CN117851127A

CN117851127A - Semantic model fault processing method and device, electronic equipment and storage medium

Info

Publication number: CN117851127A
Application number: CN202410013068.7A
Authority: CN
Inventors: 王路宝
Original assignee: Chongqing Changan Automobile Co Ltd
Current assignee: Chongqing Changan Automobile Co Ltd
Priority date: 2024-01-02
Filing date: 2024-01-02
Publication date: 2024-04-09

Abstract

The invention relates to the technical field of urban area semantic service, in particular to a semantic model fault processing method, a semantic model fault processing device, electronic equipment and a storage medium. The invention obtains a fault semantic model of faults; determining at least one candidate semantic model according to a target area corresponding to the fault semantic model; selecting a target semantic model from the at least one candidate semantic model through a model selection rule; the target semantic model is used for processing the data to be processed of the target region, and the target semantic model for replacing the fault semantic model can be accurately selected from the candidate region closest to the target region where the fault semantic model is located by using the model selection rule, so that a more efficient and accurate solution is provided for fault restoration, the application effect and user experience of a semantic service system in multiple regions are improved, and resources are further saved because backup semantic models do not need to be deployed.

Description

Semantic model fault processing method and device, electronic equipment and storage medium

Technical Field

The invention relates to the technical field of urban area semantic service, in particular to a semantic model fault processing method, a semantic model fault processing device, electronic equipment and a storage medium.

Background

With the continuous development of artificial intelligence technology, semantic models are increasingly widely applied in urban areas. These applications cover many aspects of city planning, intelligent transportation, public safety, environmental monitoring, and smart cities, and play an important role in improving the efficiency and convenience of city management. The semantic model deployment scheme of the urban area is characterized in that semantic models used in different urban areas are different, and the scheme has the characteristics of customization, dialect adaptation, user group adaptation and the like.

In conventional service fail-over schemes, when a failure occurs, the system may attempt to replace the failure model with a fixed backup or default model, however this generalized replacement approach is not effective for some special cases. For example, different semantic models are deployed in different urban areas, and due to the limitation of deployment environments, backup of the semantic models cannot be deployed, when the semantic model of a certain urban area fails, a traditional service failure recovery scheme is obviously unavailable, and the area can only stop semantic services.

Therefore, the related technology has the technical problem that the semantic service is not available because the semantic model cannot be recovered immediately after failure.

Disclosure of Invention

The invention aims to provide a semantic model fault processing method to solve the problem that semantic service is not available because the semantic model cannot be recovered immediately after fault; secondly, a semantic model fault processing device is provided; a third object is to provide an electronic device; a fourth object is to provide a computer-readable storage medium.

In order to achieve the above purpose, the technical scheme adopted by the invention is as follows:

according to an aspect of the embodiments of the present application, there is provided a semantic model fault processing method, including:

acquiring a fault semantic model with faults; determining at least one candidate semantic model according to a target area corresponding to the fault semantic model, wherein a preset geographic similarity requirement is met between the candidate area corresponding to the candidate semantic model and the target area; selecting a target semantic model from the at least one candidate semantic model through a model selection rule; and processing the data to be processed of the target area through the target semantic model.

According to the technical means, a fault semantic model of a fault is obtained; determining at least one candidate semantic model according to a target area corresponding to the fault semantic model, wherein a preset geographic similarity requirement is met between the candidate area corresponding to the candidate semantic model and the target area; selecting a target semantic model from the at least one candidate semantic model through a model selection rule; the data of the target area is processed through the target semantic model, and the target semantic model for replacing the fault semantic model can be accurately selected from the candidate area closest to the target area where the fault semantic model is located by using the model selection rule, so that a more efficient and accurate solution is provided for fault restoration, the application effect and user experience of a semantic service system in multiple areas are improved, and resources are further saved because backup semantic models do not need to be deployed.

Further, the determining at least one candidate semantic model according to the target area corresponding to the fault semantic model includes: when the preset geographic similarity requirement is a distance threshold value, acquiring the candidate region of which the distance with the target region meets the distance threshold value; and/or when the preset geographic similarity requirement is a language requirement, acquiring the candidate region which meets the language requirement with the language between the candidate region and the target region; and determining a semantic model corresponding to each candidate region as the candidate semantic model.

According to the technical means, the candidate semantic model corresponding to the candidate region is accurately selected by using the distance threshold and/or the language requirement, so that the selection range of the candidate semantic model can be effectively reduced, and the efficiency of selecting the target semantic model for replacing the fault semantic model is further improved.

Further, the selecting, by the model selection rule, the target semantic model from the at least one candidate semantic model includes: obtaining historical data of the fault semantic model, wherein the historical data comprises: history identification data and a first result obtained by identifying the history identification data; respectively inputting the history identification data into each candidate semantic model to obtain a second result output by each candidate semantic model; calculating the result similarity of the first result and each second result through the model similarity algorithm; determining a model similarity value of each candidate semantic model and the fault semantic model according to all the result similarities; and determining the candidate semantic model with the maximum model similarity value as a target semantic model in the at least one candidate semantic model.

According to the technical means, the similarity analysis is carried out on each candidate semantic model and the fault semantic model by adopting the model similarity algorithm, so that the target semantic model for replacing the fault semantic model can be accurately selected.

Further, the selecting, by the model selection rule, the target semantic model from the at least one candidate semantic model includes: acquiring data to be identified of the target area; determining a target query type corresponding to the target area according to the data to be identified; acquiring a historical query record of each candidate semantic model; and determining the target semantic model with highest accuracy on the target query type from the at least one candidate semantic model according to the target query type and the historical query record.

According to the technical means, the target semantic model with highest accuracy on the query type in the candidate models is selected through the query type of the target region, so that when the semantic model of the target region fails, a replacement semantic model can be found, and the application effect and the user experience of the semantic service system in multiple regions are improved.

Further, the selecting, by the model selection rule, the target semantic model from the at least one candidate semantic model includes: acquiring data to be identified of the target area; determining a recognition result and model performance corresponding to the candidate semantic model according to the data to be recognized; and in the at least one candidate semantic model, the identification result is successful, and the candidate semantic model with the highest model performance is determined to be a target semantic model.

