Signaling data processing method, related system and storage medium

Abstract

The embodiment of the invention provides a signaling data processing method, a related system and a storage medium. The method comprises the following steps: analyzing each message in at least one message in signaling data to obtain at least one tree structure; converting each tree structure in the at least one tree structure by using a structure division identifier to obtain at least one serialized text, the serialized text comprising the structure division identifier; and performing semantic segmentation on each serialized text in the at least one serialized text to obtain a semantic text of a cell in the at least one message in the signaling data. In this way, complete semantic representation of the signaling data can be achieved.

Description

Technical Field

[0001] The present application relates to the field of data processing technology, and in particular to a signaling data processing method and related systems and storage media. Background Art

[0002] Signaling refers to the control information transmitted between devices to coordinate the operation of various devices (network elements) in the communication network. Signaling data records the operation details of the communication network in detail. Its content and format are only related to the protocol specifications and are not affected by the equipment manufacturer. Therefore, it has extremely high analytical value in the field of operation and maintenance.

[0003] The current signaling analysis method mainly relies on manual analysis by engineers. Since the amount of data collected by signaling tracking is huge, manual analysis requires experts to have a deep understanding of protocol knowledge, so the analysis cost is high and the efficiency is low. Therefore, building an analysis device based on machine learning is a very urgent business need.

[0004] At present, through expert experience or preprocessing, the information in the signaling is compressed to reduce the amount of data, or the data association method is simplified to reduce the complexity of analysis, and some simple signaling analysis is performed based on this. Such methods sacrifice the integrity and relevance of the data to a large extent, cannot cover a wide range of signaling analysis scenarios, and the applicable fields are also relatively limited. Summary of the invention

[0005] The present application discloses a signaling data processing method and a related system and storage medium, which can realize the complete semantic representation of signaling data.

[0006] In a first aspect, an embodiment of the present application provides a signaling data processing method, the method comprising:

[0007] Parsing each message in at least one message in the signaling data to obtain at least one tree structure;

[0008] Using the structure division identifier, transform each tree structure in the at least one tree structure to obtain at least one serialized text, wherein the serialized text includes the structure division identifier;

[0009] Semantic segmentation is performed on each serialized text in the at least one serialized text to obtain the semantic text of the information element in the at least one message in the signaling data.

[0010] In an embodiment of the present application, at least one tree structure is obtained by parsing each message in the signaling data, and the tree structure is transformed to obtain at least one serialized text, and then the serialized text is semantically segmented to obtain the semantic text of the information element in at least one message. In this way, semantic overlap can be achieved between information elements, and the relevance can be reflected in the form of text. At the same time, the representation of semantic text avoids the information loss and field applicability problems caused by preprocessing in the case of a large number of categories in the prior art. The above-mentioned serialization processing also facilitates the integration of the structural information of the signaling data into the existing pre-training model, which is beneficial for the algorithm to learn the underlying combination and generation logic of the data.

[0011] The tree structure is obtained by parsing the nested relationship of cells in the message, etc. Each node in the tree structure represents a key or a value of a cell.

[0012] The serialized text can be understood as a fragment or chapter expressed in natural language.

[0013] Among them, based on the nested relationship between the keys or values ​​of the information elements in the tree structure, text serialization can be achieved by using structural division identifiers.

[0014] By converting the aforementioned tree structure into serialized text based on structural partition identifiers, the complete structural information of the signaling data can be retained, which is very necessary for tasks such as simulation that require output of structural information.

[0015] The semantic text obtained by the above segmentation can be, for example, a token text. Among them, a token is the smallest granularity semantic unit used as input in the pre-trained model. By organizing the parsed signaling text in a serialized manner, a method similar to tokenization in language processing is used to perform fine-grained segmentation on the signaling content to capture the association between the signal elements. The token sequence generated in this way can be integrated with the natural language text for unified pre-training, and the trained model can be better migrated to other signaling scenarios.

[0016] In a possible implementation, the method further includes:

[0017] Marking the key field and the value field of the information element in the at least one message to obtain a marking result of the at least one message;

[0018] An attention matrix of the semantic text of each cell in the at least one message is generated according to the marking result of the at least one message.

[0019] In the embodiments of the present application, based on the key-value characteristics of the signaling data, a sparse attention mechanism is adopted to increase the dependence on the context in training and reasoning, and at the same time can greatly reduce the signal element fragments that need to be paid attention to, which not only reduces the amount of focused fragments between messages, but also reduces the amount of focused fragments within messages, which is very important for signaling scenarios with long contexts.

[0020] The key field indicates the cell attribute, and the value field indicates the value corresponding to the cell attribute. The key field and the value field may be marked by highlighting, indicating with specific characters, etc., which is not limited in this solution.

[0021] In a possible implementation, the method further includes:

[0022] The initial model is trained according to the semantic text of the information element in the at least one message and the attention matrix of the semantic text of each information element in the at least one message to obtain a target model, or,

[0023] The semantic text of the information element in the at least one message and the attention matrix of the semantic text of each information element in the at least one message are input into the target model for processing to generate a new message. Of course, other output results can also be obtained, such as obtaining anomaly detection results, etc., which are not limited in this solution.

[0024] In a possible implementation, generating an attention matrix of the semantic text of each cell in the at least one message according to the marking result of the at least one message includes:

[0025] Performing pooling processing on the marking results of at least one message in the signaling data respectively to obtain a pooling vector of the marking results of the at least one message;

[0026] When generating the attention matrix of the semantic text of the first cell in the i+1th message, the attention matrix of the semantic text of the first cell in the i+1th message is obtained according to the pooling vector of the marking results of the i messages before the i+1th message; i is an integer not less than 0, and when i=0, the attention matrix of the semantic text of the first cell of the first message is obtained according to the empirical information;

[0027] When generating the attention matrix of the semantic text of the Mth element in the i+1th message, the attention matrix of the semantic text of the Mth element is obtained according to the pooling vector of the marking results of the layer structure above the Mth element in the tree structure corresponding to the i+1th message and the pooling vector of the marking results of i messages before the i+1th message, where M is an integer greater than 1.

[0028] For example, for each cell, only the representation of the last position token can be taken as the pooling vector, or the average of all corresponding tokens can be taken as the pooling vector, so as to obtain the pooling vector of the marking result of at least one message.

[0029] Based on the above processing, each cell token can achieve sparse attention to different other tokens.

