Semantic feature extraction method, device, equipment and communication system

Abstract

The invention relates to a semantic feature extraction method and device, equipment and a communication system. The method comprises the steps of determining to perform semantic feature extraction locally or in a cloud according to local available resources; if it is determined that semantic feature extraction is performed locally, semantic features are extracted based on a local semantic knowledge base; if it is determined that semantic feature extraction is performed at the cloud, sending indication information to a server; the indication information is used for indicating the server to extract semantic features based on the public semantic knowledge base. According to the method, the semantic feature extraction mode is determined according to the local available resources, and the sending end can be assisted to complete semantic feature extraction in an end-cloud cooperation mode, so that the semantic communication system can transmit information more effectively, and the quality and efficiency of the whole semantic communication process are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of wireless communication, and in particular to a semantic feature extraction method and device, equipment and communication system. BACKGROUND

[0002] At present, semantic communication as a new communication paradigm solves the problem of how to transmit the exact meaning of the symbol. However, in the face of the explosive growth of communication data, the traditional scheme has the problem of low efficiency in extracting semantic features. SUMMARY

[0003] The embodiments of the present application provide a semantic feature extraction method, device, equipment and communication system, which can efficiently complete semantic feature extraction.

[0004] In a first aspect, the present application provides a semantic feature extraction method applied to a sending end, which comprises:

[0005] According to the local available resources, it is determined whether to perform semantic feature extraction locally or in the cloud.

[0006] If it is determined to perform semantic feature extraction locally, the semantic features are extracted based on the local semantic knowledge base.

[0007] If it is determined to perform semantic feature extraction in the cloud, the indication information is sent to the server; the indication information is used to instruct the server to extract the semantic features based on the public semantic knowledge base.

[0008] In one of the embodiments, according to the local available resources, it is determined whether to perform semantic feature extraction locally or in the cloud, which comprises:

[0009] According to the local available resources and the resources required for semantic feature extraction, it is confirmed whether the local has the ability to complete semantic feature extraction.

[0010] If the local has the ability to complete semantic feature extraction, the performance indicators of completing semantic feature extraction locally and completing semantic feature extraction by unloading to the cloud are compared, and according to the comparison result, it is determined whether to perform semantic feature extraction locally or in the cloud.

[0011] In one of the embodiments, if it is determined to perform semantic feature extraction locally, the semantic features are extracted based on the local semantic knowledge base, which comprises:

[0012] The semantic features are extracted based on the local semantic knowledge base.

[0013] The optimization suggestion from the server is received, and the semantic features are optimized according to the optimization suggestion.

[0014] In one of the embodiments, before the step of receiving the optimization suggestion from the server and optimizing the semantic feature according to the optimization suggestion, further comprising:

[0015] Creating a prompt word based on the semantic communication target and the semantic feature;

[0016] Transmitting the prompt word and necessary data to the server to instruct the server to generate an optimization suggestion for the semantic feature based on the public semantic knowledge base, and transmitting the optimization suggestion to the sending end.

[0017] In one of the embodiments, the indication information includes the prompt word and the necessary data.

[0018] In one of the embodiments, if it is determined to perform semantic feature extraction in the cloud, the indication information is sent to the server, including:

[0019] Creating a prompt word based on the semantic communication target and the local semantic knowledge base;

[0020] Transmitting the prompt word and necessary data to the server to instruct the server to extract the semantic feature based on the public semantic knowledge base, and transmitting the semantic feature to the sending end.

[0021] In one of the embodiments, according to the local available resources and the resources required for semantic feature extraction, it is determined whether the local has the ability to complete the semantic feature extraction, including:

[0022] Evaluating the to-be-processed data to obtain data volume information of the to-be-processed data;

[0023] According to the data volume information, evaluating resource requirements of running the local semantic feature extraction algorithm, and taking the resource requirements as the resources required for semantic feature extraction;

[0024] If the local available resources are greater than the resources required for semantic feature extraction, it is determined that the local has the ability to complete the semantic feature extraction.

[0025] In one of the embodiments, the data volume information includes at least one of data size and data type; and the resource requirements include at least one of computing requirements and memory requirements.

[0026] In one of the embodiments, the resources required for semantic feature extraction include semantic feature extraction computing requirements and semantic feature extraction memory requirements.

[0027] If the local available resources are greater than the resources required for semantic feature extraction, it is determined that the local has the ability to complete the semantic feature extraction, including:

[0028] Obtaining the local available resources; the local available resources include local available computing resources and local available memory space;

[0029] If the local available computing resource is greater than the semantic feature extraction computing requirement, and the local available memory space is greater than the semantic feature extraction memory requirement, it is determined that the local has the capability to complete the semantic feature extraction.

[0030] In one of the embodiments, the performance indicators of the local completion of the semantic feature extraction include the local completion time and the local completion energy consumption; the performance indicators of the offloading to the cloud for the completion of the semantic feature extraction include the cloud completion time, the data transmission time and the data transmission energy consumption for transmitting the data to be processed to the server;

[0031] The performance indicators of the local completion of the semantic feature extraction are compared with the performance indicators of the offloading to the cloud for the completion of the semantic feature extraction, and according to the comparison result, it is determined whether to execute the semantic feature extraction locally or in the cloud, including:

[0032] The data amount information is transmitted to the server; the data amount information is used to instruct the server to calculate and feed back the cloud completion time, the data transmission time and the data transmission energy consumption;

[0033] If the local completion time is less than the sum of the data transmission time and the cloud completion time, and the local completion energy consumption is less than the data transmission energy consumption, it is determined to execute the semantic feature extraction locally.

[0034] If the local completion time is greater than or equal to the sum of the data transmission time and the cloud completion time, and / or the local completion energy consumption is greater than or equal to the data transmission energy consumption, it is determined to execute the semantic feature extraction in the cloud.

[0035] In one of the embodiments, the local semantic knowledge base includes one or more of user information, device information, target domain knowledge and context information.

[0036] In one of the embodiments, the user information includes one or more of user profile, user preference, user behavior pattern and historical interaction record; the device information includes one or more of hardware configuration, software configuration, performance parameter and device running state; the target domain knowledge includes one or more of concept definition, entity relationship, rule logic and domain model; the context information includes one or more of environment state, situational information and time information.

[0037] In one of the embodiments, the public semantic knowledge base includes one or more of general knowledge, domain professional knowledge, device interaction knowledge and model library.

[0038] In one of the embodiments, the general knowledge comprises one or more of common sense, concept definitions, and general rules; the domain expertise comprises one or more of professional terms, cases, regulations, and industry standards; the device interaction knowledge comprises one or more of device types, interfaces, communication protocols, and operation instructions; and the model library comprises one or more of statistical models, machine learning models, and deep learning models.

[0039] In one of the embodiments, the semantic features comprise one or more of entity recognition, sentiment analysis, topic classification, and semantic role labeling.

[0040] In one of the embodiments, the optimization suggestions comprise one or more of maintaining, expanding, pruning, merging, splitting, refining, and simplifying.

[0041] In one of the embodiments, the necessary data comprise one or more of semantic features, raw data, data types, data formats, data creation times, and data versions.

[0042] In one of the embodiments, the prompt words comprise one or more of communication target descriptions, semantic feature summaries, and sender intents.

[0043] In one of the embodiments, the presentation of the prompt words comprises one or more of text forms, vector forms, and structured data.

[0044] In one of the embodiments, the necessary data comprise at least one of raw data, data types, data formats, data creation times, and data versions.

[0045] In one of the embodiments, the prompt words comprise one or more of communication target descriptions, communication scenario descriptions, sender intents, and personalized labels.

[0046] In one of the embodiments, the presentation of the prompt words comprises one or more of text forms, vector forms, and structured data.

[0047] In a second aspect, the present application provides a semantic feature extraction method, applied to a server, comprising:

[0048] receiving indication information from a sender;

[0049] extracting semantic features based on a public semantic knowledge base according to the indication information.

[0050] In one of the embodiments, the indication information is sent by the sender according to local available resources in a case where the sender determines to perform semantic feature extraction in the cloud; and the indication information comprises prompt words and necessary data.

[0051] In one of the embodiments, according to the indication information, semantic features are extracted based on a public semantic knowledge base, including:

[0052] In response to receiving the prompt word and the necessary data, semantic features are extracted based on a public semantic knowledge base;

[0053] The semantic features are transmitted to the sending end.

[0054] In one of the embodiments, the sending end determines, according to the locally available resources, that the semantic feature extraction is performed locally, and semantic features are extracted based on a local semantic knowledge base.

[0055] In one of the embodiments, the method further includes:

[0056] The optimization suggestion is sent to the sending end; the optimization suggestion is used to instruct the sending end to optimize the semantic features extracted based on the local semantic knowledge base.

[0057] In one of the embodiments, before the optimization suggestion is sent to the sending end, it includes:

[0058] The prompt word and the necessary data are received from the sending end;

[0059] According to the prompt word and the necessary data, an optimization suggestion for the semantic features is generated based on a public semantic knowledge base.

[0060] In one of the embodiments, the prompt word is created by the sending end based on the semantic communication target and the semantic features.

[0061] In one of the embodiments, the method further includes:

[0062] The performance index of the semantic feature extraction completed by offloading to the cloud is transmitted to the sending end; the performance index of the semantic feature extraction completed by offloading to the cloud is used to instruct the sending end to compare the performance index of the semantic feature extraction completed locally and the performance index of the semantic feature extraction completed by offloading to the cloud, and according to the comparison result, determine whether to perform the semantic feature extraction locally or in the cloud.

[0063] In one of the embodiments, the performance index of the semantic feature extraction completed by offloading to the cloud includes the cloud completion time, and the data transmission time and data transmission energy consumption of transmitting the to-be-processed data to the server; the method further includes:

[0064] The data volume information of the to-be-processed data from the sending end is received;

[0065] According to the data volume information, the cloud completion time, the data transmission time and the data transmission energy consumption are calculated, and the cloud completion time, the data transmission time and the data transmission energy consumption are transmitted to the sending end.

[0066] In one embodiment, the data volume information includes at least one of a data size and a data type.

[0067] In a third aspect, the present application provides a semantic feature extraction device, applied to a sending end, and comprising:

[0068] A determination module is configured to determine whether to perform semantic feature extraction locally or in the cloud according to local available resources.

[0069] A local extraction module is configured to extract semantic features based on a local semantic knowledge base if it is determined to perform semantic feature extraction locally.

[0070] An indication module is configured to send indication information to a server if it is determined to perform semantic feature extraction in the cloud, and the indication information is used to instruct the server to extract semantic features based on a public semantic knowledge base.

[0071] In a fourth aspect, the present application provides a semantic feature extraction device, applied to a server, and comprising:

[0072] An information receiving module is configured to receive indication information from a sending end.

[0073] A feature extraction module is configured to extract semantic features based on a public semantic knowledge base according to the indication information.

[0074] In a fifth aspect, the present application provides a sending end, comprising a transmitter and a processor.