According to the technical means, as the candidate models are used for identifying the data to be identified of the target area and the model performance, the target semantic model which is successfully identified and has the highest model performance is selected from the candidate models, so that when the semantic model of the target area fails, a replacement semantic model can be found, and the application effect and the user experience of the semantic service system in multiple areas are improved.

Further, processing the data to be processed of the target region through the target semantic model includes: establishing a forwarding thread; and forwarding the data to be processed of the target area to the target semantic model through the forwarding thread so that the target semantic model processes the data to be processed of the target area.

According to the technical means, the data to be processed in the target area is forwarded to the target semantic model for processing through the forwarding thread, so that a more efficient and accurate solution is provided for fault restoration, and the application effect and the user experience of the semantic service system in multiple areas are improved.

Further, the obtaining a fault semantic model of the fault comprises: determining state information and performance information of a semantic model of the target area through a model running log corresponding to the semantic model of the target area, wherein the state information is used for indicating whether the semantic model of the target area runs or not, and the performance information is used for indicating the performance of the semantic model of the target area for processing semantic requests; and when the state information indicates that the semantic model of the target area is running and the performance information does not meet the preset performance requirement, determining the semantic model of the target area as the fault semantic model.

According to the technical means, the fault semantic model can be timely and accurately found by monitoring the state information and the performance information of the semantic model in the model running log of the semantic model, so that a more efficient and accurate solution is provided for fault repair, and the application effect and the user experience of the semantic service system in multiple areas are improved.

According to still another aspect of the embodiments of the present application, there is further provided a semantic model fault processing apparatus, including:

the fault semantic model acquisition module is used for acquiring a fault semantic model with faults; the candidate semantic model determining module is used for determining at least one candidate semantic model according to a target area corresponding to the fault semantic model, wherein a preset geographic similarity requirement is met between the candidate area corresponding to the candidate semantic model and the target area; the target semantic model selection module is used for selecting and obtaining a target semantic model from the at least one candidate semantic model through a model selection rule; and the data processing module is used for processing the data to be processed of the target area through the target semantic model.

According to yet another aspect of the embodiments of the present application, there is also provided an electronic device including a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory complete communication with each other through the communication bus; wherein the memory is used for storing a computer program; a processor for performing the method steps of any of the embodiments described above by running the computer program stored on the memory.

According to a further aspect of the embodiments of the present application, there is also provided a computer-readable storage medium having stored therein a computer program, wherein the computer program is arranged to perform the method steps of any of the embodiments described above when run.

The invention has the beneficial effects that:

the method comprises the steps of obtaining a fault semantic model with faults; determining at least one candidate semantic model according to a target area corresponding to the fault semantic model, wherein a preset geographic similarity requirement is met between the candidate area corresponding to the candidate semantic model and the target area; selecting a target semantic model from the at least one candidate semantic model through a model selection rule; the target semantic model is used for processing the data to be processed of the target region, and the target semantic model for replacing the fault semantic model can be accurately selected from the candidate region closest to the target region where the fault semantic model is located by using the model selection rule, so that a more efficient and accurate solution is provided for fault restoration, the application effect and user experience of a semantic service system in multiple regions are improved, and resources are further saved because backup semantic models do not need to be deployed.

Drawings

FIG. 1 is a schematic diagram of a hardware environment of an alternative semantic model fault handling method according to an embodiment of the present application;

FIG. 2 is a flow diagram of an alternative semantic model fault processing method according to an embodiment of the present application;

FIG. 3 is a flow diagram of an alternative acquisition of a fault semantic model according to an embodiment of the present application;

FIG. 4 is a flow diagram of an alternative determination of candidate semantic models by target regions according to embodiments of the present application;

FIG. 5 is a flow diagram of selecting a target semantic model by a model selection rule according to an embodiment of the present application;

FIG. 6 is a schematic illustration of an alternative principle interpretation of selecting a target semantic model according to a model similarity algorithm according to an embodiment of the present application;

FIG. 7 is a flow diagram of selecting a target semantic model by a model selection rule according to another alternative embodiment of the present application;

FIG. 8 is a flow diagram of selecting a target semantic model by a model selection rule according to another alternative embodiment of the present application;

FIG. 9 is a schematic flow diagram of an alternative target semantic model processing data to be processed for processing a target region according to an embodiment of the present application;

FIG. 10 is a block diagram of an alternative semantic model fault processing device according to an embodiment of the present application;

fig. 11 is a block diagram of an alternative electronic device according to an embodiment of the present application.

Detailed Description

In order to make the present application solution better understood by those skilled in the art, the following description will be made in detail and with reference to the accompanying drawings in the embodiments of the present application, it is apparent that the described embodiments are only some embodiments of the present application, not all embodiments. All other embodiments, which can be made by one of ordinary skill in the art based on the embodiments herein without making any inventive effort, shall fall within the scope of the present application.

It should be noted that the terms "first," "second," and the like in the description and claims of the present application and the above figures are used for distinguishing between similar objects and not necessarily for describing a particular sequential or chronological order. It is to be understood that the data so used may be interchanged where appropriate such that embodiments of the present application described herein may be implemented in sequences other than those illustrated or otherwise described herein. Furthermore, the terms "comprises," "comprising," and "having," and any variations thereof, are intended to cover a non-exclusive inclusion, such that a process, method, system, article, or apparatus that comprises a list of steps or elements is not necessarily limited to those steps or elements expressly listed but may include other steps or elements not expressly listed or inherent to such process, method, article, or apparatus.

According to one aspect of the embodiments of the present application, a semantic model fault processing method is provided. Alternatively, in the present embodiment, the above-described semantic model failure processing method may be applied to a hardware environment constituted by a terminal and a server. The server is connected with the terminal through a network, can be used for providing services (such as advertisement push service, application service and the like) for the terminal or a client installed on the terminal, and can be used for providing data storage service for the server by setting a database on the server or independent of the server.

