Semantic Processing Method, System, Device and Storage Medium Based on Edge Nodes

Abstract

The present invention discloses a semantic processing method, system, device and storage medium based on edge nodes. The method is executed by the edge nodes. First, the edge nodes obtain the corpus to be processed sent by the terminal, and perform industry feature matching on the corpus to be processed according to the industry knowledge base and the scenario corpus located at the edge nodes to determine the corpus scenario corresponding to the corpus to be processed; according to the corpus scenario, select the corresponding industry language model to process the corpus to be processed, generate the first semantic result corresponding to the corpus to be processed, and the edge nodes send the first semantic result back to the terminal to complete the current semantic processing. The embodiments of the present application propose to process the corpus to be processed through the industry language model, which helps to improve the fitting degree between the semantic processing result and the industry; in addition, the semantic processing process of the embodiments of the present application is mainly completed at the edge nodes, which helps to meet the security and privacy requirements of users and has a positive effect on the popularization of semantic processing technology.

Description

Technical Field

[0001] This application relates to the technical field of semantic processing, and in particular, to a semantic processing method, system, device, and storage medium based on edge nodes. Background Art

[0002] With the continuous development of artificial intelligence technology, semantic processing technology based on semantic processing and understanding has also developed rapidly. Taking a human-computer interaction device that applies semantic understanding as an example, through semantic processing, people can use a more natural language to more conveniently complete the interaction between humans and machines, thereby achieving the purpose of reducing the operation threshold of human-computer interaction devices and improving the efficiency of various tasks.

[0003] However, the general semantic understanding services in related technologies cannot well adapt to various specific business scenarios. For example, there are a large number of specialized vocabulary in industries such as medical and finance. Another example is that the vocabulary in some new media industries on the Internet is updated very quickly. The general semantic understanding services in related technologies are difficult to meet the scene recognition requirements of specific businesses. In addition, there are certain security and privacy requirements for information data assets within each industry, and these requirements are also difficult to meet by general semantic processing services. Summary of the Invention

[0004] This application aims to at least solve one of the technical problems in related technologies to some extent. For this purpose, this application proposes a semantic processing method, system, device, and storage medium based on edge nodes.

[0005] In a first aspect, an embodiment of this application provides a semantic processing method based on edge nodes. The method is executed by an edge node in a semantic processing system based on edge nodes. The semantic processing system based on edge nodes includes an edge node and a terminal. The method includes: obtaining a corpus to be processed; performing industry feature matching on the corpus to be processed according to an industry knowledge base and a scenario corpus located at the edge node to determine a corpus scenario; determining an industry language model according to the corpus scenario; determining a first semantic result according to the industry language model and the corpus to be processed; and returning the first semantic result to the terminal.

[0006] Optionally, the semantic processing system based on edge nodes further includes a cloud. The method further includes: calculating a first confidence level of the first semantic result according to the industry knowledge base and the scenario corpus; when the first confidence level is lower than a preset confidence level threshold, sending a collaborative processing request to the cloud so that the cloud obtains supplementary corpus and returns the supplementary corpus to the edge node; determining a plurality of second semantic results according to the supplementary corpus and the first semantic result; calculating a second confidence level of the second semantic result according to the industry knowledge base and the scenario corpus; and returning the second semantic result with the highest second confidence level to the terminal.

[0007] Optionally, the method further includes: adding the supplementary corpus to the industry knowledge base and the scenario corpus.

[0008] Optionally, enabling the cloud to obtain the supplementary corpus includes: according to a preset retrieval condition, enabling the cloud to retrieve the supplementary corpus in the Internet; wherein, the retrieval condition includes being the same as the pronunciation of the first semantic result.

[0009] Optionally, the method further includes the construction process of the industry language model, specifically including: randomly extracting N types of samples from the obtained corpus samples as the first sample set, and the first sample set contains N types of first samples; wherein, the sample categories in the first sample set include scenario results and target topics; in each type of the first samples, extracting K instances as the first instance set, and the first instance set contains K first instances; wherein, the instance is a feature word; taking all the extracted first instances as the support set, and taking all the instances in the first sample set except the first instances as the query set; wherein, the support set is used for model training, and the query set is used for model testing; using the support set and the query set to train and test the industry language model; during the training process, adopting the method of gradient weight increase to gradually increase the weight of the labeled instances in the first instances; when the number of training times reaches a preset first quantity, completing the construction of the industry language model; wherein, both N and M are positive integers.