[0030] In a possible implementation, the method further includes:

[0031] Determine a first position code and a second position code of a key field and a value field of an information element in the at least one message, wherein the first position code represents the layer depth of the key field and the value field of the information element in the tree structure in which the key field and the value field of the information element are located, and the second position code represents the fastest order in which the key field and the value field of the information element are traversed during the parallel traversal process;

[0032] The initial model is trained according to the semantic text of the information element in the at least one message and the first position coding and the second position coding of the key field and the value field of the information element to obtain the target model.

[0033] The layer depth can be understood as, starting from the layer where the root node of the tree structure is located as the first layer, the number of layers where the current layer is located is the layer depth.

[0034] The fastest order of traversal during the parallel traversal process can be understood as starting from the root node of the tree structure (whose order is 1), and the next traversal is the leaf node of the next layer of the root node (whose order is 2), and then starting to traverse other nodes in parallel based on the leaf node, and adopting the fastest sequential method to traverse to other nodes.

[0035] In the embodiment of the present application, the model is trained by encoding the two positions of the key and value fields of each cell based on the depth of the layer in the tree structure where the key and value fields of the cell correspond to and the fastest traversal order during the parallel traversal process. This encoding method can greatly reduce the dependency length within the message and significantly improve the efficiency of the model in generating signaling data.

[0036] In a possible implementation, after traversing the kth key field, the key or value field downstream of the kth key field and the adjacent key field at the same layer as the kth key field are traversed at the same time, where k is a positive integer; wherein the first position code of the kth key field and the first position code of the adjacent key field at the same layer as the kth key field are both n1, the first position code of the key or value field downstream of the kth key field is n1+1, the second position code of the kth key field is n2, the second position code of the key or value field downstream of the kth key field and the second position code of the adjacent key field at the same layer as the kth key field are both n2+1.

[0037] Based on this parallel design, the efficiency of the model in generating signaling data can be significantly improved.

[0038] In a second aspect, an embodiment of the present application provides a signaling data processing device, including:

[0039] A parsing module, used to parse each message in at least one message in the signaling data to obtain at least one tree structure;

[0040] A conversion module, used for converting each tree structure in the at least one tree structure by using the structure division identifier to obtain at least one serialized text, wherein the serialized text contains the structure division identifier;

[0041] A segmentation module is used to perform semantic segmentation on each serialized text in the at least one serialized text to obtain the semantic text of the information element in the at least one message in the signaling data.

[0042] In a possible implementation manner, the device further includes a processing module, configured to:

[0043] Marking the key field and the value field of the information element in the at least one message to obtain a marking result of the at least one message;

[0044] An attention matrix of the semantic text of each cell in the at least one message is generated according to the marking result of the at least one message.

[0045] In a possible implementation, the processing module is further configured to:

[0046] The initial model is trained according to the semantic text of the information element in the at least one message and the attention matrix of the semantic text of each information element in the at least one message to obtain a target model, or,

[0047] The semantic text of the information element in the at least one message and the attention matrix of the semantic text of each information element in the at least one message are input into the target model for processing to generate a new message.

[0048] In a possible implementation, the processing module is further configured to:

[0049] Performing pooling processing on the marking results of at least one message in the signaling data respectively to obtain a pooling vector of the marking results of the at least one message;

[0050] When generating the attention matrix of the semantic text of the first cell in the i+1th message, the attention matrix of the semantic text of the first cell in the i+1th message is obtained according to the pooling vector of the marking results of the i messages before the i+1th message; i is an integer not less than 0, and when i=0, the attention matrix of the semantic text of the first cell of the first message is obtained according to the empirical information;

[0051] When generating the attention matrix of the semantic text of the Mth element in the i+1th message, the attention matrix of the semantic text of the Mth element is obtained according to the pooling vector of the marking results of the layer structure above the Mth element in the tree structure corresponding to the i+1th message and the pooling vector of the marking results of i messages before the i+1th message, where M is an integer greater than 1.

[0052] In a possible implementation manner, the device further includes a determining module, configured to:

[0053] Determine a first position code and a second position code of a key field and a value field of an information element in the at least one message, wherein the first position code represents the layer depth of the key field and the value field of the information element in the tree structure in which the key field and the value field of the information element are located, and the second position code represents the fastest order in which the key field and the value field of the information element are traversed during the parallel traversal process;

[0054] The initial model is trained according to the semantic text of the information element in the at least one message and the first position coding and the second position coding of the key field and the value field of the information element to obtain the target model.

[0055] In a possible implementation, after traversing the kth key field, the key or value field downstream of the kth key field and the adjacent key field at the same layer as the kth key field are traversed at the same time, where k is a positive integer; wherein the first position code of the kth key field and the first position code of the adjacent key field at the same layer as the kth key field are both n1, the first position code of the key or value field downstream of the kth key field is n1+1, the second position code of the kth key field is n2, the second position code of the key or value field downstream of the kth key field and the second position code of the adjacent key field at the same layer as the kth key field are both n2+1.

[0056] In a third aspect, the present application provides a signaling data processing device, comprising a processor and a memory; wherein the memory is used to store program code, and the processor is used to call the program code to execute the described method.

[0057] In a fourth aspect, the present application provides a computer storage medium comprising computer instructions, which, when executed on an electronic device, enables the electronic device to execute a method as provided in any possible implementation of the first aspect.

[0058] In a fifth aspect, an embodiment of the present application provides a computer program product. When the computer program product runs on a computer, it enables the computer to execute a method provided in any possible implementation manner of the first aspect.

[0059] In a sixth aspect, an embodiment of the present application provides a chip, comprising a processor and a data interface, wherein the processor reads instructions stored in a memory through the data interface to execute a method provided by any possible implementation method of the first aspect.

[0060] Optionally, as an implementation method, the chip may also include a memory, in which instructions are stored, and the processor is used to execute the instructions stored in the memory. When the instructions are executed, the processor is used to execute the method provided by any possible implementation method of the first aspect.

[0061] It can be understood that the device described in the second aspect, the device described in the third aspect, the computer storage medium described in the fourth aspect, the computer program product described in the fifth aspect or the chip described in the sixth aspect provided above are all used to execute any method provided in the first aspect.

[0062] Therefore, the beneficial effects that can be achieved can refer to the beneficial effects in the corresponding methods, and will not be repeated here. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] The following is an introduction to the drawings used in the embodiments of the present application.