The network may include, but is not limited to, at least one of: wired network, wireless network. The wired network may include, but is not limited to, at least one of: a wide area network, a metropolitan area network, a local area network, and the wireless network may include, but is not limited to, at least one of: WIFI (Wireless Fidelity ), bluetooth. The terminal is not limited to a PC, a mobile phone, a tablet computer, and the like.

The semantic model fault processing method in the embodiment of the application can be executed by a server, a terminal or both. The method for processing the semantic model fault by the terminal can also be executed by a client installed on the terminal.

Taking a server to execute the semantic model fault processing method in this embodiment as an example, referring to fig. 1, fig. 1 is a schematic diagram of a hardware environment of an alternative semantic model fault processing method provided in this embodiment, and as shown in fig. 1, the hardware environment of the semantic model fault processing method includes: a terminal 102, and a server 104 connected to the terminal 102 via a network. The server 104 is configured to deploy a plurality of semantic models that provide semantic services for a certain area, and execute the semantic model fault processing method according to the embodiment of the present application when a certain semantic model of the area fails, and process a semantic service request of the area; the terminal 102 is configured to receive a semantic service requirement of a user in the area, and display a processing result of the semantic model for the semantic service requirement, where the processing result may be obtained by processing a semantic model deployed by the server 104 when the semantic model in the area fails; the region semantic model can be obtained by processing an alternative semantic model when the region semantic model fails.

The semantic model fault processing method in the embodiment can be applied to the situation that the semantic model deployment environment in the area is limited, the backup of the semantic model cannot be deployed, and when the semantic model in the area is faulty, the fault is recovered, for example: under the condition that the number of service rooms is limited and only a single region semantic service can be deployed, the fact that the single region semantic service is completed by a plurality of semantic models of the region in a coordinated manner is needed, any model in the region fails, and the semantic service of the region cannot be used.

In this embodiment, a semantic model fault processing method running on the server is provided, please refer to fig. 2, fig. 2 is a flow chart of an alternative semantic model fault processing method according to an embodiment of the present application, and as shown in fig. 2, the semantic model fault processing method of the embodiment of the present application specifically includes the following steps:

step S201, obtaining a fault semantic model of a fault.

Step S202, determining at least one candidate semantic model according to a target area corresponding to the fault semantic model, wherein a preset geographic similarity requirement is met between the candidate area corresponding to the candidate semantic model and the target area.

Step S203, selecting and obtaining a target semantic model from the at least one candidate semantic model through a model selection rule.

And S204, processing the data to be processed of the target area through the target semantic model.

Through the steps S201 to S204, firstly, the candidate area of the target area corresponding to the fault semantic model and the candidate semantic model are determined according to the preset geographic similarity requirement, secondly, the target semantic model with the highest similarity with the fault semantic model is selected from the candidate semantic models through the model selection rule, the data to be processed of the target area is processed by using the target semantic model, the problem that semantic service is unavailable due to the fact that the semantic model cannot be recovered immediately after the fault exists in the related technology is solved, and because the candidate area closest to the target area where the fault semantic model is located can be accurately selected by using the model selection rule, a more efficient and accurate solution is provided for fault repair, the application effect and user experience of the semantic service system in multiple areas are improved, and resources are further saved because the backup semantic model does not need to be deployed.

The following explains the semantic model fault processing method in the embodiment of the present application with reference to fig. 2.

In the technical solution of step S201, a fault semantic model of a fault is obtained.

In the above application scenario of the semantic model fault processing method, it is introduced that, due to the deployment environment limitation, the region cannot deploy the backup semantic model, so the fault semantic model is a semantic model that specifically fails in the semantic model that provides semantic service for the region. In practical application, the semantic model for providing semantic service for a certain area is generally a plurality of different semantic models, wherein each semantic model is matched and cooperated with each other to provide semantic service for users in the area, and if one of the semantic models fails, the semantic service in the area is unavailable.

In practical application, after the semantic model of a region is deployed, the semantic model serving each region takes the region as a node, a plurality of nodes of a plurality of regions form an interconnecting undirected graph, the graph can be displayed in a system for monitoring the semantic model, and when the semantic model of a certain region fails, the corresponding region node in the system can display the failure, and meanwhile, an alarm is triggered to inform related personnel to process. Thus, measures can be taken in time to prevent further expansion of faults. In specific practice, the system will continuously monitor the performance of the semantic model, for example: the execution speed and the execution result of the model discover potential problems in time and optimize. Through performance optimization, the stability and fault tolerance of the system can be improved, and faults are reduced. Performance optimization mainly refers to load balancing and flow rate duty cycle adjustment.

After the fault occurs, the region identifier corresponding to the fault semantic model is obtained, wherein the region identifier is the region ID. According to the semantic model fault processing method, a specific fault semantic model is searched according to the region ID, and the specific searching method is used for obtaining the content of the fault semantic model with faults.

As an alternative embodiment, obtaining a fault semantic model of a fault, please refer to fig. 3, fig. 3 is a schematic flow chart of an alternative obtaining a fault semantic model according to an embodiment of the present application, as shown in fig. 3, and specific steps include:

s11, determining state information and performance information of the semantic model of the target area through a model running log corresponding to the semantic model of the target area, wherein the state information is used for indicating whether the semantic model of the target area runs or not, and the performance information is used for indicating performance of the semantic model of the target area for processing semantic requests;

and S12, when the state information indicates that the semantic model of the target area is running and the performance information does not meet the preset performance requirement, determining the semantic model of the target area as the fault semantic model.

In specific practice, when a specific fault model is acquired, a model running log of each model of the target region, namely a region corresponding to the fault semantic model, is firstly acquired, and the model running log is specifically acquired according to the region ID; secondly, acquiring state information and performance information of the semantic model; determining a fault semantic model according to the state information and the performance information, and specifically: and when the state information indicates that the semantic model of the target area is running and the performance information does not meet the preset performance requirement, determining the semantic model of the target area as a fault semantic model.