[0010] In a second aspect, an embodiment of the present application provides a semantic processing system based on an edge node. The device is applied to an edge node in the semantic processing system based on an edge node. The semantic processing system based on an edge node includes an edge node and a terminal. The device includes: a first module, a second module, a third module, a fourth module, and a fifth module; the first module is used to obtain the corpus to be processed; the second module is used to perform industry feature matching on the corpus to be processed to determine the corpus scenario; the third module is used to determine the industry language model according to the corpus scenario; the fourth module is used to determine the first semantic result according to the industry language model and the corpus to be processed; the fifth module is used to return the first semantic result to the terminal.

[0011] Optionally, the edge node-based semantic processing system further includes a cloud, and the device further includes: a sixth module, a seventh module, an eighth module, a ninth module, and a tenth module; the sixth module is configured to calculate a first confidence level of the first semantic result according to an industry knowledge base and a scenario corpus located at the edge node; the seventh module is configured to obtain supplementary corpus from the cloud when the first confidence level is lower than a preset confidence level threshold; the eighth module is configured to determine a plurality of second semantic results according to the supplementary corpus and the first semantic result; the ninth module is configured to calculate a second confidence level of the second semantic result according to the industry knowledge base and the scenario corpus; the tenth module is configured to return the second semantic result with the highest second confidence level to the terminal.

[0012] In a third aspect, an embodiment of the present application provides a device, including: at least one processor; at least one memory for storing at least one program; when the at least one program is executed by the at least one processor, the at least one processor is caused to implement the edge node-based semantic processing method as described in the first aspect.

[0013] In a fourth aspect, an embodiment of the present application provides a computer storage medium, in which a processor-executable program is stored, and the processor-executable program is used to implement the edge node-based semantic processing method as described in the first aspect when executed by the processor.

[0014] The beneficial effects of the embodiments of the present application are as follows: This method is executed by an edge node. First, the edge node obtains the corpus to be processed sent by the terminal, and performs industry feature matching on the corpus to be processed according to the industry knowledge base and the scenario corpus located at the edge node to determine the corpus scenario corresponding to the corpus to be processed; according to the corpus scenario, a corresponding industry language model is selected to process the corpus to be processed, and a first semantic result corresponding to the corpus to be processed is generated. The edge node sends the first semantic result back to the terminal to complete the current semantic processing. The embodiments of the present application propose to process the corpus to be processed through an industry language model, which helps to improve the fitting degree of the semantic processing result with the corresponding industry and corresponding business scenario; in addition, the semantic processing process of the embodiments of the present application is mainly completed at the edge node, and the design of the edge node helps to meet the security and privacy requirements of users, and has a positive effect on the popularization of semantic processing technology. Description of the Drawings

[0015] The drawings are used to provide a further understanding of the technical solutions of the present application, and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the technical solutions of the present application, and do not constitute a limitation to the technical solutions of the present application.

[0016] Figure 1It is a schematic diagram of a semantic processing system based on edge nodes provided by an embodiment of the present application;

[0017] Figure 2 It is a flowchart of the steps of a semantic processing method based on edge nodes provided by an embodiment of the present application;

[0018] Figure 3 It is the correspondence relationship between the scenario results, feature words, and target topics provided by an embodiment of the present application;

[0019] Figure 4 It is a flowchart of the steps for updating semantic results according to third-party resources provided by an embodiment of the present application;

[0020] Figure 5 It is a schematic diagram of a semantic processing system based on edge nodes provided by an embodiment of the present application;

[0021] Figure 6 It is a schematic diagram of a device provided by an embodiment of the present application. Detailed implementation manners

[0022] In order to make the objectives, technical solutions, and advantages of the present application clearer and more understandable, the present application will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0023] It should be noted that although functional module division is performed in the system schematic diagram and the logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order from the module division in the system or the order in the flowchart. Terms such as "first" and "second" in the description, claims, and the above-mentioned drawings are used to distinguish similar objects and do not necessarily need to describe a specific order or sequence.