[0064] Figure 1 It is a schematic diagram of the architecture of a signaling data processing system provided in an embodiment of the present application;

[0065] Figure 2 It is a flowchart of a signaling data processing method provided in an embodiment of the present application;

[0066] Figure 3a It is a schematic diagram of a serialized tree structure provided in an embodiment of the present application;

[0067] Figure 3bThis is a semantic segmentation schematic diagram provided in an embodiment of the present application;

[0068] Figure 4a It is a flowchart of another signaling data processing method provided in an embodiment of the present application;

[0069] Figure 4b is a schematic diagram of generating a cell fragment provided by an embodiment of the present application;

[0070] Figure 5a It is a flowchart of another signaling data processing method provided in an embodiment of the present application;

[0071] Figure 5b It is a schematic diagram of a parallel generation field provided in an embodiment of the present application;

[0072] Figure 6 It is a flowchart of a signaling data processing device provided in an embodiment of the present application;

[0073] Figure 7 It is a flow chart of another signaling data processing device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0074] The following describes the embodiments of the present application in conjunction with the drawings in the embodiments of the present application. The terms used in the implementation method part of the embodiments of the present application are only used to explain the specific embodiments of the present application, and are not intended to limit the present application.

[0075] At present, some simple signaling analysis is done by compressing the information in the signaling to reduce the amount of data, or simplifying the way the data is associated to reduce the complexity of the analysis. This sacrifices the integrity and relevance of the data to a large extent, and cannot cover a wide range of signaling analysis scenarios. In view of this, the present application provides a signaling data processing method that can achieve a complete semantic representation of the signaling; and based on the semantic representation obtained by the processing, it can achieve better results in downstream tasks in a variety of signaling scenarios, and have a stronger model migration capability.

[0076] The following will describe the system architecture of the embodiment of the present application in detail with reference to the accompanying drawings. Figure 1 , Figure 1 It is a schematic diagram of a signaling data processing system applicable to an embodiment of the present application, the system comprising a server 101 and a client 102.

[0077] The server 101 is a device with centralized computing capabilities. Exemplarily, the server 101 can be implemented by a server, a virtual machine, a cloud, or a robot.

[0078] When the server 101 includes a server, the type of the server includes but is not limited to a general-purpose computer, a dedicated server computer, a blade server, etc. This application does not strictly limit the number of servers included in the server 101, and the number can be one or more (such as a server cluster, etc.).

[0079] A virtual machine refers to a computing module that has complete hardware system functions and runs in a completely isolated environment through software simulation. Of course, in addition to virtual machines, the server 101 can also be implemented through other computing instances, such as containers.

[0080] The cloud is a software platform that uses application virtualization technology, which enables one or more software and applications to be developed and run in an independent virtualized environment. Optionally, when the server 101 is implemented through the cloud, the cloud can be deployed on a public cloud, a private cloud, or a hybrid cloud.

[0081] The client 102 or user end is a program corresponding to the server and providing local services to the client. The client 102 of the embodiment of the present application may be a handheld terminal, a wearable device, an entertainment device, etc.

[0082] In the embodiment of the present application, the client 102 can transmit data to the server 101, and the server 101 can obtain a trained model by processing the signaling data and then inputting the processing result into the pre-trained model for processing. The server 101 can send the model parameters to the client 102 so that the client 102 can update the model parameters. Alternatively, the client 102 can transmit data to the server 101, and the server 101 can process the signaling data and then input the processing result into the target model for processing to obtain the output result of the model, and then the server 101 sends the output result to the client 102. In this example, the client 102 can also process the signaling data, and this solution does not limit this.

[0083] In another possible implementation, the signaling data processing system may include only the client 102 , that is, the client 102 may complete the signaling data processing alone.

[0084] In another possible implementation, the model parameter optimization system may only include the server 101, that is, the server 101 can complete the signaling data processing alone.

[0085] Of course, other methods may also be used, and this solution does not limit this.

[0086] Among them, the semantic text obtained based on the signaling data processing method provided in the embodiment of the present application can be used as the input of the model and can be applied to multiple downstream task scenarios: signaling anomaly detection, signaling process comparison, signaling simulation, signaling description generation, etc. For example, in the signaling anomaly detection scenario, based on the signaling process input by the user, the model gives the degree of abnormality of the process to be tested based on the common patterns identified in the training. For another example, in the signaling process comparison, the model can give the degree of matching between processes or recommend similar processes based on the semantic representation of the signaling. For another example, in the signaling simulation task, after a given preceding message, the generation ability of the model can be used to predict the next message. For another example, in the signaling description generation scenario, in view of the fusion of signaling data and natural language in the pre-training stage of the model, the above two types of data can be better associated to generate a more reasonable natural language description.

[0087] The architecture of the embodiment of the present application is described above, and the method of the embodiment of the present application is described in detail below.

[0088] Reference Figure 2 FIG. 1 is a flow chart of a signaling data processing method provided in an embodiment of the present application. Optionally, the method can be applied to the aforementioned signaling data processing system, for example Figure 1 The signaling data processing system shown in FIG. Figure 2 The signaling data processing method shown may include steps 201-203. It should be understood that for the convenience of description, this application is described in the order of 201-203, and is not intended to limit the execution to the above order. The embodiment of the present application does not limit the execution order, execution time, execution number, etc. of the above one or more steps. The following description takes the execution subject of steps 201-203 of the signaling data processing method as a server as an example, and this application is also applicable to other execution subjects. Steps 201-203 are as follows:

[0089] 201. Parse each message in at least one message in the signaling data to obtain at least one tree structure;

[0090] Signaling data records the operation details of the communication network in detail. Signaling refers to the control information transmitted between devices to coordinate the operation of various devices (network elements) in the communication network. It is used to describe the operation status of each device and put forward the connection requirements for related devices. Signaling is encapsulated in messages. A message consists of many required and optional parameters (cells). Among them, a business process corresponds to a sequence of messages. Each message is stored in binary format in the nested type, length, value (Type, Length, Value, TLV) format.

[0091] The tree structure is obtained by parsing based on the nested relationship of cells in the message, etc. Each node in the tree structure represents a key field or a value field of a cell.

[0092] Optionally, the message can be parsed into a tree structure based on a signaling parsing algorithm, etc. For example, the message is input into a signaling parsing tool for processing. Other parsing methods can also be used, and this solution does not limit this.

[0093] In a possible implementation manner, the at least one message corresponds one-to-one to at least one tree structure.