It should be noted that, the state information is used to indicate whether the semantic model of the target area is running, the state information includes running and offline, and when the state information indicates that the semantic model of the target area is offline, the model does not belong to a fault or an abnormality. The performance information is used for representing the performance of the semantic model of the target area for processing the semantic request, and the performance information comprises: the semantic model processes the time delay of the semantic request and the numerical value of the semantic request processing success in each second of the semantic model.

The performance information mentioned in the semantic model fault processing method does not meet the requirements specifically includes: the time delay of the semantic request processed by the semantic model is greater than or equal to a preset time delay threshold, and the time delay threshold is set according to actual conditions, for example: 10s, or the value of successful semantic request processing in each second of the semantic model is smaller than a preset throughput threshold, and the preset throughput threshold is set according to actual conditions.

It can be understood that when the semantic model of the target area runs and the time delay of processing the semantic request by the semantic model is greater than or equal to a preset time delay threshold, the semantic model is judged to be a fault semantic model, or when the semantic model of the target area runs and the value of the semantic request successfully processed in each second of the semantic model is smaller than a preset throughput threshold, the semantic model is judged to be a fault semantic model.

In specific practice, after the fault semantic model with the fault is obtained, further verifying whether the semantic model is true, specifically, sending a specific standard semantic processing request to the fault semantic model, and judging whether the semantic model is true according to the request processing result. If the request is not executed correctly, the semantic model determines that a failure occurred; if the request is executed correctly, the semantic model fails.

After the fault semantic model is determined, the following steps S202 to S204 in the semantic model fault processing method of the present application are executed.

In the technical scheme of step S202, at least one candidate semantic model is determined according to a target area corresponding to the fault semantic model, where a preset geographic similarity requirement is satisfied between the candidate area corresponding to the candidate semantic model and the target area.

Semantic boundaries exist between semantic models of different areas, the semantic boundaries are beneficial to improving the recognition capability of the semantic models to regional semantics in the process of semantic model training, but according to the distribution situation of natural semantics, the semantic boundaries are relatively obvious when the geographic position interval of the urban area is larger, and the semantic boundaries are relatively less obvious when the geographic interval of the area is smaller. In practical application, each regional node forms an undirected graph in deployed semantic service, the next node connected with each node in the undirected graph is the node closest to the geographical distance of the undirected graph, each node is connected with the next node to form an edge, and the number of the edges after the two nodes are connected is judged to achieve the acquisition of the candidate region corresponding to the target region.

As an optional embodiment, at least one candidate semantic model is determined according to the target area corresponding to the fault semantic model, referring to fig. 4, fig. 4 is a schematic flow chart of an optional determination of the candidate semantic model through the target area according to an embodiment of the present application, and as shown in fig. 4, specific steps include:

s21, when a preset geographic similarity requirement is a distance threshold value, acquiring the candidate region with the distance meeting the distance threshold value from the target region, and/or when the preset geographic similarity requirement is a language requirement, acquiring the candidate region with the language meeting the language requirement from the target region;

s22, determining a semantic model corresponding to each candidate region as the candidate semantic model.

It should be noted that the geographic similarity is used for selecting a candidate semantic model meeting the requirement, where the geographic similarity may be determined according to a distance between the candidate region and the target region, or may be determined according to a language used by the candidate region and the target region.

As an alternative embodiment, when determining from the distance between the candidate region and the target region, the determination is made by determining whether the distance of the two regions meets the distance threshold requirement. In specific practice, the distance threshold is the number of edges required for two nodes to connect in the undirected graph, for example: and if the distance threshold is 1, connecting the node where the selected candidate area is located with the node where the target area is located, and setting the number of required edges to be 1.

When the preset geographic similarity requirement is a distance threshold value, acquiring the candidate region with the distance between the candidate region and the target region meeting the distance threshold value, namely selecting candidate region nodes with the number of connecting edges with the target region nodes as the distance threshold value; and determining a semantic model corresponding to each candidate region as the candidate semantic model.

In specific practice, determining the region ID of the candidate region according to the region ID of the target region and the distance threshold; obtaining a model identification of a semantic model in the candidate region according to the region ID of the candidate region, wherein the model identification is an index of the semantic model; and forming a list of semantic models by the index of the fault semantic model in the target area and the index of the semantic model in the candidate area. It can be understood that the candidate model corresponding to the index of each candidate model in the list of semantic models is the candidate semantic model.

According to the technical means, the candidate semantic model corresponding to the candidate region is accurately selected by using the distance threshold, so that the selection range of the candidate semantic model can be effectively reduced, and the efficiency of selecting the target semantic model for replacing the fault semantic model is further improved.

As an alternative embodiment, when determining according to the language used by the candidate region and the target region, the determination is made by comparing whether the language used by the two regions meets the language requirement. In particular practice, the language requirements are used to indicate whether the dialects used by the user in the region of the semantic model service are the same or similar.

When the preset geographic similarity requirement is a language requirement, acquiring the candidate region which meets the language requirement with the target region by using the language, namely selecting the candidate region which is the same as or similar to the target region by using the language; and determining a semantic model corresponding to each candidate region as the candidate semantic model.

In specific practice, determining the region ID of the candidate region according to the region ID of the target region and the type of language used; obtaining a model identification of a semantic model in the candidate region according to the region ID of the candidate region, wherein the model identification is an index of the semantic model; and forming a list of semantic models by the index of the fault semantic model in the target area and the index of the semantic model in the candidate area. It can be understood that the candidate model corresponding to the index of each candidate model in the list of semantic models is the candidate semantic model.

According to the technical means, the candidate semantic model corresponding to the candidate region is accurately selected by using the language requirement, so that the selection range of the candidate semantic model can be effectively reduced, and the efficiency of selecting the target semantic model for replacing the fault semantic model is further improved.

As an alternative embodiment, the candidate semantic model may also be determined using both distance thresholds and language requirements.

Specifically, the area ID of the candidate area may be determined according to the area ID of the target area, the distance threshold, and the language type; obtaining a model identification of a semantic model in the candidate region according to the region ID of the candidate region, wherein the model identification is an index of the semantic model; and forming a list of semantic models by the index of the fault semantic model in the target area and the index of the semantic model in the candidate area. It can be understood that the candidate model corresponding to the index of each candidate model in the list of semantic models is the candidate semantic model.