[0075] The processor is configured to determine whether to perform semantic feature extraction locally or in the cloud according to local available resources, extract semantic features based on a local semantic knowledge base if it is determined to perform semantic feature extraction locally, and control the transmitter to send indication information to a server if it is determined to perform semantic feature extraction in the cloud, and the indication information is used to instruct the server to extract semantic features based on a public semantic knowledge base.

[0076] In a sixth aspect, the present application provides a server, comprising a receiver and a processor.

[0077] The receiver is configured to receive indication information from a sending end.

[0078] The processor is configured to extract semantic features based on a public semantic knowledge base according to the indication information.

[0079] In a seventh aspect, the present application provides a communication system, comprising the sending end of the fifth aspect and the server of the sixth aspect.

[0080] In an eighth aspect, the present application provides a computer readable storage medium, having stored thereon a computer program, which, when executed by a processor, implements the steps of the method of any of the above aspects.

[0081] In a ninth aspect, the present application provides a computer program product, comprising a computer program which, when executed by a processor, implements the steps of the method of any of the above aspects.

[0082] The semantic feature extraction method, device, equipment and communication system described above, the sending end determines to perform semantic feature extraction locally or in the cloud according to the local available resources, wherein if it is determined to perform semantic feature extraction locally, the semantic features are extracted based on the local semantic knowledge base, and if it is determined to perform semantic feature extraction in the cloud, the server is instructed to extract semantic features based on the public semantic knowledge base by sending instruction information to the server. The application determines the semantic feature extraction mode according to the local available resources, can assist the sending end to complete the semantic feature extraction in the end-cloud collaborative manner, so that the semantic communication system can more effectively transmit information, thereby improving the quality and efficiency of the whole semantic communication process. BRIEF DESCRIPTION OF DRAWINGS

[0083] Figure 1 An application environment diagram of the semantic feature extraction method in an embodiment;

[0084] Figure 2 A flowchart of the semantic feature extraction method in an embodiment;

[0085] Figure 3 A flowchart of determining to perform semantic feature extraction locally or in the cloud in an embodiment;

[0086] Figure 4 A flowchart of determining that the local end has the ability to complete semantic feature extraction in an embodiment;

[0087] Figure 5 A flowchart of determining that the local available resources are greater than the required resources for semantic feature extraction in an embodiment;

[0088] Figure 6 A specific flowchart of determining to perform semantic feature extraction locally or in the cloud in an embodiment;

[0089] Figure 7 A flowchart of optimizing semantic features in an embodiment;

[0090] Figure 8 A specific flowchart of optimizing semantic features in an embodiment;

[0091] Figure 9Fig. 1 is a schematic diagram of a process of semantic feature extraction in the cloud in one embodiment;

[0092] Figure 10 Fig. 2 is a schematic diagram of a process of a semantic feature extraction method in another embodiment;

[0093] Figure 11 Fig. 3 is a schematic diagram of a semantic feature extraction framework in one embodiment;

[0094] Figure 12 Fig. 4 is a schematic diagram of a specific process of a semantic feature extraction method in one embodiment;

[0095] Figure 13 Fig. 5 is a structural block diagram of a semantic feature extraction device in one embodiment;

[0096] Figure 14 Fig. 6 is a structural block diagram of a semantic feature extraction device in another embodiment;

[0097] Figure 15 Fig. 7 is an internal structural diagram of a terminal in one embodiment;

[0098] Figure 16 Fig. 8 is an internal structural diagram of a server in one embodiment. DETAILED DESCRIPTION

[0099] In order to make the purpose, technical scheme and advantages of the present application clearer, the present application is further described in detail below with reference to the 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.

[0100] The terms "first", "second", and the like in the specification and claims of the present application are used to distinguish similar objects, and are not used to describe the order or sequence of the objects. It should be understood that the data used in this way can be exchanged under appropriate circumstances, so that the embodiments of the present application can be implemented in an order other than those illustrated or described herein, and the objects distinguished by "first", "second" are generally a class, and are not limited to the number of objects, for example, the first object can be one or more. In addition, the term "and / or" is only a description of the association between the associated objects, which means that there can be three relationships, for example, A and / or B can mean that A exists alone, A and B exist together, and B exists alone. In addition, the character " / " generally represents an "or" relationship between the associated objects before and after.

[0101] In the description of the embodiments of the present application, the meaning of "a plurality of" is two or more, unless otherwise specified. As used herein, the singular forms "a," "an," and "the" can also include the plural forms, unless the context clearly indicates otherwise. It should also be understood that the term "comprises / comprising" or "has / having" specifies the presence of stated features, integers, steps, operations, components, parts, or combinations thereof, but does not preclude the presence or addition of one or more other features, integers, steps, operations, components, parts, or combinations thereof.

[0102] It is worth noting that the terms "system" and "network" in the embodiments of the present application are often used interchangeably, and the described technology can be applied to the above-mentioned systems and other systems. The following description describes a New Radio (NR) system for example purposes, and NR terminology is used in most of the following description, but these technologies can also be applied to applications other than NR system applications, such as 6th Generation (6G) communication systems, or can be applied to next-generation mobile communication systems or other similar communication systems, without limitation.

[0103] Figure 1 A schematic diagram of a wireless communication system to which embodiments of the present application can be applied. The wireless communication system includes at least one terminal 102 and a server 104, and the terminal 102 communicates with the server 104 through a network, wherein the terminal 102 and the server 104 can be connected through a wireless or wired network to complete data transmission and exchange.

[0104] The terminal 102 can be, but is not limited to, various personal computers, notebook computers, smart phones, tablet computers, Internet of Things devices, and portable wearable devices. The Internet of Things device can be a smart speaker, a smart television, a smart air conditioner, a smart vehicle device, a projection device, etc. The portable wearable device can be a smart watch, a smart bracelet, a head-mounted device, etc. The head-mounted device can be a Virtual Reality (VR) device, an Augmented Reality (AR) device, smart glasses, etc.

[0105] The server 104 can be a standalone physical server, or a server cluster or distributed system composed of multiple physical servers, or a cloud server providing cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, CDN (Content Delivery Network), and basic cloud computing services such as big data and artificial intelligence platforms.

[0106] Exemplarily, taking a server as a cloud server (referred to as a cloud server) for example, a terminal as a sending end can be directly or indirectly connected with the cloud server through wired or wireless communication mode, and the embodiments of the present application do not limit this.

[0107] In the conventional technology, only the end-side capability is used for feature extraction, and the end-side resource is usually limited, which may cause significant processing delay or even failure of the entire semantic communication process when facing data-intensive or computation-intensive tasks.

[0108] Based on the above conventional technology, the embodiments of the present application perform semantic feature extraction based on a semantic knowledge base for a semantic communication system, which can optimize the end-side feature by using a cloud-side multi-modal model. In the present application, not only the semantic feature extraction is completed by the algorithmic collaboration between the end and the cloud, but also the optimization of the semantic feature is completed by the collaboration of the local semantic knowledge base of the sending end and the public semantic knowledge base of the cloud, so as to improve the accuracy of the semantic feature. In addition, in the present application, the cloud side and the end side collaboratively complete the semantic feature extraction, which overcomes the limitation of the end-side resource and ensures that the semantic feature extraction can be efficiently completed in a resource-intensive task.

[0109] The semantic feature extraction method provided by the embodiments of the present application ensures the smooth extraction of the semantic feature when facing a resource-intensive task through the collaboration between the sending end and the cloud server. At the same time, the accuracy of the semantic feature extraction is improved by the collaboration of the local semantic knowledge base of the sending end and the public semantic knowledge base of the cloud server.

[0110] It should be noted that the beneficial effects or technical problems solved by the embodiments of the present application are not limited to this, but also other implicit or related problems, which can be referred to the description of the following embodiments.

[0111] Before introducing the specific embodiments of the present application, the professional terms involved in the present application are explained:

[0112] Device-Cloud Collaboration: an artificial intelligence technology paradigm that collaborates between the cloud side and the end side, which does not completely rely on centralized cloud computing resources, but deploys part of the AI (Artificial Intelligence) tasks on the end side, so as to solve the problems of traditional cloud intelligent services in privacy security, load, cost, etc., and improve the real-time performance and personalized service capability of the intelligent system.

[0113] Semantic Communication: focuses on the meaning of the transmitted information, rather than the accurate restoration of the communication symbols themselves, and is mainly used to solve the problem of how to transmit the exact meaning of the symbols.

[0114] Semantic Knowledge Base: A structured and memory-capable knowledge network model that can provide relevant semantic knowledge description for data information. The semantic knowledge base can provide knowledge background and storage search services for the extraction, identification, transmission, understanding and reasoning process of semantic elements in semantic communication, define an efficient search space, standardize the search path, and is one of the key enabling technologies for semantic communication. Usually, in a semantic communication system, the sending end and the receiving end have their own local semantic knowledge base, and the cloud server maintains a public semantic knowledge base.

[0115] Semantic Feature: The key information used for semantic communication after selective feature extraction of raw information. It not only contains the direct content of the data, but also may contain the context, relevance and relationship with other data of the data. The accuracy of semantic feature extraction is directly related to whether the communication system can correctly understand and convey the intention and content of the information.

[0116] The technical solutions of the present application and how the technical solutions of the present application solve the above technical problems will be described in detail below with specific examples. The following specific examples can be combined with each other, and the same or similar concepts or processes may not be described again in some examples. The embodiments of the present application will be described below with reference to the drawings.

[0117] In an exemplary embodiment, as shown in Figure 2 , a semantic feature extraction method is provided, which is applied to a terminal in Figure 1 for example (it can be understood that the terminal can communicate with the server as a sending end) includes the following steps 202 to 206. Among them:

[0118] Step 202, according to the local available resources, determine to perform semantic feature extraction locally or in the cloud.

[0119] Among them, the local available resources can be the current available resources of the sending end (such as the terminal), such as computing resources, memory space, etc., which are not limited by the present application.

[0120] Specifically, the sending end can determine the semantic feature extraction method according to the local available resources. The semantic feature extraction method in the embodiment of the present application can include a first extraction method and a second extraction method, and the second extraction method is different from the first extraction method. Exemplarily, the first extraction method can be to perform semantic feature extraction locally, and the second extraction method can be to perform semantic feature extraction in the cloud.

[0121] Optionally, the sending end can select to perform semantic feature extraction in the cloud when the locally available resources do not satisfy the corresponding conditions; or can select to perform semantic feature extraction locally when the locally available resources satisfy the corresponding conditions, thereby realizing collaborative completion of semantic feature extraction between the sending end and the cloud.

[0122] The embodiment of the application determines the semantic feature extraction mode based on the locally available resources, so that the cloud side and the end side can collaboratively complete semantic feature extraction, overcoming the resource limitation of the end side, enabling smooth extraction of semantic features when facing resource-intensive tasks, and further ensuring efficient completion of semantic feature extraction in resource-intensive tasks.