[0094] 202. Using the structure division identifier, transform each tree structure in the at least one tree structure to obtain at least one serialized text, wherein the serialized text includes the structure division identifier;

[0095] The structure division mark may be, for example, "{", "}", etc., where "{" indicates the beginning of the current layer structure, and "}" indicates the end of the current layer structure. Of course, other designs are also possible, such as using an indentation to indicate the beginning of the current layer structure, and using a line break to indicate the end of the current layer structure.

[0096] The serialized text can be understood as a fragment or chapter expressed in natural language.

[0097] Among them, based on the nested relationship between the keys or values ​​of the information elements in the tree structure, text serialization can be achieved by using structural division identifiers.

[0098] By converting the aforementioned tree structure into serialized text based on structural partition identifiers, the complete structural information of the signaling data can be retained, which is very necessary for tasks such as simulation that require output of structural information.

[0099] like Figure 3a FIG. 1 is a schematic diagram of a tree structure serialized according to an embodiment of the present application, wherein “{” and “}” are used to distinguish each layer structure.

[0100] 203. Perform semantic segmentation on each serialized text in the at least one serialized text to obtain the semantic text of the information element in the at least one message in the signaling data.

[0101] Semantic segmentation can be used to divide text into fine-grained and complete semantic parts.

[0102] Optionally, with the help of byte pair encoder (BPE), the words are first divided into single characters, and then the most frequent pair of characters is replaced with another character in turn until the number of cycles is completed. The optimal word list is generated according to the data by gradually merging bytes, and the serialized text is segmented based on this.

[0103] The semantic text obtained by the above segmentation can be, for example, a token text. Among them, token is the smallest granularity semantic unit used as input in the pre-trained model. In this way, the above semantic text can be input into the pre-trained model for processing. Of course, the above semantic text and the natural language text related to the signaling data can also be input into the pre-trained model for processing, etc., so that semantic supplementation can be performed, so that the semantics learned by the pre-trained model can be transferred to the signal elements of other scenarios, etc., and this solution does not limit this.

[0104] By organizing the parsed signaling text in a serialized manner, the signaling content is finely segmented using a tokenization method similar to that in language processing to capture the association between the cells. The token sequence generated in this way can be integrated with the natural language text for unified pre-training, and the trained model can be better migrated to other signaling scenarios.

[0105] like Figure 3b FIG. 1 is a schematic diagram of semantic segmentation provided in an embodiment of the present application. Based on the semantic segmentation processing, a token text "Initial Context Setup Failure" and the like can be obtained.

[0106] In an embodiment of the present application, at least one tree structure is obtained by parsing each message in the signaling data, and the tree structure is transformed to obtain at least one serialized text, and then the serialized text is semantically segmented to obtain the semantic text of the information element in at least one message. In this way, semantic overlap can be achieved between information elements, and the relevance can be reflected in the form of text. At the same time, the representation of semantic text avoids the information loss and field applicability problems caused by preprocessing in the case of a large number of categories in the prior art. The above-mentioned serialization processing also facilitates the integration of the structural information of the signaling data into the existing pre-training model, which is beneficial for the algorithm to learn the underlying combination and generation logic of the data.

[0107] At present, in real business scenarios, there is a problem that the length of the signaling text is too long, which leads to reduced processing efficiency. Based on this, the embodiment of the present application provides a flow chart of another signaling data processing method. The method can be applied to the aforementioned signaling data processing system, for example Figure 1 The signaling data processing system shown in FIG. Figure 4aThe signaling data processing method shown may include steps 401-405. It should be understood that for the convenience of description, this application is described in the order of 401-405, and is not intended to limit the execution to the above order. The embodiment of the present application does not limit the execution order, execution time, execution number, etc. of the above one or more steps. The following description takes the execution subject of steps 401-405 of the signaling data processing method as a server as an example, and this application is also applicable to other execution subjects. Steps 401-405 are as follows:

[0108] 401. Parse each message in at least one message in the signaling data to obtain at least one tree structure;

[0109] For the introduction of step 401, please refer to the aforementioned Figure 2 The description of step 201 in the illustrated embodiment will not be repeated here.

[0110] 402. Using the structure division identifier, transform each tree structure in the at least one tree structure to obtain at least one serialized text, wherein the serialized text includes the structure division identifier;

[0111] For the introduction of step 402, please refer to the aforementioned Figure 2 The description of step 202 in the illustrated embodiment will not be repeated here.

[0112] 403. Perform semantic segmentation on each serialized text in the at least one serialized text to obtain a semantic text of a cell in the at least one message in the signaling data;

[0113] For the introduction of step 403, please refer to the aforementioned Figure 2 The description of step 203 in the illustrated embodiment will not be repeated here.

[0114] 404. Mark the key field and the value field of the information element of each message in the at least one message to obtain a marking result of the at least one message;

[0115] The key field indicates the cell attribute, and the value field indicates the value corresponding to the cell attribute. The key field and the value field may be marked by highlighting, indicating with specific characters, etc., which is not limited in this solution.

[0116] Optionally, the tag content may include the continuous token text corresponding to the complete key field or value field, which is used to generate the pooling vector of the information element. The tag content may also include the nested relationship between fields, which is used to generate a sparse matrix that only focuses on the upper-level information element in the message.

[0117] 405. Generate an attention matrix of the semantic text of each cell in the at least one message according to the marking result of the at least one message.

[0118] In a possible implementation, the marking results of at least one message in the signaling data are pooled respectively to obtain a pooled vector of the marking results of the at least one message. For example, for each information element, only the representation of the last position token can be taken as the pooled vector, or the average of all corresponding tokens can be taken as the pooled vector, so as to obtain the pooled vector of the marking results of the at least one message.

[0119] Then, when generating the attention matrix of the semantic text of the first cell in the i+1th message, the attention matrix of the semantic text of the first cell in the i+1th message is obtained according to the pooling vector of the marking results of the i messages before the i+1th message.

[0120] Wherein, i is an integer not less than 0. When i=0, the attention matrix of the semantic text of the first cell of the first message can be obtained based on experience information, such as historical experience data.

[0121] When generating the attention matrix of the semantic text of the Mth element in the i+1th message, the attention matrix of the semantic text of the Mth element is obtained according to the pooling vector of the marking results of the layer structure above the Mth element in the tree structure corresponding to the i+1th message and the pooling vector of the marking results of i messages before the i+1th message, where M is an integer greater than 1.

[0122] In this way, each cell token can achieve sparse attention to different other tokens.

[0123] Optionally, the initial model is trained according to the semantic text of the information element in the at least one message and the attention matrix of the semantic text of each information element in the at least one message to obtain the target model.