[0024] The following further elaborates on the embodiments of the present application with reference to the accompanying drawings.

[0025] Refer to Figure 1 , Figure 1 It is a schematic diagram of a semantic processing system based on edge nodes provided by an embodiment of the present application, as shown in Figure 1As shown, the system 100 includes a terminal 110 and an edge node 120. In the embodiments of the present application, the terminal can be any electronic device capable of submitting a semantic processing request to the edge node, such as a mobile phone, a smart phone, a personal digital assistant (PDA), a wearable device, a pocket PC (PPC), a tablet computer, etc. It can be understood that the terminal can perform human-computer interaction through one or more of a keyboard, a touchpad, a touch screen, a remote control, voice interaction, or a handwriting device. The terminal device can submit a semantic processing request to the edge node and receive the semantic processing result returned by the edge node.

[0026] The edge node can be any electronic device capable of performing semantic processing, such as a mobile phone, a smart phone, a personal digital assistant (PDA), a wearable device, a pocket PC (PPC), a tablet computer, etc. The edge node can receive a semantic processing request from the terminal, perform semantic processing services, and return the semantic processing result to the terminal.

[0027] In some other embodiments, as Figure 1 shown, the semantic processing system based on the edge node proposed in the embodiments of the present application further includes a cloud 130. In the embodiments of the present application, the cloud is a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, CDN (Content Delivery Network), and big data and artificial intelligence platforms. The cloud can receive a collaborative processing request from the edge node, execute the corresponding network search service, and return supplementary corpus to the edge node.

[0028] Through the semantic processing system based on the edge node as Figure 1 shown, the semantic processing method based on the edge node proposed in the embodiments of the present application can be implemented. The specific implementation process of this method will be described in the following content.

[0029] Referring to Figure 2 , Figure 2 is the step flowchart of the semantic processing method based on the edge node provided in the embodiments of the present application. This method is executed by the edge node 120 in the semantic processing system based on the edge node as Figure 1 shown. This method includes but is not limited to steps S200 - S240:

[0030] S200, Obtain the corpus to be processed;

[0031] Specifically, the edge node obtains the corpus to be processed that needs semantic processing, and the corpus to be processed is sent by the terminal 110 in Figure 1 . The edge node performs semantic processing on the corpus to be processed according to the semantic processing request of the terminal. In the embodiment of the present application, the corpus to be processed is the corpus within the same industry, and the forms of the corpus include but are not limited to user conversations, common instructions, industry papers, industry-related news, etc. Using the corpus within the industry as the training corpus for semantic processing can improve the fitting degree of the semantic processing service with this industry and help improve the accuracy of the semantic processing method based on the edge node.

[0032] S210. According to the industry knowledge base and the scenario corpus located in the edge node, perform industry feature matching on the corpus to be processed to determine the corpus scenario;

[0033] Specifically, different industries generally have different proper nouns. During the semantic analysis process, it is necessary to rely on the explanations of these nouns to assist in semantic analysis. Therefore, the embodiment of the present application proposes an industry knowledge base stored in the edge node, and the content in this industry knowledge base is generally the explanations of proper nouns within the industry. For example, in the transportation industry, the industry knowledge base can store the explanations of nouns such as "morning rush hour", "accident-prone section", "blind spot", etc. When performing semantic analysis on a sentence, not only the specific meanings of some industry proper nouns in the sentence need to be clarified, but also the intention of the sentence needs to be clarified through the relationships between the words in the sentence, etc. Therefore, the embodiment of the present application also proposes a scenario corpus stored in the edge node, and through this scenario corpus, the specific meaning of the sentence can be analyzed. For example, in the field of vehicle control, when the input sentence is: "Open the driver's seat window", then according to the industry knowledge base proposed in the embodiment of the present application, the specific meanings of "driver's seat" and "window" can be determined; and through the scenario corpus, it can be analyzed and determined that the intention of this sentence should be to open the specified window, and then the action that the vehicle needs to perform next can be determined.