[0123] In step 204, if it is determined to perform semantic feature extraction locally, the semantic features are extracted based on the local semantic knowledge base.

[0124] Specifically, when it is determined to perform semantic feature extraction locally, the sending end can extract semantic features based on the local semantic knowledge base. The local semantic knowledge base can be the semantic knowledge base of the sending end.

[0125] For example, the sending end can extract semantic features based on the local semantic knowledge base when it is determined to perform semantic feature extraction locally according to the locally available resources.

[0126] In step 206, if it is determined to perform semantic feature extraction in the cloud, the sending end sends indication information to the server; the indication information is used to instruct the server to extract semantic features based on the public semantic knowledge base.

[0127] Specifically, when it is determined to perform semantic feature extraction in the cloud, the sending end can send indication information to the server to instruct the server to extract semantic features based on the public semantic knowledge base. The public semantic knowledge base can be a semantic knowledge base maintained by the server, such as a public semantic knowledge base maintained on a cloud server.

[0128] For example, taking the cloud server as an example, the sending end can send relevant data (such as indication information) to the cloud server when it is determined to perform semantic feature extraction in the cloud according to the locally available resources, and the cloud server extracts semantic features based on the public semantic knowledge base.

[0129] The above semantic feature extraction method, in which the sending end determines the semantic feature extraction mode according to the locally available resources, can face a semantic communication system and utilize an end-cloud collaborative manner to assist the sending end to complete semantic feature extraction. The embodiment of the application utilizes the collaboration of the local semantic knowledge base of the sending end and the public semantic knowledge base of the server (such as a cloud server) to complete semantic feature extraction, so that the semantic communication system can more effectively transmit information, thereby improving the quality and efficiency of the entire semantic communication process.

[0130] In one embodiment, as shown in FIG. 3, step 202 can include steps 302-304. In which: Figure 3

[0131] In step 302, the required resources for semantic feature extraction are extracted according to the locally available resources and semantic feature extraction, and it is determined whether the local has the ability to complete the semantic feature extraction.

[0132] Specifically, the sender can extract the required resources for semantic feature extraction according to the locally available resources and semantic feature extraction, and determine whether the local has the ability to complete the semantic feature extraction. The required resources for semantic feature extraction can refer to the resource requirement of semantic feature extraction. For example, the sender can evaluate the locally available resources and the resource requirement of semantic feature extraction, and determine whether the local has the ability to complete the semantic feature extraction.

[0133] Optionally, the sender can compare the locally available resources and the required resources for semantic feature extraction to determine whether the locally available resources can meet the resource requirement of semantic feature extraction, and then determine whether the local has the ability to complete the semantic feature extraction. In the embodiment of the present application, the sender can determine the semantic feature extraction mode according to the locally available resources, and then realize the mode of assisting the sender to complete the semantic feature extraction by the end-cloud collaboration.

[0134] Further, the sender can determine the locally available resources through local resource evaluation, and obtain the required resources for semantic feature extraction through semantic feature extraction resource evaluation. In addition, the sender can also perform channel resource evaluation to determine the communication between the sender and the server, for example, in a semantic communication system, the available channel resources are analyzed and evaluated to ensure effective signal transmission and optimal network performance. For example, when it is determined that the current situation is not suitable for signal transmission based on the result of channel resource evaluation, it can be determined that the semantic feature extraction is performed locally.

[0135] In step 304, if the local has the ability to complete the semantic feature extraction, the performance indicators of completing the semantic feature extraction locally and completing the semantic feature extraction by offloading to the cloud are compared, and according to the comparison result, it is determined whether to perform the semantic feature extraction locally or in the cloud.

[0136] Specifically, if the local has the ability to complete the semantic feature extraction, the sender can compare the performance indicators of completing the semantic feature extraction locally and completing the semantic feature extraction by offloading to the cloud, and then determine whether to perform the semantic feature extraction locally or in the cloud according to the comparison result.

[0137] ​In the embodiments of the present application, the sending end can determine whether the local has the capability to complete the semantic feature extraction, and then assist the sending end to complete the semantic feature extraction in the mode of end-cloud cooperation in the case that the local has the capability to complete the semantic feature extraction, so as to ensure the smooth extraction of the semantic feature in the face of resource-intensive tasks, and the optimization of the semantic feature can be completed by the cooperation of the local semantic knowledge base of the sending end and the public semantic knowledge base of the cloud server, the accuracy of the semantic feature extraction is improved, the semantic communication system can more effectively transmit information, and thus the quality and efficiency of the whole semantic communication process are improved.

[0138] In one exemplary embodiment, as shown in FIG. 4, step 302 can include steps 402 to 406. Wherein: Figure 4

[0139] In step 402, the to-be-processed data is evaluated, and the data amount information of the to-be-processed data is obtained.

[0140] Specifically, the to-be-processed data can refer to the data that needs to be processed, wherein the sending end can evaluate the to-be-processed data and obtain the data amount information of the to-be-processed data.

[0141] Exemplarily, the data amount information can include at least one of a data size and a data type. Optionally, the sending end can obtain the data size and the data type of the to-be-processed data.

[0142] In step 404, according to the data amount information, the resource requirement of running the local semantic feature extraction algorithm is evaluated, and the resource requirement is taken as the required resource for the semantic feature extraction.

[0143] Specifically, the sending end can check whether the local has the semantic feature extraction algorithm, and in the case that the local has the semantic feature extraction algorithm, the resource requirement of running the algorithm is evaluated. Exemplarily, the resource requirement can include at least one of a computing requirement and a memory requirement.

[0144] In the case that the sending end evaluates the resource requirement of running the local semantic feature extraction algorithm according to the data amount information, the resource requirement can be taken as the required resource for the semantic feature extraction.

[0145] In step 406, if the local available resource is greater than the required resource for the semantic feature extraction, it is determined that the local has the capability to complete the semantic feature extraction.

[0146] Specifically, the sending end can determine whether the local available resource is greater than the required resource for the semantic feature extraction, for example, compare the local available resource and the required resource for the semantic feature extraction to determine whether the local available resource is greater than the required resource for the semantic feature extraction.

[0147] ​When the local available resource is greater than the resource required for semantic feature extraction, the sending end can determine that the local has the ability to complete the semantic feature extraction, otherwise, it is determined that the local does not have the ability to complete the semantic feature extraction.

[0148] Further, taking the resource requirement including the computing requirement and the memory requirement as an example, in one embodiment, the resource required for semantic feature extraction can include semantic feature extraction computing requirement and semantic feature extraction memory requirement; as Figure 5 As shown in FIG. 5, step 406 can include steps 502-504. Wherein:

[0149] Step 502, obtaining the local available resource; the local available resource includes the local available computing resource and the local available memory space.

[0150] Specifically, the sending end can obtain the local available resource, for example, check the local available resource; wherein the local available resource can include the local available computing resource and the local available memory space.

[0151] Step 504, if the local available computing resource is greater than the semantic feature extraction computing requirement, and the local available memory space is greater than the semantic feature extraction memory requirement, it is determined that the local has the ability to complete the semantic feature extraction.

[0152] Specifically, the sending end judges whether the local available resource is greater than the resource required for extracting semantic features, if the local available computing resource > semantic feature extraction computing requirement, and the local available memory space > semantic feature extraction memory requirement, it is indicated that the local has the ability to complete the semantic feature extraction. Otherwise, it does not have.

[0153] In actual application, if the local has the ability to complete the semantic feature extraction, the sending end can compare the performance index of completing the semantic feature extraction locally and offloading to the cloud to complete the semantic feature extraction, and decide to execute the semantic feature extraction locally or in the cloud according to the comparison result. In one exemplary embodiment, as Figure 6 As shown in FIG. 6, step 504 can include steps 602-606. Wherein:

[0154] Step 602, transmitting the data amount information to the server; the data amount information is used to instruct the server to calculate and feedback the cloud completion time, data transmission time and data transmission energy consumption.

[0155] Specifically, the performance index of completing the semantic feature extraction locally can include the local completion time and the local completion energy consumption, wherein the local completion time represents the time required for completing the semantic feature extraction locally, and the local completion energy consumption represents the energy consumption required for completing the semantic feature extraction locally. Exemplarily, the sending end can calculate the time and energy consumption required for completing the semantic feature extraction.

[0156] Optionally, the performance indicators of offloading to the cloud to complete the semantic feature extraction can include a cloud completion time, and a data transmission time and a data transmission energy consumption of transmitting the to-be-processed data to the server; wherein the cloud completion time represents a time required for completing the semantic feature extraction in the cloud.

[0157] In the embodiments of the present application, the sending end can transmit the data volume information to the server to instruct the server to calculate the cloud completion time, the data transmission time and the data transmission energy consumption, and feed back the cloud completion time, the data transmission time and the data transmission energy consumption to the sending end. Taking the server as an example, the sending end can send the data scale and the data type to the cloud server, the cloud server calculates the time and energy consumption required for completing the semantic feature extraction, and transmits the results to the sending end. The sending end compares the time and energy consumption of completing locally and in the cloud, and decides to execute the semantic feature extraction locally or in the cloud.

[0158] In step 604, if the local completion time is less than the sum of the data transmission time and the cloud completion time, and the local completion energy consumption is less than the data transmission energy consumption, it is determined to execute the semantic feature extraction locally.

[0159] Specifically, if the local completion time is less than the sum of the data transmission time and the cloud completion time, and the local completion energy consumption is less than the data transmission energy consumption, it is determined to execute the semantic feature extraction locally.

[0160] In step 606, if the local completion time is greater than or equal to the sum of the data transmission time and the cloud completion time, and / or the local completion energy consumption is greater than or equal to the data transmission energy consumption, it is determined to execute the semantic feature extraction in the cloud.

[0161] Specifically, if the local completion time is greater than or equal to the sum of the data transmission time and the cloud completion time, and / or the local completion energy consumption is greater than or equal to the data transmission energy consumption, the sending end determines to execute the semantic feature extraction in the cloud.

[0162] It can be understood that taking the terminal as an example, the terminal is usually powered by a battery with limited capacity, and the energy consumption of the terminal device needs to be particularly concerned when the terminal completes a complex task. In contrast, the cloud has higher energy efficiency and larger energy capacity; based on this, the embodiments of the present application compare the performance indicators of completing the semantic feature extraction locally and offloading to the cloud to complete the semantic feature extraction, to determine to execute the semantic feature extraction locally or in the cloud, and then realize the use of end-cloud collaboration to assist the sending end to complete the semantic feature extraction.

[0163] The above semantic feature extraction method ensures the smooth extraction of semantic features when facing resource-intensive tasks through the cooperation between the sending end and the cloud server. At the same time, the accuracy of semantic feature extraction is improved by the cooperation of the local semantic knowledge base of the sending end and the public semantic knowledge base of the cloud server.

[0164] In an example embodiment, as shown in FIG. 8, step 704 can include steps 802-804. Wherein: Figure 7

[0165] Step 802, based on the semantic communication target and the semantic features, create a prompt word.