[0124] For another example, the semantic text of the information element in the at least one message and the attention matrix of the semantic text of each information element in the at least one message are input into the target model for processing to generate a new message, or to obtain anomaly detection results, etc. This solution does not limit this.

[0125] like Figure 4b As shown, it is a schematic diagram of generating a cell fragment provided by an embodiment of the present application. When generating the value K of the cell field J, only the cell fragments at the upper layer of its tree structure need to be concerned, such as fields "A", "G", "J", etc.

[0126] In an embodiment of the present application, each message in the signaling data is parsed to obtain at least one tree structure, and the tree structure is transformed to obtain at least one serialized text, and the serialized text is semantically segmented to obtain the semantic text of the information element in at least one message. Then, based on the key-value characteristics of the signaling data, a sparse attention mechanism is used to increase the dependence on the context in training and reasoning, and at the same time, the information element fragments that need to be paid attention to are greatly reduced, which not only reduces the amount of focused fragments between messages, but also reduces the amount of focused fragments within messages, which plays a very important role in signaling scenarios with long contexts.

[0127] Reference Figure 5a FIG. 2 is a flow chart of another signaling data processing method provided in an embodiment of the present application. Optionally, the method can be applied to the aforementioned signaling data processing system, for example Figure 1 The signaling data processing system shown in FIG. Figure 5a The signaling data processing method shown may include steps 501-505. It should be understood that for the convenience of description, this application is described in the order of 501-505, and is not intended to limit the execution to the above order. The embodiment of the present application does not limit the execution order, execution time, execution number, etc. of the above one or more steps. The following description takes the execution subject of steps 501-505 of the signaling data processing method as a server as an example, and this application is also applicable to other execution subjects. Steps 501-505 are as follows:

[0128] 501. Parse each message in at least one message in the signaling data to obtain at least one tree structure;

[0129] For the introduction of step 501, please refer to the aforementioned Figure 2 The description of step 201 in the illustrated embodiment will not be repeated here.

[0130] 502. Using the structure division identifier, transform each tree structure in the at least one tree structure to obtain at least one serialized text, wherein the serialized text includes the structure division identifier;

[0131] For the introduction of step 502, please refer to the aforementioned Figure 2 The description of step 202 in the illustrated embodiment will not be repeated here.

[0132] 503. Perform semantic segmentation on each serialized text in the at least one serialized text to obtain a semantic text of a cell in the at least one message in the signaling data;

[0133] For the introduction of step 503, please refer to the aforementioned Figure 2 The description of step 203 in the illustrated embodiment will not be repeated here.

[0134] 504. Determine a first position code and a second position code of a key field and a value field of an information element in the at least one message, wherein the first position code represents a layer depth of the key field and the value field of the information element in the tree structure, and the second position code represents an order in which the key field and the value field of the information element are traversed fastest during a parallel traversal process;

[0135] The layer depth can be understood as, starting from the layer where the root node of the tree structure is located as the first layer, the number of layers where the current layer is located is the layer depth.

[0136] The fastest order of traversal during the parallel traversal process can be understood as starting from the root node of the tree structure (whose order is 1), and the next traversal is the leaf node of the next layer of the root node (whose order is 2), and then starting the parallel traversal of other nodes based on the leaf node, and using the fastest sequential traversal method to traverse other nodes. For example, after traversing the kth key field, the key or value field located downstream of the kth key field and the adjacent key field at the same layer as the kth key field are traversed at the same time.

[0137] Wherein, k is a positive integer. It can be understood that the first position code of the kth key field and the first position code of the adjacent key field at the same level of the kth key field are both n1, the first position code of the key or value field downstream of the kth key field is n1+1, the second position code of the kth key field is n2, the second position code of the key or value field downstream of the kth key field and the second position code of the adjacent key field at the same level of the kth key field are both n2+1.

[0138] 505. Train the initial model according to the semantic text of the information element in the at least one message and the first position code and the second position code of the key field and the value field of the information element to obtain a target model.

[0139] The first position code and the second position code can be introduced into the generative pre-trained model through the absolute position information and the relative position information. For example, a new position code can be generated by concatenating two types of position codes and used as the model input together with the word embedding representation. Figure 4a The attention calculation in the illustrated embodiment introduces weights regarding relative position encoding to take into account position information, etc.

[0140] Step 505 is only described by taking model training as an example. It is understandable that the same is also applicable in the reasoning process. For example, Figure 5bAs shown in the figure, after the model finishes generating a key field, it can start generating the downstream key field or value field, as well as the sibling key field at the same level: for the downstream key field or value field, the code representing the layer depth is increased by one, and the code representing the traversal position is increased by one; for the sibling key field at the same level, the code representing the layer depth remains unchanged, and the code representing the traversal position is increased by one. By introducing the two-dimensional position encoding of tree information, the dependence on the absolute position information of the serialized text is avoided, the parallelization of model generation is achieved, and the generation efficiency is accelerated.

[0141] like Figure 5b As shown, according to the tree structure of the signaling, the order of its traversal can be determined, and the position coding of the field can be obtained in combination with the layer depth information. The first position coding and the second position coding of field A are 1, 1 respectively; the first position coding and the second position coding of field B are 2, 2 respectively; the first position coding and the second position coding of field C are 3, 3 respectively; the first position coding and the second position coding of field D are 2, 3 respectively; the first position coding and the second position coding of field E are 3, 4 respectively; the first position coding and the second position coding of field F are 4, 5 respectively; the first position coding and the second position coding of field G are 2, 4 respectively; the first position coding and the second position coding of field H are 3, 5 respectively; the first position coding and the second position coding of field I are 4, 6 respectively; the first position coding and the second position coding of field J are 3, 6 respectively; the first position coding and the second position coding of field K are 4, 7 respectively.

[0142] Should Figure 5b The example is a possible implementation method, and other methods can be used. For example, the field G is traversed in the second order, and the field D and field J are traversed in the third order. Of course, the field D can also be traversed in the second order, and this solution does not limit this.

[0143] In the embodiment of the present application, each message in the signaling data is parsed to obtain at least one tree structure, and the tree structure is transformed to obtain at least one serialized text, and the serialized text is semantically segmented to obtain the semantic text of the information element in at least one message, and the two position codes of the key and value fields of each information element are determined based on the layer depth of the tree structure where the key and value fields of the information element correspond to, and the order of being traversed the fastest during the parallel traversal process, so as to train the model. This encoding method can greatly reduce the dependency length within the message and significantly improve the efficiency of the model in generating signaling data.