[0034] In this step, the obtained corpus to be processed is subjected to industry feature matching with the industry knowledge base and the scenario corpus, so as to determine which industry the current corpus to be processed belongs to. It can be understood that even within the same industry, the corpus features for different services may be different. Therefore, the corpus within the industry can be further segmented according to different business scenarios. In this step, the industry feature matching can also identify which business scenario within the industry the corpus to be processed belongs to, so as to determine the corpus scenario corresponding to the current corpus to be processed.

[0035] In some embodiments, the industry feature matching can specifically be to first label the key words or so-called keyword in the industry knowledge base, match the words in the to-be-processed corpus with these keywords. Among different industry knowledge bases, the industry knowledge base with the most keywords matching the to-be-processed corpus can represent the industry corresponding to the to-be-processed corpus.

[0036] In some other embodiments, the corpus overlap degree of the industry knowledge bases of similar industries may also be relatively high. That is to say, when performing industry feature matching, the distinguishability of similar industries may be relatively low, and it is difficult to match the correct industry for the to-be-processed corpus. Then, when performing industry matching, the word frequency of the same keyword in different industry knowledge bases can also be comprehensively considered. When the number of keywords matched by the to-be-processed corpus in two industry knowledge bases is similar, it can be considered to determine the industry where the to-be-processed corpus is located by comparing the word frequencies of the keywords in different industry knowledge bases.

[0037] S220. Determine an industry language model according to the corpus scenario;

[0038] Specifically, in the edge node of the embodiment of the present application, language models corresponding to multiple industries are stored. Therefore, according to the corpus scenario determined in step S210, the corresponding industry language model can be determined. The industry language model adds industry feature language attributes on the basis of the general language model; the general language model generally defines the probability distribution of the token sequence in natural language, and the token can generally be a word, a character or a byte, etc. This language model is used to identify the corpus and generate a semantic processing result. This industry language model is a lightweight model set in the edge node and can be trained with relatively few corpus samples. The training method of this industry language model will be described in the following content.

[0039] In the embodiment of the present application, the industry language model needs to be trained. First, considering that the corpus of some industries is relatively scarce, the method of few shot learning is adopted for training. Specifically, a large number of corpus samples of this industry are obtained, and N types of samples are randomly selected from these corpus samples as the first sample set, and the first sample set contains N types of first samples; among them, the sample categories in the first sample set include scenario results and target topics; in each type of first sample, K instances are selected as the first instance set, and the first instance set contains K first instances; among them, the instance is a feature word; all the extracted first instances are used as the support set, and all the instances in the first sample set except the first instances are used as the query set; among them, the support set is used for model training, and the query set is used for model testing; the industry language model is trained and tested by using the support set and the query set. Refer to Figure 3 , Figure 3 For the corresponding relationship between the scenario result, feature word and target topic provided by the embodiment of the present application, such asFigure 3 As shown, this correspondence is non-linear, where M is the number of correspondences and N is the number of feature words. During the training process, the method of increasing the gradient weight is used to gradually increase the weight of the labeled instances in the first instance. The labeled instances are a part of the first instance that are manually labeled, that is, manually increase Figure 3 the weight of some of the labeled feature words in it. The basis for increasing the weight can be self-defined content such as the word frequency of the feature word, the word sequence, etc. For example, in scenario-based semantic recognition, if it is considered that the labeling role of the feature word is relatively significant, then the feature words that are crucial for scenario-based recognition are given a fixed high weight, and the feature words with secondary roles are given a lower weight, and so on. Through the method of increasing the gradient weight, the goal of enabling the industry language model to support personalized scenario semantic understanding is achieved. When the number of training times reaches a preset first quantity, the construction of the industry language model is completed.

[0040] S230. Determine the first semantic result according to the industry language model and the corpus to be processed;

[0041] Specifically, according to the industry language model determined in the above step S220, the corpus to be processed is processed, that is, the corpus to be processed is input into the industry language model, and the industry language model outputs the corresponding semantic processing result, which is called the first semantic result. The specific content of the first semantic result includes but is not limited to the request type and the request result.