[0166] Specifically, the sending end can create a prompt word based on the semantic communication target and the semantic features after obtaining the semantic features. Wherein, the semantic communication target can refer to the target information to be obtained through this semantic communication in the current application scenario; for example, in the automatic driving scenario, the semantic communication target can be to obtain the surrounding environment information including pedestrians, vehicles, traffic signals, etc. through the semantic communication between vehicles and vehicles, vehicles and traffic infrastructure.

[0167] Step 804, send the prompt word to the server.

[0168] Specifically, the sending end can send the prompt word to the server. Wherein, the server can obtain the prompt word based on the public semantic knowledge base, and then obtain the optimization suggestion based on the prompt word. For example, the server can obtain the prompt word based on the public semantic knowledge base, and then obtain the optimization suggestion based on the prompt word. It can be understood that the application embodiment does not limit the way the server obtains the optimization suggestion.

[0169] Optionally, the optimization suggestion can be obtained by the server based on the public semantic knowledge base. For example, the sending end can send the semantic features and related data extracted based on the local semantic knowledge base to the server, so that the server obtains the optimization suggestion based on the public semantic knowledge base. It can be understood that the application embodiment does not limit the way the server obtains the optimization suggestion.

[0170] In an example embodiment, as shown in FIG. 8, step 704 can include steps 802-804. Wherein: Figure 8

[0171] Wherein:

[0172] Step 802, based on the semantic communication target and the semantic features, create a prompt word.

[0173] Specifically, the sending end can create a prompt word based on the semantic communication target and the semantic features after obtaining the semantic features. Wherein, the semantic communication target can refer to the target information to be obtained through this semantic communication in the current application scenario; for example, in the automatic driving scenario, the semantic communication target can be to obtain the surrounding environment information including pedestrians, vehicles, traffic signals, etc. through the semantic communication between vehicles and vehicles, vehicles and traffic infrastructure.

[0174] ​​Exemplarily, according to a specific application scenario (for example, a semantic communication scenario), the obtaining manner of the semantic communication target can include that the semantic communication parties jointly confirm before a formal communication process, or the semantic communication target can be obtained from a server (for example, a cloud server).

[0175] It can be understood that, for the purpose of distinction, the prompt word created by the sending end based on the semantic communication target and the semantic feature in the embodiments of the present application can be referred to as a first prompt word.

[0176] Step 804, transmitting the prompt word and necessary data to the server to instruct the server to generate an optimization suggestion for the semantic feature based on the public semantic knowledge base, and transmitting the optimization suggestion to the sending end.

[0177] Specifically, the sending end can transmit the prompt word and necessary data to the server, and then instruct the server to generate an optimization suggestion for the semantic feature based on the public semantic knowledge base, and transmit the optimization suggestion to the sending end.

[0178] The necessary data can refer to data related to semantic feature extraction. Further, when it is determined that the semantic feature extraction is performed locally, the necessary data can include the semantic feature extracted by the sending end based on the local semantic knowledge base. It can be understood that, for the purpose of distinction, the necessary data used to instruct the server to generate an optimization suggestion in the embodiments of the present application can be referred to as first necessary data.

[0179] Further, the optimization suggestion can be used to instruct the sending end to optimize the semantic feature result (that is, the semantic feature) extracted locally. Exemplarily, taking the server as a cloud server as an example, the sending end can transmit the prompt word and necessary data to the cloud server, the cloud server receives the prompt word and necessary data, generates an optimization suggestion for the semantic feature based on the public semantic knowledge base, and transmits the optimization suggestion to the sending end.

[0180] In the embodiments of the present application, by transmitting the prompt word and necessary data to the server, when it is determined that the semantic feature extraction is performed locally, the sending end can integrate the semantic communication target, create and transmit the prompt word to guide the cloud to more accurately understand the semantic communication scenario and the intention of the sending end, so as to realize more accurate optimization suggestion or more accurate semantic feature extraction. The embodiments of the present application make the processing and optimization work of the cloud more closely match the personalized needs of the sending end and the context, thereby improving the accuracy and efficiency of the entire semantic feature extraction process.

[0181] The server is a cloud server, and the sending end determines to perform semantic feature extraction locally. The sending end extracts semantic features based on a local semantic knowledge base. The sending end creates a prompt word based on a semantic communication target and the extracted semantic features. The sending end transmits the prompt word and necessary data to the cloud server. The cloud server receives the prompt word and the necessary data, generates an optimization suggestion for the semantic features based on a public semantic knowledge base, transmits the optimization suggestion to the sending end, and the sending end receives the suggestion from the cloud server and optimizes the semantic feature result. The embodiment of the application utilizes the cooperation of the local semantic knowledge base of the sending end and the public semantic knowledge base of the cloud server to improve the accuracy of semantic feature extraction.

[0182] The above semantic feature extraction method determines the semantic feature extraction method according to the local available resources of the sending end, and can extract and optimize the semantic features with the help of the local semantic knowledge base on the sending end side and the public semantic knowledge base on the cloud side, thereby finally improving the accuracy of semantic feature extraction. In the embodiment of the application, the semantic communication system can be faced, and the end-to-cloud cooperation can be used to assist the sending end to complete semantic feature extraction. Further, by utilizing the cooperation of the local semantic knowledge base of the sending end and the public semantic knowledge base of the cloud server, the extraction and optimization of the semantic features are completed, the accuracy of the semantic feature extraction is improved, the semantic communication system can more effectively transmit information, and thus the quality and efficiency of the whole semantic communication process are improved.

[0183] In one embodiment, the necessary data includes one or more of the semantic features, the original data, the data type, the data format, the data creation time, and the data version.

[0184] Specifically, when it is confirmed that the semantic feature extraction is performed locally, the necessary data can include the semantic features extracted by the sending end. Optionally, the necessary data can also include at least one of the original data, the data type, the data format, the data creation time, and the data version. The original data can be data collected by the sending end through a sensor and desensitized by the device end locally. The data type can be the modality of the data collected by the sending end, such as text, voice, image, video, etc.

[0185] Exemplarily, taking the server as a cloud server, if it is determined that the semantic feature extraction is performed locally, the sending end can extract semantic features based on a local semantic knowledge base, and then create a prompt word based on a semantic communication target and the extracted semantic features. The sending end transmits the prompt word and necessary data to the cloud server, wherein the necessary data can include one or more of the semantic features, the original data, the data type, the data format, the data creation time, and the data version.

[0186] Further, in an exemplary embodiment, the prompt word includes one or more of a communication target description, a semantic feature summary, and a sending end intention.

[0187] Specifically, when it is confirmed that the semantic feature extraction is performed locally, and the sending end extracts the semantic features based on the local semantic knowledge base, for the prompt word (i.e., the first prompt word) created by the sending end based on the semantic communication target and the extracted semantic features, the prompt word can include one or more of a communication target description, a semantic feature summary, and a sending end intent.

[0188] Exemplarily, the communication target description can refer to a description of the semantic communication target; the semantic feature summary can be a semantic-level abstraction and representation of the semantic features, such as keywords, topics, etc.; and the sending end intent can be an operation to be performed by the cloud, information to be understood, or a behavior to be triggered.

[0189] In one embodiment, the presentation mode of the prompt word includes one or more of a text form, a vector form, and structured data.

[0190] Optionally, the presentation mode of the prompt word includes but is not limited to a text form, a vector form, and structured data. For example, the prompt word can be presented in a text form, or can be presented in a vector form or structured data, which is not limited in the present application.

[0191] In some embodiments, the local semantic knowledge base includes one or more of user information, device information, target domain knowledge, and context information.

[0192] Specifically, the local semantic knowledge base of the sending end can include at least one of user information, device information, target domain knowledge, and context information. The user information can refer to information related to user identity and user usage mode; the device information can refer to information related to device attributes, device configuration, and device running; the target domain knowledge can refer to structured knowledge for describing information, concepts, and their mutual relationships of a certain target domain; and the context information can be information related to the application environment of the sending end.

[0193] In one possible implementation, the user information can include one or more of user profile, user preference, user behavior mode, and historical interaction record; the device information can include one or more of hardware configuration, software configuration, performance parameter, and device running state; the target domain knowledge can include one or more of concept definition, entity relationship, rule logic, and domain model; and the context information can include one or more of environment state, situational information, and time information.

[0194] In one embodiment, the semantic features include one or more of entity recognition, sentiment analysis, topic classification, and semantic role labeling.

[0195] Specifically, when it is determined to perform semantic feature extraction locally, the semantic features (i.e., semantic features) extracted by the sending end based on the local semantic knowledge base can include one or more of entity recognition, sentiment analysis, topic classification, and semantic role labeling. Among them, entity recognition is used to represent entities of the target meaning, such as names, place names, organization names, dates and times, and proper nouns; sentiment analysis can refer to identifying the sentiment expressed by the input data (such as positive, negative, or neutral, or positive, negative, and neutral); topic classification can refer to classifying input data into predefined topics or categories; and semantic role labeling can refer to assigning semantic roles to each component in the input data, such as the performer of an action and the recipient of an action.

[0196] Regarding the public semantic knowledge base and the optimization suggestion, in some embodiments, the public semantic knowledge base can include one or more of general knowledge, domain expertise, device interaction knowledge, and a model library; and the optimization suggestion can include one or more of maintaining, expanding, reducing, merging, splitting, refining, and simplifying.

[0197] Specifically, the public semantic knowledge base maintained by the server can include at least one of general knowledge, domain expertise, device interaction knowledge, and a model library; wherein the general knowledge can refer to general knowledge, the domain expertise can refer to professional knowledge of the target domain, and the device interaction knowledge can relate to interaction information between devices, including the functions, interface standards, and communication protocols of the devices. The model library can contain various models and algorithms.

[0198] In one possible implementation, the general knowledge includes one or more of common sense, concept definitions, and general rules; the domain expertise includes one or more of professional terms, cases, regulations, and industry standards; the device interaction knowledge includes one or more of device types, interfaces, communication protocols, and operation instructions; and the model library includes one or more of statistical models, machine learning models, and deep learning models.

[0199] Further, the optimization suggestion in the embodiments of the present application can include at least one of maintaining, expanding, reducing, merging, splitting, refining, and simplifying. After receiving the optimization suggestion transmitted by the server, the sending end can optimize the semantic feature results, for example, by performing operations such as maintaining, expanding, reducing, merging, splitting, refining, and / or simplifying on the semantic features extracted based on the local semantic knowledge base, to improve the accuracy of semantic feature extraction.

[0200] It can be understood that when it is determined to perform semantic feature extraction in the cloud, the sending end can send relevant data (such as indication information) to the server to instruct the server to extract semantic features based on the public semantic knowledge base. In one embodiment, the indication information can include prompt words and necessary data.

[0201] The prompt word and the necessary data transmitted by the sending end to the server can be used to instruct the server to extract semantic features based on a public semantic knowledge base.