[0144] It should be noted that, in the various embodiments of the present application, if there is no special explanation and logical conflict, the terms and / or descriptions between the various embodiments are consistent and can be referenced to each other, and the technical features in different embodiments can be combined to form new embodiments according to their internal logical relationships. For example, Figure 4a The illustrated embodiment can be used with Figure 5a The embodiments shown are combined. That is, before model processing, an attention matrix of the semantic text of each cell in the at least one message can be generated; and the first position code and the second position code of the key field and the value field of the cell in the at least one message can be determined; and then, processing is performed based on the semantic text of the cell in at least one message, the attention matrix of the semantic text of each cell, and the first position code and the second position code of the key field and the value field of the cell in at least one message. For example, the semantic text of the cell in at least one message, the attention matrix of the semantic text of each cell, and the first position code and the second position code of the key field and the value field of the cell in at least one message are input into the initial model for training to obtain the target model. Of course, reasoning application can also be performed, and this scheme does not limit the above-mentioned implementation method and implementation order.

[0145] The above detailed description of the method of the embodiment of the present application, the following provides a device of the embodiment of the present application. It can be understood that in the various device embodiments of the present application, the division of multiple units or modules is only a logical division according to function, and is not used as a limitation on the specific structure of the device. In a specific implementation, some functional modules may be subdivided into more small functional modules, and some functional modules may also be combined into one functional module, but no matter whether these functional modules are subdivided or combined, the general process performed by the device is the same. For example, some devices contain a receiving unit and a sending unit. In some designs, the sending unit and the receiving unit can also be integrated into a communication unit, which can implement the functions implemented by the receiving unit and the sending unit. Usually, each unit corresponds to its own program code (or program instruction), and when the program code corresponding to each of these units is run on the processor, the unit is controlled by the processing unit to execute the corresponding process to implement the corresponding function.

[0146] The embodiments of the present application also provide a device for implementing any of the above methods. For example, a signaling data processing device is provided, including a module (or means) for implementing each step performed by the server in any of the above methods.

[0147] For example, refer to Figure 6 FIG. 1 is a schematic diagram of the structure of a signaling data processing device provided in an embodiment of the present application. The signaling data processing device is used to implement the aforementioned signaling data processing method, for example Figure 2 , Figure 4a, Figure 5a The signaling data processing method shown.

[0148] like Figure 6 As shown, the device may include a parsing module 601, a conversion module 602 and a segmentation module 603, which are specifically as follows:

[0149] The parsing module 601 is used to parse each message in at least one message in the signaling data to obtain at least one tree structure;

[0150] A conversion module 602, configured to convert each tree structure in the at least one tree structure using the structure division identifier to obtain at least one serialized text, wherein the serialized text includes the structure division identifier;

[0151] The segmentation module 603 is used to perform semantic segmentation on each serialized text in the at least one serialized text to obtain the semantic text of the information element in the at least one message in the signaling data.

[0152] In a possible implementation manner, the device further includes a processing module, configured to:

[0153] Marking the key field and the value field of the information element in the at least one message to obtain a marking result of the at least one message;

[0154] An attention matrix of the semantic text of each cell in the at least one message is generated according to the marking result of the at least one message.

[0155] In a possible implementation, the processing module is further configured to:

[0156] The initial model is trained according to the semantic text of the information element in the at least one message and the attention matrix of the semantic text of each information element in the at least one message to obtain a target model, or,

[0157] The semantic text of the information element in the at least one message and the attention matrix of the semantic text of each information element in the at least one message are input into the target model for processing to generate a new message.

[0158] In a possible implementation, the processing module is further configured to:

[0159] Performing pooling processing on the marking results of at least one message in the signaling data respectively to obtain a pooling vector of the marking results of the at least one message;

[0160] When generating the attention matrix of the semantic text of the first cell in the i+1th message, the attention matrix of the semantic text of the first cell in the i+1th message is obtained according to the pooling vector of the marking results of the i messages before the i+1th message; i is an integer not less than 0, and when i=0, the attention matrix of the semantic text of the first cell of the first message is obtained according to the empirical information;

[0161] When generating the attention matrix of the semantic text of the Mth element in the i+1th message, the attention matrix of the semantic text of the Mth element is obtained according to the pooling vector of the marking results of the layer structure above the Mth element in the tree structure corresponding to the i+1th message and the pooling vector of the marking results of i messages before the i+1th message, where M is an integer greater than 1.

[0162] In a possible implementation manner, the device further includes a determining module, configured to:

[0163] Determine a first position code and a second position code of a key field and a value field of an information element in the at least one message, wherein the first position code represents the layer depth of the key field and the value field of the information element in the tree structure in which the key field and the value field of the information element are located, and the second position code represents the fastest order in which the key field and the value field of the information element are traversed during the parallel traversal process;

[0164] The initial model is trained according to the semantic text of the information element in the at least one message and the first position coding and the second position coding of the key field and the value field of the information element to obtain the target model.

[0165] In a possible implementation, after traversing the kth key field, the key or value field downstream of the kth key field and the adjacent key field at the same layer as the kth key field are traversed at the same time, where k is a positive integer; wherein the first position code of the kth key field and the first position code of the adjacent key field at the same layer as the kth key field are both n1, the first position code of the key or value field downstream of the kth key field is n1+1, the second position code of the kth key field is n2, the second position code of the key or value field downstream of the kth key field and the second position code of the adjacent key field at the same layer as the kth key field are both n2+1.

[0166] For the introduction of the above modules, please refer to the description of the aforementioned method embodiment, which will not be repeated here.

[0167] It should be understood that the division of the modules in the above devices is only a division of logical functions. In actual implementation, they can be fully or partially integrated into one physical entity, or they can be physically separated. In addition, the modules in the signaling data processing device can be implemented in the form of a processor calling software; for example, the signaling data processing device includes a processor, the processor is connected to a memory, the memory stores instructions, and the processor calls the instructions stored in the memory to implement any of the above methods or implement the functions of the modules of the device, wherein the processor is, for example, a general-purpose processor, such as a central processing unit (CPU) or a microprocessor, and the memory is a memory inside the device or a memory outside the device. Alternatively, the modules in the device can be implemented in the form of hardware circuits, and the functions of some or all units can be realized by designing the hardware circuits, and the hardware circuits can be understood as one or more processors; for example, in one implementation, the hardware circuit is an application-specific integrated circuit (ASIC), and the functions of some or all of the above units are realized by designing the logical relationship of the components in the circuit; for another example, in another implementation, the hardware circuit can be realized by a programmable logic device (PLD), taking a field programmable gate array (FPGA) as an example, which can include a large number of logic gate circuits, and the connection relationship between the logic gate circuits is configured by a configuration file, so as to realize the functions of some or all of the above units. All modules of the above devices can be realized in the form of a processor calling software, or in the form of hardware circuits, or in part by a processor calling software, and the rest by hardware circuits.