[0042] S240. Return the first semantic result to the terminal;

[0043] Specifically, the edge node returns the first semantic result to the terminal, and the terminal that obtains the first semantic result can execute the corresponding service according to this semantic result, thus completing the entire semantic processing process from the terminal to the edge node and then from the edge node to the terminal.

[0044] Refer to Figure 1 , the embodiment of the present application provides a semantic processing system based on an edge node. This system includes a terminal and an edge node. Through steps S200 - S240, the embodiment of the present application provides a semantic processing method based on an edge node. This method is performed by Figure 1It is executed by the edge node shown. First, the edge node obtains the corpus to be processed sent by the terminal, and based on the industry knowledge base and the scenario-based corpus located at the edge node, performs industry feature matching on the corpus to be processed to determine the corpus scenario corresponding to the corpus to be processed. According to the corpus scenario, the corresponding industry language model is selected to process the corpus to be processed, generating a first semantic result corresponding to the corpus to be processed. The edge node sends this first semantic result back to the terminal to complete this semantic processing. The embodiments of the present application propose to process the corpus to be processed through an industry language model, which helps to improve the fitting degree of the semantic processing result with the corresponding industry and corresponding business scenario. In addition, the semantic processing process of the embodiments of the present application is mainly completed at the edge node, and the design of the edge node helps to meet the security and privacy requirements of users, and has a positive effect on the popularization of semantic processing technology.

[0045] In some embodiments, the semantic processing method based on the edge node proposed in the embodiments of the present application further includes the step of updating the semantic result according to third-party resources, referring to Figure 4 , Figure 4 is the flowchart of the step of updating the semantic result according to third-party resources provided by the embodiments of the present application. The method includes but is not limited to steps S400 - S440:

[0046] S400: Calculate the first confidence level of the first semantic result according to the industry knowledge base and the scenario-based corpus;

[0047] Specifically, according to the industry knowledge base and the scenario-based corpus corresponding to the corpus to be processed, calculate the first confidence level of the first semantic result. The first confidence level is used to characterize the fitting degree of the first semantic result with the corresponding industry and corresponding business scenario. According to the first confidence level, the reliability of the first semantic result can be reflected.

[0048] S410: When the first confidence level is lower than the preset confidence threshold, initiate a collaborative processing request to the cloud so that the cloud obtains supplementary corpus and returns the supplementary corpus to the edge node;

[0049] Specifically, in step S400 above, the first confidence level of the first semantic result is calculated, and this first confidence level is compared with the pre-set confidence threshold. The confidence threshold is used to characterize the lowest confidence level required for the reliability of the first semantic result. Then it can be understood that if the first confidence level is higher than or equal to the preset confidence threshold, it means that the current first semantic result is relatively reliable, and the edge node can directly return this first semantic result to the terminal. On the contrary, if the first confidence level is lower than the preset confidence threshold, it means that the current first semantic result is not reliable enough, or rather, it can be considered that the current industry language model has no corresponding processing result.

[0050] If it is determined that the current first semantic result is not reliable according to the first confidence level, the edge node will initiate a collaborative processing request to the cloud and receive the supplementary corpus sent back by the cloud. Since the storage capacity of the industry knowledge base and the scenario corpus in the edge node is indeed limited, in the embodiments of the present application, the supplementary corpus refers to the corpus within the same industry other than the content recorded in the industry knowledge base and the scenario corpus.

[0051] The specific way for the cloud to obtain the supplementary corpus can be that the cloud retrieves in the industry knowledge base and scenario corpus including but not limited to third parties, or in the social network, and different search engines can also be switched during the retrieval to obtain more comprehensive retrieval results. When retrieving, the preset retrieval condition can be the content with the same pronunciation as the first semantic result. For example, retrieval can be performed through the entire pinyin unit of the word or the first letter of the word pronunciation. In some other embodiments, retrieval can also be performed on the content similar to the first semantic result based on information such as co-occurrence of words. The embodiments of the present application do not specifically limit the retrieval method and retrieval path of the cloud. What the present application wants to illustrate is that when there is no processing result corresponding to the first semantic result in the industry knowledge base and scenario corpus in the edge node, the scope of industry knowledge can be expanded through the collaborative method of the cloud to obtain the supplementary corpus that can supplement the industry knowledge base and scenario corpus.