It should be noted that, regarding determining the area ID of the candidate area according to the area ID of the target area, the distance threshold value and the language type, the area ID of the first candidate area may be determined according to the area ID of the target area and the distance threshold value; determining the region ID of a second candidate region from the region IDs of the first candidate regions according to the region IDs of the target regions and the language types, wherein the region ID of the second candidate region is the region ID of the candidate region; the method can also determine the region ID of the first candidate region according to the region ID of the target region and the language type; and determining the region ID of the second candidate region from the region IDs of the first candidate region according to the region ID of the target region and the distance threshold value, wherein the region ID of the second candidate region is the region ID of the candidate region.

According to the technical means, the candidate semantic model corresponding to the candidate region is more accurately selected by using the distance threshold and the language requirement, so that the selection range of the candidate semantic model can be effectively reduced, and the efficiency of selecting the target semantic model for replacing the fault semantic model is further improved.

In the technical solution of step S203, a target semantic model is selected from the at least one candidate semantic model by a model selection rule.

In specific practice, a target semantic model with highest similarity with the fault semantic model is selected through a model selection rule. The model selection rule can be a model similarity algorithm, a query type matching method and a model identification result and model performance selection method. The model similarity algorithm is used for obtaining the model similarity of the candidate semantic model and the fault semantic model according to the similarity of the recognition results of the candidate semantic model and the fault semantic model, and finally selecting the target semantic model with the highest similarity with the fault semantic model. The query type matching method is to classify the data to be identified of the target area in the fault recovery period according to the target query types, select a semantic model with highest identification accuracy in the target query types in the candidate models as a target semantic model, wherein the target query types are operation types used by users, such as: navigation, music, control instructions, etc. The recognition result of the model and the model performance selection method are that the data to be recognized of a target area in a fault recovery period are divided into different groups, each group processes the data by using different candidate semantic models to obtain a processing result and model performance, the data to be recognized can be successfully recognized in the candidate models according to the model performance, the semantic models with better model performance are the target semantic models, the model recognition result comprises success and failure, and the model performance specifically comprises: the candidate semantic model processes the time delay of the semantic request and the numerical value of the success of the semantic request processing in each second of the candidate semantic model.

As an optional embodiment, when the model selection rule is a model similarity algorithm, the selecting, by the model selection rule, a target semantic model is selected from the at least one candidate semantic model, referring to fig. 5, and fig. 5 is a schematic flow chart of selecting, by the model selection rule, the target semantic model according to an optional embodiment of the application, as shown in fig. 5, and specific steps include:

s31, acquiring historical data of the fault semantic model, wherein the historical data comprises: history identification data and a first result obtained by identifying the history identification data;

s32, respectively inputting the history identification data into each candidate semantic model to obtain a second result output by each candidate semantic model;

s33, calculating the result similarity of the first result and each second result through the model similarity algorithm;

s34, determining a model similarity value of each candidate semantic model and the fault semantic model according to all the result similarities;

and S35, determining the candidate semantic model with the maximum model similarity value as a target semantic model in the at least one candidate semantic model.

It should be noted that, the historical data is data identified by the fault semantic model, where the historical data includes: history identification data and a first result obtained by identifying the history identification data. The model similarity algorithm is used for calculating a result similarity value output by the fault semantic model and the candidate semantic model which respectively process the same historical identification data, and the result similarity can be used for representing the model similarity of each candidate semantic model and the fault semantic model, namely the result similarity value is the model similarity value. The first result is the result of the recognition of the historical recognition data by the fault semantic model, and the second result is the result of the recognition of the historical recognition data by the candidate semantic model.

In specific practice, the model similarity algorithm specifically comprises the following steps of: acquiring a list of history identification data, a first result and a semantic model; creating an array for storing the result similarity values; respectively inputting the history identification data into candidate models corresponding to model indexes in a list of semantic models to obtain a second result output by each candidate semantic model; calculating a result similarity value of a second result and a first result corresponding to each candidate semantic model by using a cosine function; storing the result similarity value corresponding to each candidate semantic model into an array; acquiring an index of the fault semantic model according to the first result; removing the result similarity value corresponding to the index of the fault semantic model in the array; and selecting the model index with the maximum similarity value from the rest results in the array, and outputting the model index corresponding to the similarity value, wherein the model corresponding to the model index is the target semantic model.

In order to better explain the principle of selecting the target semantic model according to the model similarity algorithm in this embodiment, the regional model a fault in the urban regional semantic model cluster is illustrated by taking the regional model B, C, D as a candidate model as an example. Referring to fig. 6, fig. 6 is an optional schematic diagram illustrating the principle of selecting a target semantic model according to a model similarity algorithm according to an embodiment of the present application, as shown in fig. 6, after a monitored service monitors a fault of a region model a, a model similarity algorithm is called to calculate model similarity values of the region model a and a region model B, C, D, respectively, according to fig. 3, it can be known that the model similarity value of the region model a and the region model B is 0.88, the model similarity value of the region model a and the region model C is 0.90, and the model similarity value of the region model a and the region model D is 0.76, then a region model C corresponding to 0.90 with the largest model similarity value is selected as the target semantic model, and the region model C is used for replacing the region model a with the fault to provide the semantic service for the region corresponding to the region model a.

According to the technical means, the similarity analysis is carried out on each candidate semantic model and the fault semantic model by adopting the model similarity algorithm, so that the target semantic model for replacing the fault semantic model can be accurately selected. By selecting the model with the highest similarity with the state before the fault, the fluctuation of user experience caused by model switching is reduced.

As an optional embodiment, when the model selection rule is a query type matching method, the selecting, by the model selection rule, a target semantic model from the at least one candidate semantic model, please refer to fig. 7, fig. 7 is a schematic flow chart of selecting, by the model selection rule, the target semantic model according to another optional embodiment of the present application, as shown in fig. 7, and the specific steps include:

s41, acquiring data to be identified of the target area;

s42, determining a target query type corresponding to the target area according to the data to be identified;

s43, acquiring a history query record of each candidate semantic model;

s44, determining the target semantic model with highest accuracy on the target query type from the at least one candidate semantic model according to the target query type and the historical query record.