[0202] In some embodiments, as shown in FIG. 9, step 206 can include steps 902-904. In step 902, a prompt word is created based on the semantic communication target and the local semantic knowledge base. Figure 9

[0203] In step 902, a prompt word is created based on the semantic communication target and the local semantic knowledge base.

[0204] Specifically, when it is determined that semantic feature extraction is performed in the cloud, the sending end can create a prompt word based on the semantic communication target and the local semantic knowledge base. It can be understood that, for the sake of distinction, the prompt word created by the sending end based on the semantic communication target and the local semantic knowledge base in the embodiments of the present application can be referred to as a second prompt word.

[0205] The semantic communication target refers to target information to be obtained through the current semantic communication in the current application scenario. For example, in an automatic driving scenario, the semantic communication target can be the surrounding environment information, including pedestrians, vehicles, traffic signals, etc., obtained through semantic communication between vehicles and vehicles or between vehicles and traffic infrastructure.

[0206] Optionally, according to the specific application scenario, the semantic communication target can be obtained in the following ways: the semantic communication target can be confirmed by both parties before the formal communication process, or it can be obtained from a server (such as a cloud server).

[0207] In step 904, the prompt word and the necessary data are transmitted to the server to instruct the server to extract semantic features based on a public semantic knowledge base and transmit the semantic features to the sending end.

[0208] Specifically, when it is determined that semantic feature extraction is performed in the cloud, the sending end can send instruction information to the server, wherein the instruction information can include the prompt word and the necessary data. Further, the server can extract semantic features based on a public semantic knowledge base and transmit the semantic features to the sending end when receiving the prompt word and the necessary data.

[0209] It can be understood that, for the sake of distinction, the necessary data used to instruct the server to extract semantic features based on a public semantic knowledge base in the embodiments of the present application can be referred to as second necessary data. The semantic features extracted by the server based on a public semantic knowledge base in the embodiments of the present application can be referred to as second semantic features.

[0210] ​Taking a server as an example, when it is determined to perform semantic feature extraction in the cloud, the sending end creates a prompt word based on the semantic communication target and the local semantic knowledge base; the sending end transmits the prompt word and necessary data to the cloud server, and the cloud server receives the prompt word and necessary data, extracts semantic features based on the public semantic knowledge base; the cloud server transmits the extracted semantic features to the sending end.

[0211] When it is determined to extract semantic features in the cloud, the sending end can integrate the semantic communication target, create and transmit the prompt word to guide the cloud to more accurately understand the semantic communication scene and the sending end intent, and then generate more accurate optimization suggestions or implement more accurate semantic feature extraction. By sending indication information to the cloud, the processing and optimization work of the cloud can be more closely matched with the personalized needs of the sending end and the context, thereby improving the accuracy and efficiency of the entire semantic feature extraction process.

[0212] In one of the embodiments, the necessary data includes at least one of original data, data type, data format, data creation time, and data version.

[0213] Specifically, when it is determined to extract semantic features in the cloud, the necessary data can include one or more of original data, data type, data format, data creation time, and data version. The original data can be data collected by the sending end through a sensor and desensitized by the device end, and the data type can be the modality of the data collected by the sending end, such as text, voice, image, video, etc.

[0214] In some embodiments, the prompt word includes one or more of a communication target description, a communication scene description, a sending end intent, and a personalized label.

[0215] Specifically, when it is determined to extract semantic features in the cloud, the prompt word created by the sending end based on the semantic communication target and the local semantic knowledge base can include at least one of a communication target description, a communication scene description, a sending end intent, and a personalized label, wherein the communication target description can be a description of the semantic communication target, the communication scene description can be a description of the current application scene, the sending end intent can be an operation to be performed by the cloud, information to be understood, or a behavior to be triggered, and the personalized label can be an identifier generated based on the relevant information of the semantic communication target, used to describe the target feature or content attribute.

[0216] In some embodiments, the presentation mode of the prompt word includes one or more of a text form, a vector form, and structured data.

[0217] Specifically, the presentation of the prompt word includes but is not limited to text form, vector form and structured data. For example, the prompt word can be presented in text form, or in vector form or structured data, which is not limited in the present application.

[0218] It should be noted that in the case of semantic feature extraction in the cloud, the form of the above local semantic database and public semantic database can refer to the definition of the local semantic database and public semantic database in the local semantic feature extraction, which will not be repeated here.

[0219] The above semantic feature extraction method, whether in local or cloud semantic feature extraction, the sending end can integrate the semantic communication target, create and transmit the prompt word, to guide the cloud to more accurately understand the semantic communication scene and the sending end intention, so as to realize the generation of more accurate optimization suggestions or the realization of more accurate semantic feature extraction. The embodiments of the present application make the processing and optimization work of the cloud more closely match the personalized needs of the sending end and the context, thereby improving the accuracy and efficiency of the entire semantic feature extraction process.

[0220] In an exemplary embodiment, as shown in Figure 10 , a semantic feature extraction method is provided, which is applied to the server in Figure 1 for example (it can be understood that the server can be a cloud server) includes the following steps 1002 to step 1004. Among them:

[0221] Step 1002, receiving indication information from the sending end.

[0222] Specifically, the server can respond to the received indication information from the sending end to perform semantic feature extraction. Among them, the indication information is sent by the sending end according to the local available resources in the case of semantic feature extraction in the cloud.

[0223] Step 1004, extracting semantic features based on the public semantic knowledge base according to the indication information.

[0224] Specifically, the server can extract semantic features based on the public semantic knowledge base according to the indication information.

[0225] Further, regarding the execution of semantic feature extraction in the cloud, in some embodiments, the indication information is sent by the sending end according to the local available resources in the case of semantic feature extraction in the cloud; wherein the indication information includes the prompt word and the necessary data. In the embodiments of the present application, the server can extract semantic features based on the public semantic knowledge base according to the prompt word and the necessary data sent by the sending end.

[0226] In one embodiment, according to the indication information, semantic features are extracted based on a public semantic knowledge base, including:

[0227] In response to receiving the prompt word and the necessary data, semantic features are extracted based on a public semantic knowledge base;

[0228] The semantic features are transmitted to the sending end.

[0229] Specifically, the server receives the prompt word and the necessary data, extracts semantic features based on a public semantic knowledge base, and transmits the semantic features to the sending end.

[0230] In some embodiments, the necessary data includes at least one of raw data, data type, data format, data creation time, and data version.

[0231] Specifically, regarding the execution of semantic feature extraction in the cloud, the necessary data sent by the sending end to the server can include one or more of raw data, data type, data format, data creation time, and data version.

[0232] In one exemplary embodiment, the prompt word includes one or more of a communication target description, a communication scenario description, a sending end intent, and a personalized label.

[0233] In one embodiment, the presentation mode of the prompt word includes one or more of text form, vector form, and structured data.

[0234] In one exemplary embodiment, the sending end determines to perform semantic feature extraction locally according to locally available resources, and extracts semantic features based on a local semantic knowledge base.

[0235] In one embodiment, the method further includes:

[0236] Sending an optimization suggestion to the sending end; the optimization suggestion is used to instruct the sending end to optimize the semantic features extracted based on the local semantic knowledge base.

[0237] Specifically, when the sending end determines to perform semantic feature extraction locally, the server can send an optimization suggestion to the sending end, and then the sending end optimizes the semantic features extracted based on the local semantic knowledge base.

[0238] In one embodiment, the optimization suggestion can include one or more of maintaining, expanding, reducing, merging, splitting, refining, and simplifying.

[0239] In one embodiment, before sending the optimization suggestion to the sending end, it includes:

[0240] Receiving the prompt word and the necessary data from the sending end;

[0241] According to the prompt word and the necessary data, an optimized suggestion for the semantic feature is generated based on a public semantic knowledge base.

[0242] In some embodiments, the prompt word is created by the sending end based on the semantic communication target and the semantic feature.

[0243] Specifically, when the sending end determines to perform the semantic feature extraction locally, the prompt word sent by the sending end to the server can be created by the sending end based on the semantic communication target and the semantic feature.

[0244] In one embodiment, the prompt word includes one or more of a communication target description, a semantic feature summary, and a sending end intention.

[0245] In some embodiments, the presentation mode of the prompt word includes one or more of a text form, a vector form, and structured data.

[0246] In one embodiment, the necessary data includes one or more of a semantic feature, original data, a data type, a data format, a data creation time, and a data version.

[0247] Specifically, when the sending end determines to perform the semantic feature extraction locally, the necessary data sent by the sending end to the server can include one or more of a semantic feature, original data, a data type, a data format, a data creation time, and a data version, wherein the semantic feature included in the necessary data is a semantic feature extracted by the sending end based on a local semantic knowledge base.

[0248] In one embodiment, the local semantic knowledge base includes one or more of user information, device information, target domain knowledge, and context information.

[0249] In one embodiment, the user information includes one or more of user profile, user preference, user behavior pattern, and historical interaction record; the device information includes one or more of hardware configuration, software configuration, performance parameter, and device running state; the target domain knowledge includes one or more of concept definition, entity relationship, rule logic, and domain model; and the context information includes one or more of environment state, situational information, and time information.

[0250] In one embodiment, the semantic feature includes one or more of entity recognition, sentiment analysis, topic classification, and semantic role labeling.

[0251] In one embodiment, the public semantic knowledge base includes one or more of general knowledge, domain professional knowledge, device interaction knowledge, and model library.

[0252] In one embodiment, the general knowledge comprises one or more of common sense, concept definitions, and general rules; the domain expertise comprises one or more of professional terms, cases, regulations, and industry standards; the device interaction knowledge comprises one or more of device types, interfaces, communication protocols, and operation instructions; and the model library comprises one or more of statistical models, machine learning models, and deep learning models.

[0253] In one embodiment, the method further comprises:

[0254] transmitting the performance index of completing the semantic feature extraction to the cloud to the sender; the performance index of completing the semantic feature extraction to the cloud is used to instruct the sender to compare the performance index of completing the semantic feature extraction locally and the performance index of completing the semantic feature extraction to the cloud, and determine to execute the semantic feature extraction locally or to the cloud according to the comparison result, in the case that the sender determines that the local has the ability to complete the semantic feature extraction according to the local available resources and the required resources of the semantic feature extraction.

[0255] In one embodiment, the performance index of completing the semantic feature extraction to the cloud comprises a cloud completion time, and a data transmission time and a data transmission energy consumption of transmitting the to-be-processed data to the server; the method further comprises:

[0256] receiving data volume information of the to-be-processed data from the sender;

[0257] calculating the cloud completion time, the data transmission time and the data transmission energy consumption according to the data volume information, and transmitting the cloud completion time, the data transmission time and the data transmission energy consumption to the sender.

[0258] In one embodiment, the data volume information comprises at least one of data size and data type.

[0259] It can be understood that the specific limitations in the above-mentioned semantic feature extraction method embodiment executed by the server can refer to the limitations of the semantic feature extraction method executed by the sender described above, which will not be described here.

