at least one first object; The first processing module is used to perform keyword extraction processing on the first message to obtain multiple keyword fields corresponding to the first message; The second processing module is used to process the first message through a preset model to obtain multiple semantic fields corresponding to the first message; The determining module is used to determine the reply message corresponding to the first message based on the plurality of keyword fields and the plurality of semantic fields, and to send the reply message to a preset device; When determining the reply message corresponding to the first message based on the plurality of keyword fields and the plurality of semantic fields, the determining module is specifically configured to: determine at least one keyword field corresponding to each first object among the plurality of keyword fields, and determine at least one semantic field corresponding to each first object among the plurality of semantic fields; merge the at least one keyword field and the at least one semantic field corresponding to each first object to obtain the target field corresponding to each first object; obtain the current state of each first object, the current state including non-transaction state, transaction state, or transaction completed state; for any first object, determine the first reply message corresponding to the first object based on the target field corresponding to the first object and the current state of the first object; and determine the reply message corresponding to the first message based on the first reply message corresponding to each first object.
9. A message reply device, characterized in that, include: At least one processor; as well as A memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor to enable the at least one processor to perform the method of any one of claims 1 to 7.
10. A non-transitory computer-readable storage medium storing computer instructions, characterized in that, in, The computer instructions are used to cause the computer to perform the method according to any one of claims 1 to 7.
11. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the method of any one of claims 1 to 7.