It should be noted that the data to be identified is unprocessed data acquired by the system during the fault recovery. The target query type refers to a recognition scene type with high use frequency in the data to be recognized of the target area, for example: navigation, music, control instructions, etc. The historical query record refers to a set of recognition results for candidate semantic model historical recognition scenarios, each recognition scenario.

Taking the type of target query in the data to be identified as navigation as an example, the target semantic model selected in the embodiment is the semantic model with the highest identification accuracy in the navigation identification scene in the candidate models.

According to the technical means, the user in the fault recovery period is classified according to the query types, and the candidate semantic model with good performance of the corresponding type is selected as the target semantic model from the candidate semantic models for each type.

As an optional embodiment, when the model selection rule is a recognition result of a model and a model performance selection method, the selecting, by the model selection rule, a target semantic model from the at least one candidate semantic model is selected, please refer to fig. 8, fig. 8 is a schematic flow chart of selecting, by the model selection rule, the target semantic model according to another optional embodiment of the present application, as shown in fig. 8, including the specific steps of:

S51, acquiring data to be identified of the target area;

s52, determining a recognition result and model performance corresponding to the candidate semantic model according to the data to be recognized;

and S53, in the at least one candidate semantic model, the identification result is successful, and the candidate semantic model with the highest model performance is determined to be a target semantic model.

It should be noted that the data to be identified is unprocessed data acquired by the system during the fault recovery. The recognition result refers to the result of the candidate semantic model after recognizing the data to be recognized. Model performance refers to the performance of candidate semantic models in identifying data to be identified, and specifically includes: the candidate semantic model processes the time delay of the semantic request and the numerical value of the success of the semantic request processing in each second of the candidate semantic model.

The target semantic model selected in the embodiment is a semantic model which can successfully identify the data to be identified in the candidate models and has better model performance. Specifically, if the identification result corresponding to the candidate semantic model is successfully displayed and the time delay of the candidate semantic model for processing the semantic request is minimum, the candidate semantic model is the target semantic model; under the same time delay condition, selecting the target semantic model with the largest value of successful semantic request processing in each second as the candidate semantic model.

In specific practice, the target semantic model can be further selected according to feedback in the process of processing the data to be identified by the user on the candidate semantic model or feedback of an identification result on the basis of the conditions. The feedback of the user specifically includes: emotional language of the user, manual operation, etc. If after the processing result is obtained, the user speaks the emotion language satisfactory to the result, the candidate semantic model which provides the semantic service currently can be judged to meet the requirement of the user, and the candidate semantic model can be used for providing the semantic service for the current user continuously. If the user has manual operation in the process of processing the data to be identified by the candidate semantic model, the current candidate semantic model can be considered to be incapable of well processing the requirements of the current user, and other candidate semantic models can be used for continuously providing semantic services for the current user.

According to the technical means, users in a target area are divided into different groups during fault recovery, each group uses different candidate semantic models to carry out semantic service, and candidate semantic models with good model performance and good user feedback are selected as target semantic models by comparing the model performance of each group with user feedback.

In the technical solution of step S204, the data to be processed of the target area is processed through the target semantic model.

Because the semantic service of each region needs to be completed in a co-operation way, one semantic model fails, other semantic models of the region cannot provide semantic service, and the parameter configuration of the semantic model of each region is different and not universal, so that the data to be processed of the target region cannot be processed by directly replacing the failed semantic model after the target semantic model is obtained, and other semantic models in the region corresponding to the target semantic model are required to be matched for use. In this embodiment, the forwarding thread is established to replace the area ID of the target area corresponding to the fault semantic model with the area ID of the candidate area corresponding to the target semantic model, so that the target semantic model processes the data to be processed of the target area.

As an optional embodiment, the target semantic model processes the data to be processed of the target area, referring to fig. 9, fig. 9 is a schematic flow chart of an optional target semantic model processing the data to be processed of the target area according to an embodiment of the present application, as shown in fig. 9, and specific steps include:

s61, establishing a forwarding thread;

s62, forwarding the data to be processed of the target area to the target semantic model through the forwarding thread so that the target semantic model processes the data to be processed of the target area.

It should be noted that, the forwarding thread forwards the data to be processed in the target area to the target semantic model, so that the target semantic model processes the data to be processed.

In specific practice, firstly, an index forwarding list is established, corresponding area IDs of target semantic models are stored in the index forwarding list, and when a semantic service request of a target area is acquired, a forwarding thread is used for replacing the area IDs of the target area in the semantic service request with the area IDs corresponding to the target models, so that the purpose of processing data to be processed by using the target semantic models is achieved.

In specific practice, because one semantic model of a certain area fails, other semantic models of the area cannot provide semantic services for the area, the forwarding thread actually replaces the area ID of the target area in the semantic service request with the area ID corresponding to the target semantic model, so that the target semantic model processes the data to be processed.

In specific practice, after the target semantic model is used for processing the data to be processed of the target area, a system for monitoring the semantic model can start a monitoring program to monitor the semantic fault model, and when the model is recovered to be normal, a monitoring thread is closed, and a forwarding thread is closed.

In specific practice, after the semantic model of the target area fails, backup is needed to be performed on the data of the failed semantic model, and the backup data comprises: user log data and model parameter data of the fault model; the user log data, namely the historical data, can be used for a subsequent replacement model, and the fault parameter data is used for the recovered use of the fault model.

It should be noted that, for simplicity of description, the foregoing method embodiments are all expressed as a series of action combinations, but it should be understood by those skilled in the art that the present application is not limited by the order of actions described, as some steps may be performed in other order or simultaneously in accordance with the present application. Further, those skilled in the art will also appreciate that the embodiments described in the specification are all preferred embodiments, and that the acts and modules referred to are not necessarily required in the present application.