[0168] Reference Figure 7 FIG. 1 is a schematic diagram of the hardware structure of another signaling data processing device provided in an embodiment of the present application. Figure 7 The signaling data processing device 700 shown (the device 700 may be a computer device) includes a memory 701, a processor 702, a communication interface 703 and a bus 704. The memory 701, the processor 702 and the communication interface 703 are connected to each other through the bus 704.

[0169] The memory 701 may be a read only memory (ROM), a static storage device, a dynamic storage device or a random access memory (RAM).

[0170] The memory 701 can store programs. When the program stored in the memory 701 is executed by the processor 702, the processor 702 and the communication interface 703 are used to execute the various steps of the signaling data processing method of the embodiment of the present application.

[0171] The processor 702 is a circuit with signal processing capability. In one implementation, the processor 702 may be a circuit with instruction reading and running capability, such as a central processing unit CPU, a microprocessor, a graphics processing unit (GPU) (which may be understood as a microprocessor), or a digital signal processor (DSP), etc.; in another implementation, the processor 702 may implement certain functions through the logical relationship of a hardware circuit, and the logical relationship of the hardware circuit may be fixed or reconfigurable, such as the processor 702 being a hardware circuit implemented by an ASIC or a programmable logic device PLD, such as an FPGA. In a reconfigurable hardware circuit, the process of the processor loading a configuration document to implement the hardware circuit configuration may be understood as the process of the processor loading instructions to implement the functions of some or all of the above modules. In addition, it may also be a hardware circuit designed for artificial intelligence, which may be understood as an ASIC, such as a neural network processing unit (NPU), a tensor processing unit (TPU), a deep learning processing unit (DPU), etc. The processor 702 is used to execute relevant programs to implement the functions that need to be performed by the units in the signaling data processing device of the embodiment of the present application, or to execute the signaling data processing method of the method embodiment of the present application.

[0172] It can be seen that each module in the above device can be one or more processors (or processing circuits) configured to implement the above method, such as: CPU, GPU, NPU, TPU, DPU, microprocessor, DSP, ASIC, FPGA, or a combination of at least two of these processor forms.

[0173] In addition, the modules in the above device can be fully or partially integrated together, or can be implemented independently. In one implementation, these modules are integrated together and implemented in the form of a system-on-a-chip (SOC). The SOC may include at least one processor for implementing any of the above methods or implementing the functions of the modules of the device. The type of the at least one processor may be different, for example, including a CPU and an FPGA, a CPU and an artificial intelligence processor, a CPU and a GPU, etc.

[0174] The communication interface 703 uses a transceiver device such as, but not limited to, a transceiver to implement communication between the device 700 and other devices or a communication network. For example, data can be obtained through the communication interface 703.

[0175] The bus 704 may include a path for transmitting information between various components of the device 700 (eg, the memory 701 , the processor 702 , and the communication interface 703 ).

[0176] It should be noted that although Figure 7 The device 700 shown only shows a memory, a processor, and a communication interface, but in the specific implementation process, those skilled in the art should understand that the device 700 also includes other devices necessary for normal operation. At the same time, according to specific needs, those skilled in the art should understand that the device 700 may also include hardware devices for implementing other additional functions. In addition, those skilled in the art should understand that the device 700 may also only include the devices necessary to implement the embodiments of the present application, and does not necessarily include Figure 7 All devices shown in .

[0177] An embodiment of the present application also provides a computer-readable storage medium, which stores instructions. When the computer-readable storage medium is executed on a computer or a processor, the computer or the processor executes one or more steps in any of the above methods.

[0178] The embodiment of the present application further provides a computer program product including instructions. When the computer program product is executed on a computer or a processor, the computer or the processor executes one or more steps in any of the above methods.

[0179] It should be understood that in the description of the present application, unless otherwise specified, " / " indicates that the objects associated before and after are in an "or" relationship, for example, A / B can represent A or B; wherein A and B can be singular or plural. Also, in the description of the present application, unless otherwise specified, "multiple" refers to two or more than two. "At least one of the following" or similar expressions refers to any combination of these items, including any combination of single items or plural items. For example, at least one of a, b, or c can represent: a, b, c, ab, ac, bc, or abc, wherein a, b, c can be single or multiple. In addition, in order to facilitate the clear description of the technical solutions of the embodiments of the present application, in the embodiments of the present application, the words "first", "second", etc. are used to distinguish the same items or similar items with substantially the same functions and effects. Those skilled in the art can understand that the words "first", "second", etc. do not limit the quantity and execution order, and the words "first", "second", etc. do not limit them to be necessarily different. Meanwhile, in the embodiments of the present application, words such as "exemplary" or "for example" are used to indicate examples, illustrations or descriptions. Any embodiment or design described as "exemplary" or "for example" in the embodiments of the present application should not be interpreted as being more preferred or more advantageous than other embodiments or designs. Specifically, the use of words such as "exemplary" or "for example" is intended to present related concepts in a concrete manner for ease of understanding.

[0180] In the several embodiments provided in the present application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the division of the unit is only a logical function division, and there may be other division methods in actual implementation, for example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. The mutual coupling, direct coupling, or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0181] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0182] In the above embodiments, it can be implemented in whole or in part by software, hardware, firmware or any combination thereof. When implemented using software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the process or function according to the embodiment of the present application is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions can be stored in a computer-readable storage medium or transmitted through the computer-readable storage medium. The computer instructions can be transmitted from a website site, computer, server or data center to another website site, computer, server or data center by wired (e.g., coaxial cable, optical fiber, digital subscriber line (digital subscriber line, DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) mode. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that contains one or more available media integrations. The available medium may be a read-only memory (ROM), or a random access memory (RAM), or a magnetic medium, such as a floppy disk, a hard disk, a tape, a magnetic disk, or an optical medium, such as a digital versatile disc (DVD), or a semiconductor medium, such as a solid state disk (SSD), etc.