[0260] In one embodiment, Figure 11 A flow chart of interaction between a cloud server and a sender (such as a terminal) in a semantic feature extraction method is provided, as shown in Figure 12 The flow chart can include the following steps.

[0261] 1) The sender evaluates the resource requirement of the semantic feature extraction and the local available resources. The sender determines whether the local has the ability to complete the semantic feature extraction according to the evaluation result; wherein the sender can determine whether the local has the ability to complete the semantic feature extraction locally by evaluating the local available resources, channel resources and resource requirement of the semantic feature extraction.

[0262] 2) If the sender has the ability to complete semantic feature extraction locally, the sender can compare the performance indicators of local semantic feature extraction and the performance indicators of semantic feature extraction offloaded to the cloud, and decide whether to perform semantic feature extraction locally or in the cloud according to the comparison result. The comparison result can indicate whether the performance indicators of local semantic feature extraction are better than the performance indicators of semantic feature extraction offloaded to the cloud.

[0263] 3) If semantic feature extraction is performed locally at the sender, then:

[0264] a) The sender extracts semantic features based on the local semantic knowledge base;

[0265] b) The sender creates a prompt word based on the semantic communication target and the extracted semantic features;

[0266] c) The sender transmits the prompt word and necessary data to the cloud server;

[0267] d) The cloud server receives the prompt word and necessary data, generates optimization suggestions for the semantic features based on the public semantic knowledge base;

[0268] e) The sender receives the optimization suggestions of the cloud server and optimizes the semantic features.

[0269] 4) If semantic feature extraction is performed locally in the cloud, then:

[0270] a) The sender creates a prompt word based on the semantic communication target and the local semantic knowledge base;

[0271] b) The sender transmits the prompt word and necessary data to the cloud server;

[0272] c) The cloud server receives the prompt word and necessary data, extracts semantic features based on the public semantic knowledge base;

[0273] d) The cloud server transmits the extracted semantic features to the sender.

[0274] The following is described with reference to a specific example. In an autonomous driving scenario, vehicle A extracts semantic features based on a local semantic knowledge base, including time, weather, road type, road condition, vehicle condition, traffic signal, and the like. Vehicle A creates a prompt word based on the semantic communication target and the extracted semantic features: "current location is a certain highway branch, no vehicle in front of the road, reflector on the left side of the road is interrupted, whether it can continue to drive normally". Vehicle A sends the prompt word and the road photo to the cloud server. The cloud server receives the prompt word and the road photo, and generates an optimization suggestion for the semantic features based on a public semantic knowledge base: "collapse appears on the left side of the road in front". The cloud server sends the optimization suggestion to vehicle A.

[0275] It should be understood that, although each step in the flowchart involved in each of the above-described embodiments is shown in sequence according to the direction of the arrow, these steps are not necessarily executed in sequence according to the direction of the arrow. Unless otherwise specified herein, the execution of these steps is not strictly limited in sequence, and these steps can be executed in other sequences. Moreover, at least some of the steps in the flowchart involved in each of the above-described embodiments can include multiple steps or stages, which are not necessarily executed at the same time, but can be executed at different times, and the execution sequence of these steps or stages is not necessarily sequential, but can be executed alternately or alternately with at least some of the other steps or steps or stages in other steps.

[0276] Based on the same inventive concept, the embodiments of the present application also provide a semantic feature extraction device for implementing the above-mentioned semantic feature extraction method. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme described in the above method, so the specific limitations in one or more semantic feature extraction device embodiments provided below can refer to the limitations of the semantic feature extraction method described above, which will not be repeated here.

[0277] In one exemplary embodiment, as shown in Figure 13 a semantic feature extraction device is provided, applied to a sending end, and the device includes:

[0278] The extraction mode determination module 1101 is configured to determine whether to perform semantic feature extraction locally or in the cloud according to the locally available resources.

[0279] The local extraction module 1102 is configured to, if it is determined to perform semantic feature extraction locally, extract semantic features based on a local semantic knowledge base.

[0280] The indication module 1103 is configured to send indication information to the server if it is determined that the semantic feature extraction is performed in the cloud; the indication information is used to instruct the server to extract the semantic feature based on the public semantic knowledge base.

[0281] In one of the embodiments, the extraction mode determination module 1101 comprises:

[0282] The capability determination module is configured to determine whether the local device has the capability to complete the semantic feature extraction according to the local available resources and the required resources for the semantic feature extraction; and

[0283] The index comparison module is configured to compare the performance index of the local completion of the semantic feature extraction and the performance index of the cloud completion of the semantic feature extraction if the local device has the capability to complete the semantic feature extraction, and determine whether to perform the semantic feature extraction in the local device or in the cloud according to the comparison result.

[0284] In one of the embodiments, the capability determination module is configured to evaluate the to-be-processed data, obtain data volume information of the to-be-processed data, and evaluate resource requirement of the local semantic feature extraction algorithm according to the data volume information, wherein the resource requirement is used as the required resources for the semantic feature extraction; and if the local available resources are greater than the required resources for the semantic feature extraction, it is determined that the local device has the capability to complete the semantic feature extraction.

[0285] In one of the embodiments, the data volume information comprises at least one of data size and data type; and the resource requirement comprises at least one of calculation requirement and memory requirement.

[0286] In one of the embodiments, the required resources for the semantic feature extraction comprise semantic feature extraction calculation requirement and semantic feature extraction memory requirement.

[0287] The capability determination module is configured to obtain the local available resources, wherein the local available resources comprise local available calculation resources and local available memory space; and if the local available calculation resources are greater than the semantic feature extraction calculation requirement and the local available memory space is greater than the semantic feature extraction memory requirement, it is determined that the local device has the capability to complete the semantic feature extraction.

[0288] In one of the embodiments, the performance index of the local completion of the semantic feature extraction comprises local completion time and local completion energy consumption; and the performance index of the cloud completion of the semantic feature extraction comprises cloud completion time, data transmission time and data transmission energy consumption of the to-be-processed data transmitted to the server.

[0289] The index comparison module is configured to transmit data volume information to the server, wherein the data volume information is used to instruct the server to calculate and feed back cloud-side completion time, data transmission time and data transmission energy consumption; if local completion time is less than the sum of data transmission time and cloud-side completion time, and local completion energy consumption is less than data transmission energy consumption, it is determined that semantic feature extraction is performed locally; if local completion time is greater than or equal to the sum of data transmission time and cloud-side completion time, and / or local completion energy consumption is greater than or equal to data transmission energy consumption, it is determined that semantic feature extraction is performed in the cloud.

[0290] In one embodiment, the local extraction module 1102 includes:

[0291] The local feature extraction module is configured to extract semantic features based on a local semantic knowledge base.

[0292] The optimization module is configured to receive optimization suggestions from the server and optimize the semantic features according to the optimization suggestions.

[0293] In one embodiment, the local extraction module 1102 further includes:

[0294] The prompt creation module is configured to create a prompt word based on a semantic communication target and semantic features.

[0295] The data transmission module is configured to transmit the prompt word and necessary data to the server to instruct the server to generate optimization suggestions for the semantic features based on a public semantic knowledge base, and transmit the optimization suggestions to the sending end.

[0296] In one embodiment, the necessary data includes one or more of semantic features, original data, data type, data format, data creation time and data version.

[0297] In one embodiment, the prompt word includes one or more of a communication target description, a semantic feature summary and a sending end intention.

[0298] In one embodiment, the presentation mode of the prompt word includes one or more of a text form, a vector form and structured data.

[0299] In one embodiment, the local semantic knowledge base includes one or more of user information, device information, target domain knowledge and context information.

[0300] In one of the embodiments, the user information includes one or more of user profile, user preference, user behavior pattern, and historical interaction record; the device information includes one or more of hardware configuration, software configuration, performance parameter, and device running status; the target domain knowledge includes one or more of concept definition, entity relationship, rule logic, and domain model; and the context information includes one or more of environment status, situational information, and time information.

[0301] In one of the embodiments, the semantic features include one or more of entity recognition, sentiment analysis, topic classification, and semantic role labeling.

[0302] In one of the embodiments, the public semantic knowledge base includes one or more of general knowledge, domain expertise, device interaction knowledge, and model library.

[0303] In one of the embodiments, the general knowledge includes one or more of common sense, concept definition, and general rule; the domain expertise includes one or more of professional term, case, regulation, and industry standard; the device interaction knowledge includes one or more of device type, interface, communication protocol, and operation instruction; and the model library includes one or more of statistical model, machine learning model, and deep learning model.

[0304] In one of the embodiments, the optimization suggestion includes one or more of maintaining, extending, pruning, merging, splitting, refining, and simplifying.

[0305] In one of the embodiments, the indication information includes prompt word and necessary data.

[0306] In one of the embodiments, the indication module 1003 includes:

[0307] a prompt word creation module configured to create a prompt word based on the semantic communication target and the local semantic knowledge base;

[0308] a transmission module configured to transmit the prompt word and the necessary data to a server, so as to instruct the server to extract semantic features based on a public semantic knowledge base and transmit the semantic features to the sending end.

[0309] In one of the embodiments, the necessary data includes at least one of original data, data type, data format, data creation time, and data version.

[0310] In one of the embodiments, the prompt word includes one or more of communication target description, communication scenario description, sending end intention, and personalized label.

[0311] The presentation mode of the prompt word includes one or more of text form, vector form, and structured data.

[0312] In one example embodiment, as shown in Figure 14 A semantic feature extraction apparatus is provided, applied to a server, and the apparatus comprises:

[0313] An information receiving module 1201 is configured to receive indication information from a sending end.

[0314] A feature extraction module 1202 is configured to extract semantic features based on a public semantic knowledge base according to the indication information.

[0315] In one example embodiment, the indication information is sent by the sending end according to local available resources in a case where the sending end determines to perform semantic feature extraction in the cloud.

[0316] In one example embodiment, the feature extraction module 1102 is configured to extract semantic features based on the public semantic knowledge base in response to receiving the prompt word and the necessary data, and transmit the semantic features to the sending end.

[0317] In one example embodiment, the necessary data comprises at least one of original data, data type, data format, data creation time and data version.

[0318] In one example embodiment, the prompt word comprises one or more of a communication target description, a communication scenario description, a sending end intention and a personalized label.

[0319] In one example embodiment, the presentation mode of the prompt word comprises one or more of a text form, a vector form and structured data.

[0320] In one example embodiment, the sending end extracts semantic features based on a local semantic knowledge base in a case where the sending end determines to perform semantic feature extraction locally according to local available resources.

[0321] In one example embodiment, the apparatus further comprises:

[0322] An optimization module is configured to send an optimization suggestion to the sending end, and the optimization suggestion is used to instruct the sending end to optimize the semantic features extracted based on the local semantic knowledge base.

[0323] In one example embodiment, the optimization module comprises:

[0324] A data receiving module is configured to receive a prompt word and necessary data from the sending end.