From the description of the above embodiments, it will be clear to a person skilled in the art that the method according to the above embodiments may be implemented by means of software plus the necessary general hardware platform, but of course also by means of hardware, but in many cases the former is a preferred embodiment. Based on such understanding, the technical solution of the present application may be embodied essentially or in a part contributing to the prior art in the form of a software product stored in a storage medium (such as ROM (Read-Only Memory)/RAM (Random Access Memory ), magnetic disk, optical disc), including instructions for causing a terminal device (which may be a mobile phone, a computer, a server, or a network device, etc.) to perform the method described in the embodiments of the present application.

According to another aspect of the embodiment of the application, there is also provided a semantic model fault processing device for implementing the semantic model fault processing method. Referring to fig. 10, fig. 10 is a block diagram illustrating an alternative semantic model fault processing apparatus according to an embodiment of the present application, as shown in fig. 10, the apparatus may include:

A fault semantic model obtaining module 401, configured to obtain a fault semantic model that has a fault;

a candidate semantic model determining module 402, configured to determine at least one candidate semantic model according to a target area corresponding to the fault semantic model, where a candidate area corresponding to the candidate semantic model and the target area meet a preset geographic similarity requirement;

a target semantic model selection module 403, configured to select a target semantic model from the at least one candidate semantic model according to a model selection rule;

and the data processing module 404 is used for processing the data to be processed of the target area through the target semantic model.

It should be noted that, the failure semantic model obtaining module 401 in this embodiment may be used to perform the above step S201, the candidate semantic model determining module 402 in this embodiment may be used to perform the above step S202, the target semantic model selecting module 403 in this embodiment may be used to perform the above step S203, and the data processing module 404 in this embodiment may be used to perform the above step S204.

With respect to the semantic model failure processing apparatus in the present embodiment, the specific manner in which the failure semantic model acquisition module 401, the candidate semantic model determination module 402, the target semantic model selection module 403, and the data processing module 404 execute the above-described semantic model failure processing method has been described in detail in the embodiment regarding the semantic model failure processing method, and will not be described in detail here.

It can be understood that, according to the technical scheme provided by the embodiment, each module in the semantic model fault processing device acquires a fault semantic model with faults; determining at least one candidate semantic model according to a target area corresponding to the fault semantic model, wherein a preset geographic similarity requirement is met between the candidate area corresponding to the candidate semantic model and the target area; selecting a target semantic model from the at least one candidate semantic model through a model selection rule; the target semantic model is used for processing the data to be processed of the target area, so that the problem that semantic service is not available due to the fact that the semantic model cannot be recovered immediately after failure in the related technology is solved, and the target semantic model for replacing the failure semantic model can be accurately selected from a candidate area closest to the target area where the failure semantic model is located by using the model selection rule, so that a more efficient and accurate solution is provided for failure repair, the application effect and user experience of a semantic service system in multiple areas are improved, and resources are further saved due to the fact that backup semantic models do not need to be deployed.

The apparatus in this embodiment may further include, in addition to the above-described modules, modules that execute any of the methods in the foregoing embodiments of any of the semantic model fault processing methods.

It should be noted that the above modules are the same as examples and application scenarios implemented by the corresponding steps, but are not limited to what is disclosed in the above embodiments. It should be noted that the above modules may be implemented as part of an apparatus in a hardware environment implementing the method shown in fig. 1, and may be implemented by software, or may be implemented by hardware, where the hardware environment includes a network environment.

According to still another aspect of the embodiments of the present application, there is further provided an electronic device for implementing the foregoing semantic model fault processing method, where the electronic device may be a server, a terminal, or a combination thereof.

In accordance with another embodiment of the present application, referring to fig. 11, fig. 11 is a block diagram of an alternative electronic device according to an embodiment of the present application, as shown in fig. 11, the electronic device may include: the device comprises a processor 1501, a communication interface 1502, a memory 1503 and a communication bus 1504, wherein the processor 1501, the communication interface 1502 and the memory 1503 are in communication with each other through the communication bus 1504.

A memory 1503 for storing a computer program;

the processor 1501, when executing the program stored in the memory 1503, performs the following steps:

step S201, obtaining a fault semantic model of a fault.

It can be understood that, according to the technical scheme provided by the embodiment, the processor of the electronic device obtains the fault semantic model with faults; determining at least one candidate semantic model according to a target area corresponding to the fault semantic model, wherein a preset geographic similarity requirement is met between the candidate area corresponding to the candidate semantic model and the target area; selecting a target semantic model from the at least one candidate semantic model through a model selection rule; the target semantic model is used for processing the data to be processed of the target area, so that the problem that semantic service is not available due to the fact that the semantic model cannot be recovered immediately after failure in the related technology is solved, and the target semantic model for replacing the failure semantic model can be accurately selected from a candidate area closest to the target area where the failure semantic model is located by using the model selection rule, so that a more efficient and accurate solution is provided for failure repair, the application effect and user experience of a semantic service system in multiple areas are improved, and resources are further saved due to the fact that backup semantic models do not need to be deployed.

Alternatively, in the present embodiment, the above-described communication bus may be a PCI (Peripheral Component Interconnect, peripheral component interconnect standard) bus or an EISA (Extended Industry Standard Architecture ) bus or the like. The communication bus may be classified as an address bus, a data bus, a control bus, or the like. For ease of illustration, the figures are shown with only one bold line, but not with only one bus or one type of bus. The communication interface is used for communication between the electronic device and other devices.

The Memory may include random access Memory (Random Access Memory, RAM) or may include Non-Volatile Memory (NVM), such as at least one disk Memory. Optionally, the memory may also be at least one memory device located remotely from the aforementioned processor.

The processor may be a general purpose processor and may include, but is not limited to: CPU (Central Processing Unit ), NP (Network Processor, network processor), etc.; but also DSP (Digital Signal Processor ), ASIC (Application Specific Integrated Circuit, application specific integrated circuit), FPGA (Field-Programmable Gate Array, field programmable gate array) or other programmable logic device, discrete gate or transistor logic device, discrete hardware components.

The embodiment of the application also provides a computer readable storage medium, wherein the storage medium comprises a stored program, and the program executes the method steps of the method embodiment.