[0183] The above is only a specific implementation of the embodiment of the present application, but the protection scope of the embodiment of the present application is not limited thereto, and any changes or replacements within the technical scope disclosed in the embodiment of the present application should be included in the protection scope of the embodiment of the present application. Therefore, the protection scope of the embodiment of the present application should be based on the protection scope of the claims.

Claims

1. A signaling data processing method, characterized in that: include: Parsing each message in at least one message in the signaling data to obtain at least one tree structure; Using the structure division identifier, transform each tree structure in the at least one tree structure to obtain at least one serialized text, wherein the serialized text includes the structure division identifier; Semantic segmentation is performed on each serialized text in the at least one serialized text to obtain the semantic text of the information element in the at least one message in the signaling data.

2. The method according to claim 1, characterized in that The method further comprises: Marking the key field and the value field of the information element in the at least one message to obtain a marking result of the at least one message; An attention matrix of the semantic text of each cell in the at least one message is generated according to the marking result of the at least one message.

3. The method according to claim 2, characterized in that The method further comprises: The initial model is trained according to the semantic text of the information element in the at least one message and the attention matrix of the semantic text of each information element in the at least one message to obtain a target model, or, The semantic text of the information element in the at least one message and the attention matrix of the semantic text of each information element in the at least one message are input into the target model for processing to generate a new message.

4. The method according to claim 2, characterized in that: The step of generating an attention matrix of the semantic text of each cell in the at least one message according to the marking result of the at least one message comprises: Performing pooling processing on the marking results of at least one message in the signaling data respectively to obtain a pooling vector of the marking results of the at least one message; When generating the attention matrix of the semantic text of the first cell in the i+1th message, the attention matrix of the semantic text of the first cell in the i+1th message is obtained according to the pooling vector of the marking results of the i messages before the i+1th message; i is an integer not less than 0, and when i=0, the attention matrix of the semantic text of the first cell of the first message is obtained according to the empirical information; When generating the attention matrix of the semantic text of the Mth element in the i+1th message, the attention matrix of the semantic text of the Mth element is obtained according to the pooling vector of the marking results of the layer structure above the Mth element in the tree structure corresponding to the i+1th message and the pooling vector of the marking results of i messages before the i+1th message, where M is an integer greater than 1.

5. The method according to claim 1, 2 or 4, characterized in that: The method further comprises: Determine a first position code and a second position code of a key field and a value field of an information element in the at least one message, wherein the first position code represents the layer depth of the key field and the value field of the information element in the tree structure in which the key field and the value field of the information element are located, and the second position code represents the fastest order in which the key field and the value field of the information element are traversed during the parallel traversal process; The initial model is trained according to the semantic text of the information element in the at least one message and the first position coding and the second position coding of the key field and the value field of the information element to obtain the target model.

6. The method according to claim 5, characterized in that After traversing the kth key field, the key or value field downstream of the kth key field and the adjacent key field at the same layer as the kth key field are traversed at the same time, where k is a positive integer; wherein the first position code of the kth key field and the first position code of the adjacent key field at the same layer as the kth key field are both n1, the first position code of the key or value field downstream of the kth key field is n1+1, the second position code of the kth key field is n2, the second position code of the key or value field downstream of the kth key field and the second position code of the adjacent key field at the same layer as the kth key field are both n2+1.

7. A signaling data processing device, characterized in that: include: A parsing module, used to parse each message in at least one message in the signaling data to obtain at least one tree structure; A conversion module, used for converting each tree structure in the at least one tree structure by using the structure division identifier to obtain at least one serialized text, wherein the serialized text contains the structure division identifier; A segmentation module is used to perform semantic segmentation on each serialized text in the at least one serialized text to obtain the semantic text of the information element in the at least one message in the signaling data.

8. The device according to claim 7, characterized in that The device also includes a processing module, configured to: Marking the key field and the value field of the information element in the at least one message to obtain a marking result of the at least one message; An attention matrix of the semantic text of each cell in the at least one message is generated according to the marking result of the at least one message.

9. The device according to claim 8, characterized in that The processing module is further used for: The initial model is trained according to the semantic text of the information element in the at least one message and the attention matrix of the semantic text of each information element in the at least one message to obtain a target model, or, The semantic text of the information element in the at least one message and the attention matrix of the semantic text of each information element in the at least one message are input into the target model for processing to generate a new message.

10. The device according to claim 8, characterized in that The processing module is further used for: Performing pooling processing on the marking results of at least one message in the signaling data respectively to obtain a pooling vector of the marking results of the at least one message; When generating the attention matrix of the semantic text of the first cell in the i+1th message, the attention matrix of the semantic text of the first cell in the i+1th message is obtained according to the pooling vector of the marking results of the i messages before the i+1th message; i is an integer not less than 0, and when i=0, the attention matrix of the semantic text of the first cell of the first message is obtained according to the empirical information; When generating the attention matrix of the semantic text of the Mth element in the i+1th message, the attention matrix of the semantic text of the Mth element is obtained according to the pooling vector of the marking results of the layer structure above the Mth element in the tree structure corresponding to the i+1th message and the pooling vector of the marking results of i messages before the i+1th message, where M is an integer greater than 1.

11. The device according to claim 7, 8 or 10, characterized in that The device also includes a determining module, configured to: Determine a first position code and a second position code of a key field and a value field of an information element in the at least one message, wherein the first position code represents the layer depth of the key field and the value field of the information element in the tree structure in which the key field and the value field of the information element are located, and the second position code represents the fastest order in which the key field and the value field of the information element are traversed during the parallel traversal process; The initial model is trained according to the semantic text of the information element in the at least one message and the first position coding and the second position coding of the key field and the value field of the information element to obtain the target model.

12. The device according to claim 11, characterized in that After traversing the kth key field, the key or value field downstream of the kth key field and the adjacent key field at the same layer as the kth key field are traversed at the same time, where k is a positive integer; wherein the first position code of the kth key field and the first position code of the adjacent key field at the same layer as the kth key field are both n1, the first position code of the key or value field downstream of the kth key field is n1+1, the second position code of the kth key field is n2, the second position code of the key or value field downstream of the kth key field and the second position code of the adjacent key field at the same layer as the kth key field are both n2+1.

13. A signaling data processing device, characterized in that: It comprises a processor and a memory; wherein the memory is used to store program code, and the processor is used to call the program code to execute the method according to any one of claims 1 to 6.

14. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and the computer program is executed by a processor to implement the method according to any one of claims 1 to 6.

15. A computer program product, characterized in that When the computer program product runs on a computer, the computer is caused to execute the method according to any one of claims 1 to 6.

Images