[0325] An optimization suggestion generation module is configured to generate an optimization suggestion for the semantic features based on the public semantic knowledge base according to the prompt word and the necessary data.

[0326] In one example embodiment, the prompt word is created by the sending end based on a semantic communication target and semantic features.

[0327] In one of the embodiments, the necessary data includes one or more of semantic features, raw data, data type, data format, data creation time, and data version.

[0328] In one of the embodiments, the prompt words include one or more of communication target description, semantic feature summary, and sender intent.

[0329] In one of the embodiments, the presentation of the prompt words includes one or more of text form, vector form, and structured data.

[0330] In one of the embodiments, the semantic features include one or more of entity recognition, sentiment analysis, topic classification, and semantic role labeling.

[0331] In one of the embodiments, the optimization suggestions include one or more of maintaining, expanding, pruning, merging, splitting, refining, and simplifying.

[0332] In one of the embodiments, the local semantic knowledge base includes one or more of user information, device information, target domain knowledge, and context information.

[0333] In one of the embodiments, the user information includes one or more of user profile, user preference, user behavior pattern, and historical interaction record; the device information includes one or more of hardware configuration, software configuration, performance parameter, and device running state; the target domain knowledge includes one or more of concept definition, entity relationship, rule logic, and domain model; and the context information includes one or more of environment state, situational information, and time information.

[0334] In one of the embodiments, the public semantic knowledge base includes one or more of general knowledge, domain-specific knowledge, device interaction knowledge, and model library.

[0335] In one of the embodiments, the general knowledge includes one or more of common sense, concept definition, and general rule; the domain-specific knowledge includes one or more of professional terminology, case, regulation, and industry standard; the device interaction knowledge includes one or more of device type, interface, communication protocol, and operation instruction; and the model library includes one or more of statistical model, machine learning model, and deep learning model.

[0336] In one of the embodiments, the apparatus further includes:

[0337] The index transmission module is configured to transmit the performance index of the semantic feature extraction completed by offloading to the cloud to the sending end; the performance index of the semantic feature extraction completed by offloading to the cloud is used to instruct the sending end to compare the performance index of the semantic feature extraction completed locally and the performance index of the semantic feature extraction completed by offloading to the cloud in the case that the sending end has the capability to complete the semantic feature extraction locally according to the local available resources and the required resources for the semantic feature extraction, and determine whether to perform the semantic feature extraction locally or by offloading to the cloud according to the comparison result.

[0338] In one of the embodiments, the performance index of the semantic feature extraction completed by offloading to the cloud includes a cloud completion time, and a data transmission time and a data transmission energy consumption of transmitting the data to be processed to the server;

[0339] The index transmission module is configured to receive data volume information of the data to be processed from the sending end; calculate the cloud completion time, the data transmission time and the data transmission energy consumption according to the data volume information, and transmit the cloud completion time, the data transmission time and the data transmission energy consumption to the sending end.

[0340] In one of the embodiments, the data volume information includes at least one of a data size and a data type.

[0341] The modules in the semantic feature extraction apparatus can be realized by software, hardware or a combination thereof. The modules can be embedded in or independent of a processor in a computer device in hardware form, or stored in a memory in the computer device in software form, so as to be called and executed by a processor to perform the operations corresponding to the modules.

[0342] In one of the embodiments, a sending end is provided, which is taken as a terminal device (terminal for short) in the embodiment, and the structure of the terminal device is shown in Figure 15 . Figure 15 is a structure diagram of the terminal device provided in the embodiment. Figure 15 The terminal device 700 shown in the figure includes at least one processor 701, a memory 707, at least one network interface 708 and a user interface 703. The components in the terminal device 700 are coupled together through a bus system 705. It can be understood that the bus system 705 is used to realize the connection and communication between the components. The bus system 705 includes a data bus, a power supply bus, a control bus and a state signal bus. However, for the purpose of clear illustration, all the buses are marked as the bus system 705 in the figure. In addition, a transceiver 706 is also included in the embodiment, which can be multiple elements, i.e., including a transmitter and a receiver, and providing a unit for communicating with various other devices on a transmission medium. Figure 15

[0343] ​The user interface 703 can include a display, a keyboard, or a pointing device (e.g., a mouse, a trackball, a touchpad, or a touchscreen).

[0344] It can be appreciated that the memory 707 in the embodiments of the present application can be a volatile memory or a nonvolatile memory, or can include both volatile and nonvolatile memory. The nonvolatile memory can be a Read-Only Memory (ROM), a Programmable ROM (PROM), an Erasable PROM (EPROM), an Electrically EPROM (EEPROM), or a flash memory. The volatile memory can be a Random Access Memory (RAM), which is used as the external cache. By way of example, and not limitation, many forms of RAM are available, such as Static RAM (SRAM), Dynamic RAM (DRAM), Synchronous DRAM (SDRAM), Double Data Rate SDRAM (DDR SDRAM), Enhanced SDRAM (ESDRAM), Synchlink DRAM (SLDRAM), and Direct Rambus RAM (DRRAM). The memory 707 of the system and method described in the embodiments of the present application is intended to include, without being limited to, these and any other suitable types of memory.

[0345] In some embodiments, the memory 707 stores the following elements, executable modules or data structures, or a subset thereof, or an extended set thereof: an operating system 7071 and an application program 7072.

[0346] The operating system 7071 includes various system programs, such as a framework layer, a core library layer, a driver layer, and the like, for implementing various basic services and processing hardware-based tasks. The application program 7072 includes various application programs, such as a media player (Media Player), a browser (Browser), and the like, for implementing various application services. The program for implementing the method of the embodiments of the present application can be included in the application program 7072.

[0347] In the embodiments of the present application, the processor 701 is configured to determine whether to perform semantic feature extraction locally or in the cloud according to locally available resources, extract semantic features based on a local semantic knowledge base if it is determined to perform semantic feature extraction locally, and control the transmitter to send indication information to a server if it is determined to perform semantic feature extraction in the cloud, where the indication information is used to instruct the server to extract semantic features based on a public semantic knowledge base.

[0348] Part or all of the methods disclosed in the embodiments of the present application can also be applied to the processor 701, or implemented by the processor 701 or in cooperation with other elements (such as a transceiver). The processor 701 can be an integrated circuit chip with a signal processing capability. In the implementation process, the steps of the above methods can be completed by integrated logic circuits or software form instructions in the processor 701. The processor 701 can be a general-purpose processor 701, a digital signal processor 701 (DSP), an application specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components. The disclosed methods, steps and logic block diagrams in the embodiments of the present application can be implemented or executed. The general-purpose processor 701 can be a microprocessor or any conventional processor 701. The steps of the methods disclosed in conjunction with the embodiments of the present application can be directly embodied as a hardware code processor 701 for execution, or a combination of hardware and software modules in the code processor 701 for execution. The software module can be located in a random access memory, a flash memory, a read-only memory, a programmable read-only memory or an electrically erasable programmable memory, a register or other mature storage medium in the art. The storage medium is located in the memory 707, and the processor 701 reads the information in the memory 707 and completes the steps of the above method in cooperation with the hardware.

[0349] It can be understood that the embodiments described in the embodiments of the application can be implemented by hardware, software, firmware, middleware, microcode or a combination thereof. For hardware implementation, the processing unit can be implemented in one or more application specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), general-purpose processors 701, controllers, micro-controllers, microprocessors 701, other electronic units for performing the functions described in the application, or a combination thereof.

[0350] For software implementation, the technologies described in the embodiments of the application can be implemented by modules (such as processes, functions, etc.) for performing the functions described in the embodiments of the application. The software code can be stored in the memory and executed by the processor 701. The memory can be implemented in the processor 701 or outside the processor 701.

[0351] In one embodiment, the processor 701 is specifically configured to extract required resources according to local available resources and semantic feature extraction, determine whether the local has the ability to complete the semantic feature extraction; if the local has the ability to complete the semantic feature extraction, compare the performance index of completing the semantic feature extraction locally and the performance index of completing the semantic feature extraction by offloading to the cloud, and determine to execute the semantic feature extraction locally or in the cloud according to the comparison result.

[0352] In one embodiment, the processor 701 is specifically configured to evaluate the to-be-processed data, obtain data volume information of the to-be-processed data; according to the data volume information, evaluate resource demand of running a local semantic feature extraction algorithm, take the resource demand as required resources for semantic feature extraction; if the available resources of the local are greater than the required resources for semantic feature extraction, determine that the local has the ability to complete the semantic feature extraction.

[0353] In one embodiment, the data volume information includes at least one of data size and data type; the resource demand includes at least one of computing demand and memory demand.

[0354] In one embodiment, the required resources for semantic feature extraction include semantic feature extraction computing demand and semantic feature extraction memory demand.

[0355] The processor 701 is specifically configured to acquire a local available resource; the local available resource includes a local available computing resource and a local available memory space; if the local available computing resource is greater than a semantic feature extraction computing requirement and the local available memory space is greater than a semantic feature extraction memory requirement, it is determined that the local has the capability to complete the semantic feature extraction.

[0356] In one embodiment, the performance indicators of the local completion of the semantic feature extraction include a local completion time and a local completion energy consumption; the performance indicators of the semantic feature extraction completed by the cloud include a cloud completion time, a data transmission time and a data transmission energy consumption of transmitting the data to be processed to the server;

[0357] The processor 701 is specifically configured to control the transmitter to transmit data amount information to the server; the data amount information is used to instruct the server to calculate and feed back the cloud completion time, the data transmission time and the data transmission energy consumption; if the local completion time is less than a sum of the data transmission time and the cloud completion time and the local completion energy consumption is less than the data transmission energy consumption, it is determined to execute the semantic feature extraction locally; if the local completion time is greater than or equal to the sum of the data transmission time and the cloud completion time and / or the local completion energy consumption is greater than or equal to the data transmission energy consumption, it is determined to execute the semantic feature extraction in the cloud.

[0358] In one embodiment, the processor 701 is specifically configured to extract the semantic feature based on a local semantic knowledge base, and receive an optimization suggestion from the server through the receiver and optimize the semantic feature according to the optimization suggestion.

[0359] In one embodiment, the processor 701 is further configured to create a prompt word based on the semantic communication target and the semantic feature, and control the transmitter to transmit the prompt word and necessary data to the server to instruct the server to generate an optimization suggestion for the semantic feature based on a public semantic knowledge base and transmit the optimization suggestion to the sending end.

[0360] In one embodiment, the necessary data includes one or more of the semantic feature, the original data, the data type, the data format, the data creation time and the data version.

[0361] In one embodiment, the prompt word includes one or more of a communication target description, a semantic feature summary and a sending end intention.

[0362] In one embodiment, the presentation mode of the prompt word includes one or more of a text form, a vector form and structured data.

[0363] In one embodiment, the semantic feature includes one or more of entity recognition, sentiment analysis, topic classification and semantic role labeling.

[0364] In an embodiment, the optimization suggestions include one or more of maintaining, expanding, pruning, merging, splitting, refining, and simplifying.