[0052] S420. Determine a number of second semantic results according to the supplementary corpus and the first semantic result;

[0053] Specifically, the first semantic result is supplemented with the supplementary corpus sent back by the cloud. The form of supplementation can be to select several key texts in the supplementary corpus and add them to the first semantic result, so as to generate several second semantic results on the basis of the first semantic result.

[0054] It can be understood that if different numbers or different key texts are selected from the supplementary corpus and added to the first semantic result, different second semantic results can be obtained. Therefore, for the same supplementary corpus and the same first semantic result, multiple second semantic results may be obtained.

[0055] S430. Calculate the second confidence level of the second semantic result according to the industry knowledge base and the scenario corpus;

[0056] Specifically, similar to the above step S400, calculate the second confidence levels corresponding to the several second semantic results obtained in step S420. Similar to the first confidence level, the second confidence level is used to represent the fitting degree of the second semantic result with the corresponding industry and corresponding business scenario. According to the second confidence level, the reliability of the second semantic result can be reflected.

[0057] S440. Return the second semantic result with the second highest confidence level to the terminal;

[0058] Specifically, select the one with the highest value among the several second confidence levels calculated in step S430. The second semantic result corresponding to the highest second confidence level can be considered the most reliable semantic processing result. Therefore, the edge node uses this second semantic result as the result of this semantic processing process and returns it to the terminal.

[0059] In some embodiments, the supplementary corpus obtained in step S410 above can be supplemented into the corresponding industry knowledge base and scenario-based corpus. After each new supplementary corpus is added, the industry language model can be iteratively updated according to the updated industry knowledge base and scenario-based corpus, so that the industry language model can keep up with the corpus update speed of the corresponding industry as much as possible, enabling the industry language model to better fit the corresponding industry and business scenario and obtaining a semantic processing result with a higher confidence level.

[0060] Through steps S400 - S440, the embodiments of the present application provide a solution for updating semantic results based on third-party resources. For industries with relatively fast industry corpus updates, the embodiments of the present application can timely update the industry knowledge base and scenario-based corpus according to third-party resources, and use the updated industry knowledge base and scenario-based corpus to update the industry language model, so that the industry language model always has a high degree of fit with the corpus of this industry, which helps to improve the accuracy of semantic recognition and has a positive effect on semantic processing services in different industry scenarios.

[0061] Through one or more of the above embodiments, the embodiments of the present application propose a semantic understanding method executed by an edge node. First, the edge node obtains the corpus to be processed sent by the terminal, and performs industry feature matching on the corpus to be processed according to the industry knowledge base and scenario-based corpus located at the edge node to determine the corpus scenario corresponding to the corpus to be processed; according to the corpus scenario, select the corresponding industry language model to process the corpus to be processed and generate a first semantic result corresponding to the corpus to be processed. The edge node sends this first semantic result back to the terminal to complete this semantic processing. The embodiments of the present application propose to process the corpus to be processed through an industry language model, which helps to improve the degree of fit between the semantic processing result and the corresponding industry and business scenario; in addition, the semantic processing process of the embodiments of the present application is mainly completed at the edge node, and the design of the edge node helps to meet the security and privacy requirements of users and has a positive effect on the popularization of semantic processing technology.

[0062] Refer to Figure 5 , Figure 5The figure is a schematic diagram of a semantic processing system based on edge nodes provided by an embodiment of the present application. The device is applied to an edge node in the semantic processing system based on edge nodes. The device 500 includes a first module 510, a second module 520, a third module 530, a fourth module 540, and a fifth module 550. The first module is used to obtain the corpus to be processed. The second module is used to perform industry feature matching on the corpus to be processed and determine the corpus scenario. The third module is used to determine the industry language model according to the corpus scenario. The fourth module is used to determine the first semantic result according to the industry language model and the corpus to be processed. The fifth module is used to return the first semantic result to the terminal.