Alternatively, in the present embodiment, the storage medium may include, but is not limited to: various media capable of storing program codes, such as a U disk, ROM, RAM, a mobile hard disk, a magnetic disk or an optical disk.

The foregoing embodiment numbers of the present application are merely for describing, and do not represent advantages or disadvantages of the embodiments.

The integrated units in the above embodiments may be stored in the above-described computer-readable storage medium if implemented in the form of software functional units and sold or used as separate products. Based on such understanding, the technical solution of the present application may be embodied in essence or a part contributing to the prior art or all or part of the technical solution in the form of a software product stored in a storage medium, including several instructions to cause one or more computer devices (which may be personal computers, servers or network devices, etc.) to perform all or part of the steps of the methods described in the various embodiments of the present application.

In the foregoing embodiments of the present application, the descriptions of the embodiments are emphasized, and for a portion of this disclosure that is not described in detail in this embodiment, reference is made to the related descriptions of other embodiments.

In several embodiments provided in the present application, it should be understood that the disclosed client may be implemented in other manners. The above-described embodiments of the apparatus are merely exemplary, and the division of the units, such as the division of the units, is merely a logical function division, and may be implemented in another manner, for example, multiple units or components may be combined or may be integrated into another system, or some features may be omitted, or not performed. Alternatively, the coupling or direct coupling or communication connection shown or discussed with each other may be through some interfaces, units or modules, or may be in electrical or other forms.

The units described as separate units may or may not be physically separate, and units shown as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution provided in the present embodiment.

In addition, each functional unit in each embodiment of the present application may be integrated in one processing unit, or each unit may exist alone physically, or two or more units may be integrated in one unit. The integrated units may be implemented in hardware or in software functional units.

The foregoing is merely a preferred embodiment of the present application and it should be noted that modifications and adaptations to those skilled in the art may be made without departing from the principles of the present application and are intended to be comprehended within the scope of the present application.

Claims

1. A semantic model fault handling method, comprising:

acquiring a fault semantic model with faults;

determining at least one candidate semantic model according to a target area corresponding to the fault semantic model, wherein a preset geographic similarity requirement is met between the candidate area corresponding to the candidate semantic model and the target area;

selecting a target semantic model from the at least one candidate semantic model through a model selection rule;

and processing the data to be processed of the target area through the target semantic model.

2. The semantic model fault processing method according to claim 1, wherein determining at least one candidate semantic model according to the target region corresponding to the fault semantic model comprises:

when the preset geographic similarity requirement is a distance threshold value, acquiring the candidate region of which the distance with the target region meets the distance threshold value;

and/or the number of the groups of groups,

when the preset geographic similarity requirement is a language requirement, acquiring the candidate region which meets the language requirement with the language between the candidate region and the target region;

and determining a semantic model corresponding to each candidate region as the candidate semantic model.

3. The semantic model fault processing method according to claim 1, wherein the selecting, by a model selection rule, a target semantic model from the at least one candidate semantic model includes:

obtaining historical data of the fault semantic model, wherein the historical data comprises: history identification data and a first result obtained by identifying the history identification data;

respectively inputting the history identification data into each candidate semantic model to obtain a second result output by each candidate semantic model;

Calculating the result similarity of the first result and each second result through the model similarity algorithm;

determining a model similarity value of each candidate semantic model and the fault semantic model according to all the result similarities;

and determining the candidate semantic model with the maximum model similarity value as a target semantic model in the at least one candidate semantic model.

4. The semantic model fault processing method according to claim 1, wherein the selecting, by a model selection rule, a target semantic model from the at least one candidate semantic model includes:

acquiring data to be identified of the target area;

determining a target query type corresponding to the target area according to the data to be identified;

acquiring a historical query record of each candidate semantic model;

and determining the target semantic model with highest accuracy on the target query type from the at least one candidate semantic model according to the target query type and the historical query record.

5. The semantic model fault processing method according to claim 1, wherein the selecting, by a model selection rule, a target semantic model from the at least one candidate semantic model includes:

Acquiring data to be identified of the target area;

determining a recognition result and model performance corresponding to the candidate semantic model according to the data to be recognized;

and in the at least one candidate semantic model, the identification result is successful, and the candidate semantic model with the highest model performance is determined to be a target semantic model.

6. The semantic model failure processing method according to claim 1, wherein processing the data to be processed of the target region by the target semantic model includes:

establishing a forwarding thread;

and forwarding the data to be processed of the target area to the target semantic model through the forwarding thread so that the target semantic model processes the data to be processed of the target area.

7. The semantic model fault processing method according to claim 1, wherein the obtaining a fault semantic model of a fault comprises:

determining state information and performance information of a semantic model of the target area through a model running log corresponding to the semantic model of the target area, wherein the state information is used for indicating whether the semantic model of the target area runs or not, and the performance information is used for indicating the performance of the semantic model of the target area for processing semantic requests;

And when the state information indicates that the semantic model of the target area is running and the performance information does not meet the preset performance requirement, determining the semantic model of the target area as the fault semantic model.

8. A semantic model fault processing apparatus, comprising:

the fault semantic model acquisition module is used for acquiring a fault semantic model with faults;

the candidate semantic model determining module is used for determining at least one candidate semantic model according to a target area corresponding to the fault semantic model, wherein a preset geographic similarity requirement is met between the candidate area corresponding to the candidate semantic model and the target area;

the target semantic model selection module is used for selecting and obtaining a target semantic model from the at least one candidate semantic model through a model selection rule;

and the data processing module is used for processing the data to be processed of the target area through the target semantic model.

9. An electronic device comprising a processor, a communication interface, a memory and a communication bus, wherein the processor, the communication interface and the memory communicate with each other via the communication bus, characterized in that,

The memory is used for storing a computer program;

the processor for executing the semantic model fault processing method according to any one of claims 1 to 7 by running the computer program stored on the memory.

10. A computer-readable storage medium, characterized in that the storage medium has stored therein a computer program, wherein the computer program is arranged to execute the semantic model fault processing method according to any of claims 1 to 7 at run-time.