[0063] In some other embodiments, the semantic processing system based on edge nodes provided by the embodiments of the present application further includes a sixth module, a seventh module, an eighth module, a ninth module, and a tenth module. The sixth module is used to calculate the first confidence level of the first semantic result according to the industry knowledge base and the scenario corpus located at the edge node. The seventh module is used to obtain supplementary corpus from the cloud when the first confidence level is lower than a preset confidence level threshold. The eighth module is used to determine a number of second semantic results according to the supplementary corpus and the first semantic result. The ninth module is used to calculate the second confidence level of the second semantic result according to the industry knowledge base and the scenario corpus. The tenth module is used to return the second semantic result with the highest second confidence level to the terminal.

[0064] Reference Figure 6 , Figure 6 The figure is a schematic diagram of a device provided by an embodiment of the present application. The device 600 includes at least one processor 610 and further includes at least one memory 620 for storing at least one program. Figure 6 In the example, one processor and one memory are used.

[0065] The processor and the memory can be connected by a bus or other means. Figure 6 In the example, connection by a bus is used.

[0066] As a non-transitory computer-readable storage medium, the memory can be used to store non-transitory software programs and non-transitory computer-executable programs. In addition, the memory can include high-speed random access memory, and can also include non-transitory memory, such as at least one magnetic disk storage device, a flash memory device, or other non-transitory solid-state storage devices. In some embodiments, the memory can optionally include a memory remotely set relative to the processor, and these remote memories can be connected to the device through a network. Examples of the above networks include, but are not limited to, the Internet, an enterprise intranet, a local area network, a mobile communication network, and combinations thereof.

[0067] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separated, that is, they may be located in one place or distributed to multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0068] The embodiments of the present application also disclose a computer storage medium, which stores a program executable by a processor. The program executable by the processor is used to implement the semantic processing method based on edge nodes proposed by the present application when executed by the processor.

[0069] Those of ordinary skill in the art can understand that all or some of the steps and systems in the methods disclosed above can be implemented as software, firmware, hardware, and their appropriate combinations. Some physical components or all physical components can be implemented as software executed by a processor, such as a central processing unit, a digital signal processor, or a microprocessor, or implemented as hardware, or implemented as an integrated circuit, such as an application-specific integrated circuit. Such software can be distributed on a computer-readable medium, which can include a computer storage medium (or non-transitory medium) and a communication medium (or transitory medium). As is well known to those of ordinary skill in the art, the term computer storage medium includes volatile and non-volatile, removable and non-removable media implemented in any method or technology for storing information, such as computer-readable instructions, data structures, program modules, or other data. Computer storage media includes, but is not limited to, RAM, ROM, EEPROM, flash memory, or other memory technologies, CD-ROM, digital versatile disk (DVD), or other optical disk storage, magnetic cassette, tape, magnetic disk storage, or other magnetic storage devices, or any other medium that can be used to store the desired information and can be accessed by a computer. In addition, as is well known to those of ordinary skill in the art, a communication medium typically includes computer-readable instructions, data structures, program modules, or other data in a modulated data signal such as a carrier wave or other transmission mechanism, and can include any information delivery medium.

[0070] The above is a specific description of the preferred embodiments of the present application, but the present application is not limited to the above embodiments. Those skilled in the art can make various equivalent deformations or substitutions without departing from the spirit of the present application, and these equivalent deformations or substitutions are all included within the scope defined by the claims of the present application.

Claims

1. A semantic processing method based on edge nodes, characterized in that, The method is executed by an edge node in a semantic processing system based on edge nodes, and the semantic processing system based on edge nodes includes edge nodes and terminals. The method includes: Obtain the corpus to be processed; According to the industry knowledge base and the scenario corpus located at the edge node, perform industry feature matching on the corpus to be processed to determine the corpus scenario; Determine an industry language model according to the corpus scenario; Determine a first semantic result according to the industry language model and the corpus to be processed; the first semantic result includes a request type and a request intention; Return the first semantic result to the terminal; Wherein, the semantic processing system based on edge nodes further includes a cloud, and the method further includes: Calculate a first confidence level of the first semantic result according to the industry knowledge base and the scenario corpus; the first confidence level is used to characterize the fitting degree of the first semantic result with the corresponding industry and the corresponding business scenario; When the first confidence level is lower than a preset confidence level threshold, initiate a collaborative processing request to the cloud so that the cloud obtains supplementary corpus and returns the supplementary corpus to the edge node; Select several key texts in the supplementary corpus and add them to the first semantic result to generate several second semantic results; Calculate a second confidence level of the second semantic result according to the industry knowledge base and the scenario corpus; the second confidence level is used to characterize the fitting degree of the second semantic result with the corresponding industry and the corresponding business scenario; Return the second semantic result with the highest second confidence level to the terminal; Wherein, the supplementary corpus is corpus within the same industry other than the content recorded in the industry knowledge base and the scenario corpus. The step of enabling the cloud to obtain supplementary corpus includes: According to preset retrieval conditions, enable the cloud to retrieve the supplementary corpus on the Internet; wherein, the retrieval conditions include being the same as the pronunciation of the first semantic result; Wherein, the method further includes the construction process of the industry language model, which specifically includes: Randomly extract N types of samples from the obtained corpus samples as a first sample set, and the first sample set contains N types of first samples; wherein, the sample categories in the first sample set include scenario results and target topics; In each type of the first samples, extract K instances as a first instance set, and the first instance set contains K first instances; wherein, the instance is a feature word; Take all the extracted first instances as a support set, and take all the instances in the first sample set except the first instances as a query set; wherein, the support set is used for model training, and the query set is used for model testing; Use the support set and the query set to train and test the industry language model; During the training process, adopt the method of gradient weight increase to gradually increase the weight of the labeled instances in the first instances; When the number of training times reaches a preset first quantity, complete the construction of the industry language model; Wherein, both N and M are positive integers.

2. The semantic processing method based on edge nodes according to claim 1, characterized in that, The method further includes: Add the supplementary corpus to the industry knowledge base and the scenario corpus.

3. A semantic processing system based on edge nodes, characterized in that, The system is applied to the edge-node-based semantic processing method according to any one of claims 1-2. The system is applied to an edge node in an edge-node-based semantic processing system, and the edge-node-based semantic processing system includes an edge node and a terminal. The system includes: a first module, a second module, a third module, a fourth module, and a fifth module; The first module is configured to obtain a corpus to be processed; The second module is configured to perform industry feature matching on the corpus to be processed to determine a corpus scenario; The third module is configured to determine an industry language model according to the corpus scenario; The fourth module is configured to determine a first semantic result according to the industry language model and the corpus to be processed; The fifth module is configured to return the first semantic result to the terminal; Wherein, the edge-node-based semantic processing system further includes a cloud, and the system further includes: a sixth module, a seventh module, an eighth module, a ninth module, and a tenth module; The sixth module is configured to calculate a first confidence level of the first semantic result according to an industry knowledge base and a scenario-based knowledge base located at the edge node; the first confidence level is used to characterize the fitting degree of the first semantic result with the corresponding industry and the corresponding business scenario; The seventh module is configured to, when the first confidence level is lower than a preset confidence level threshold, obtain supplementary corpus through the cloud; the supplementary corpus is corpus within the same industry other than the content recorded in the industry knowledge base and the scenario-based corpus; The eighth module is configured to select several key texts from the supplementary corpus and add them to the first semantic result to generate several second semantic results; The ninth module is configured to calculate a second confidence level of the second semantic result according to the industry knowledge base and the scenario-based knowledge base; the second confidence level is used to characterize the fitting degree of the second semantic result with the corresponding industry and the corresponding business scenario; The tenth module is configured to return the second semantic result with the highest second confidence level to the terminal.

4. A device, characterized in that, Including: At least one processor; At least one memory for storing at least one program; When the at least one program is executed by the at least one processor, the at least one processor implements the edge-node-based semantic processing method according to any one of claims 1-2.

5. A computer storage medium storing a program executable by a processor, characterized in that, The program executable by the processor, when executed by the processor, is used to implement the edge-node-based semantic processing method according to any one of claims 1-2.

Images