[0365] In an embodiment, the local semantic knowledge base includes one or more of user information, device information, target domain knowledge, and context information.

[0366] In an embodiment, the user information includes one or more of user profile, user preference, user behavior pattern, and historical interaction record; the device information includes one or more of hardware configuration, software configuration, performance parameter, and device running state; the target domain knowledge includes one or more of concept definition, entity relationship, rule logic, and domain model; and the context information includes one or more of environment state, situational information, and time information.

[0367] In an embodiment, the public semantic knowledge base includes one or more of general knowledge, domain expertise, device interaction knowledge, and model library.

[0368] In an embodiment, the general knowledge includes one or more of common sense, concept definition, and general rule; the domain expertise includes one or more of professional terminology, case, regulation, and industry standard; the device interaction knowledge includes one or more of device type, interface, communication protocol, and operation instruction; and the model library includes one or more of statistical model, machine learning model, and deep learning model.

[0369] In an embodiment, the indication information includes a prompt word and necessary data.

[0370] In an embodiment, the processor 701 is specifically configured to create a prompt word based on the semantic communication target and the local semantic knowledge base; and control the transmitter to transmit the prompt word and the necessary data to the server, so as to instruct the server to extract semantic features based on the public semantic knowledge base and transmit the semantic features to the sending end.

[0371] In an embodiment, the necessary data includes at least one of original data, data type, data format, data creation time, and data version.

[0372] In an embodiment, the prompt word includes one or more of communication target description, communication scenario description, sending end intention, and personalized label.

[0373] In an embodiment, the presentation mode of the prompt word includes one or more of text form, vector form, and structured data.

[0374] Figure 16 FIG. 1 is a structural schematic diagram of a server provided by an embodiment of the present application. Figure 16The illustrated server 800 includes at least one processor 801, a memory 803, and at least one network interface 805. The various components of server 800 are coupled together by a bus system 807, which is used for communicating information between the components. The bus system 807 includes a power bus, a control bus, and a status bus, among others. For clarity, the various buses are illustrated as a single bus system 807. However, the Figure 16 bus system 807 can include a number of busses that are configured to carry different information in different manners. In some embodiments, the server 800 includes a transceiver 809, which can be a number of components, including a transmitter and a receiver, for communicating with various other apparatuses over a transmission medium.

[0375] It is to be understood that the memory 803 of the present embodiments can be volatile or nonvolatile memory, or can include both volatile and nonvolatile memory. In one embodiment, the nonvolatile memory can be read-only memory (ROM), programmable ROM (PROM), erasable PROM (EPROM), electrically EPROM (EEPROM), or flash memory. The volatile memory can be random access memory (RAM), which is used as external cache. By way of example, and not limitation, many forms of RAM are available, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous dynamic RAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), Synchlink DRAM (SLDRAM), and direct Rambus RAM (DRRAM). The memory 803 of the system and method described herein is intended to include, without being limited to, these and any other suitable types of memory.

[0376] In some embodiments, the memory 803 stores the following elements, executable modules, or data structures, or a subset thereof, or their extensions: an operating system 8031. The operating system 8031 contains various system programs, such as a framework layer, a core library layer, a driver layer, etc., for implementing various basic services and processing hardware-based tasks.

[0377] In the embodiments of the present application, the receiver is configured to receive the indication information from the sending end by invoking the program or instruction stored in the memory 803; and the processor 801 is configured to extract the semantic features based on the public semantic knowledge base according to the indication information.

[0378] Part or all of the methods disclosed in the embodiments of the present application can also be applied to the processor 801, or implemented by the processor 801, or implemented by the processor 801 in cooperation with other elements (such as the transceiver). The processor 801 can be an integrated circuit chip with a signal processing capability. In the implementation process, the steps of the above methods can be completed by the integrated logic circuits or the instructions in the software form in the processor 801. The processor 801 mentioned above can be a general purpose processor, a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components. The methods, steps and logical block diagrams disclosed in the embodiments of the present application can be implemented or executed. The general purpose processor can be a microprocessor or the processor can also be any conventional processor. The steps of the methods disclosed in conjunction with the embodiments of the present application can be directly embodied as a hardware code processor to execute, or a combination of hardware and software modules in the code processor to execute. The software module can be located in a random access memory, a flash memory, a read only memory, a programmable read only memory or an electrically erasable programmable memory, a register, or other mature storage medium in the art. The storage medium is located in the memory 803, and the processor 801 reads the information in the memory 803 and completes the steps of the above method in combination with the hardware.

[0379] It can be understood that the embodiments described in the embodiments of the application can be implemented by hardware, software, firmware, middleware, microcode or a combination thereof. For hardware implementation, the processing unit can be implemented in one or more application specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), general purpose processors, controllers, microcontrollers, microprocessors, other electronic units for performing the functions described in the embodiments of the application, or a combination thereof.

[0380] For software implementation, the technologies described in the embodiments of the application can be implemented by modules (such as processes, functions, etc.) for performing the functions described in the embodiments of the application. The software code can be stored in the memory and executed by the processor 801. The memory can be implemented in the processor 801 or outside the processor 801.

[0381] In one of the embodiments, the indication information is sent by the sending end in the case of determining to perform semantic feature extraction in the cloud according to the locally available resources; the indication information includes a prompt word and necessary data.

[0382] In one of the embodiments, the processor 801 is further configured to extract semantic features based on a public semantic knowledge base in response to receiving the prompt word and the necessary data.

[0383] The transmitter is configured to transmit the semantic features to the sending end.

[0384] In one of the embodiments, the necessary data includes at least one of original data, data type, data format, data creation time and data version.

[0385] In one of the embodiments, the prompt word includes one or more of a communication target description, a communication scene description, a sending end intention and a personalized label.

[0386] In one of the embodiments, the presentation mode of the prompt word includes one or more of a text form, a vector form and structured data.

[0387] In one of the embodiments, the transmitter is further configured to send an optimization suggestion to the sending end; the optimization suggestion is used to instruct the sending end to optimize the semantic features extracted based on the local semantic knowledge base.

[0388] In an embodiment, the receiver is further configured to receive the prompt word and the necessary data from the sender.

[0389] The processor 801 is further configured to generate the optimization suggestion for the semantic feature based on the common semantic knowledge base according to the prompt word and the necessary data.

[0390] In an embodiment, the semantic feature is extracted by the sender based on a local semantic knowledge base; and the prompt word is created by the sender based on the semantic communication target and the semantic feature.

[0391] In an embodiment, the semantic feature includes one or more of entity recognition, sentiment analysis, topic classification, and semantic role labeling.

[0392] In an embodiment, the necessary data includes one or more of the semantic feature, raw data, data type, data format, data creation time, and data version.

[0393] In an embodiment, the prompt word includes one or more of a communication target description, a semantic feature summary, and a sender intent.

[0394] In an embodiment, the presentation mode of the prompt word includes one or more of a text form, a vector form, and structured data.

[0395] In an embodiment, the optimization suggestion includes one or more of maintaining, expanding, reducing, merging, splitting, refining, and simplifying.

[0396] In an embodiment, the local semantic knowledge base includes one or more of user information, device information, target domain knowledge, and context information.

[0397] In an embodiment, the user information includes one or more of user profile, user preference, user behavior pattern, and historical interaction record; the device information includes one or more of hardware configuration, software configuration, performance parameter, and device running state; the target domain knowledge includes one or more of concept definition, entity relationship, rule logic, and domain model; and the context information includes one or more of environment state, situational information, and time information.

[0398] In an embodiment, the common semantic knowledge base includes one or more of general knowledge, domain-specific knowledge, device interaction knowledge, and model library.

[0399] In an embodiment, the general knowledge comprises one or more of common sense, concept definitions, and general rules; the domain expertise comprises one or more of professional terms, cases, regulations, and industry standards; the device interaction knowledge comprises one or more of device types, interfaces, communication protocols, and operation instructions; and the model library comprises one or more of statistical models, machine learning models, and deep learning models.

[0400] In an embodiment, the transmitter is further configured to transmit, to the sender, a performance indicator of performing the semantic feature extraction on the cloud; the performance indicator of performing the semantic feature extraction on the cloud is used to instruct the sender to compare the performance indicator of performing the semantic feature extraction locally and the performance indicator of performing the semantic feature extraction on the cloud, and determine to perform the semantic feature extraction locally or on the cloud according to a comparison result.

[0401] In an embodiment, the performance indicator of performing the semantic feature extraction on the cloud comprises a cloud completion time, a data transmission time of transmitting the to-be-processed data to the server, and a data transmission energy consumption.

[0402] The receiver is further configured to receive data volume information of the to-be-processed data from the sender.

[0403] The processor 801 is specifically configured to calculate the cloud completion time, the data transmission time, and the data transmission energy consumption according to the data volume information.

[0404] The transmitter is further configured to transmit, to the sender, the cloud completion time, the data transmission time, and the data transmission energy consumption.

[0405] In an embodiment, the data volume information comprises at least one of a data scale and a data type.

[0406] In an embodiment, a computer readable storage medium is provided, and the computer readable storage medium stores a computer program, which is executed by a processor to implement the steps in the above method embodiments.

[0407] In an embodiment, a computer program product is provided, and the computer program product comprises a computer program, which is executed by a processor to implement the steps in the above method embodiments.

[0408] It should be noted that the user information (including but not limited to user equipment information, user personal information, etc.) and data (including but not limited to data for analysis, stored data, displayed data, etc.) involved in the present application are all information and data authorized by the user or authorized by all parties, and the collection, use, and processing of the related data need to comply with relevant regulations.

[0409] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a non-volatile computer readable storage medium, and when executed, can include the processes of the above-mentioned embodiment methods. Any reference to memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile memory and volatile memory. The non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical storage, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. The volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration but not limitation, the RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM), etc. The database involved in the embodiments provided in the present application can include at least one of a relational database and a non-relational database. The non-relational database can include a distributed database based on a block chain, etc., without being limited thereto. The processor involved in the embodiments provided in the present application can be a general-purpose processor, a central processing unit, a graphics processing unit, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, an artificial intelligence (AI) processor, etc., without being limited thereto.

[0410] The technical features of the above embodiments can be combined in any manner. To make the description concise, not all possible combinations of the technical features in the above embodiments are described, but as long as the combinations of the technical features do not exist contradictions, they should be considered as the scope of the present application.

[0411] The above-described embodiments are merely illustrative of several embodiments of the present application, and the description is relatively specific and detailed, but should not be understood as a limitation on the scope of the patent. It should be noted that for those skilled in the art, without departing from the concept of the present application, a number of modifications and improvements can be made, which are all within the scope of the present application. Therefore, the scope of protection of the present application should be subject to the appended claims.

Claims

1. A semantic feature extraction method, characterized by, Applied to the sending end, the method includes: Based on available local resources, determine whether to perform semantic feature extraction locally or in the cloud; If it is determined that semantic feature extraction will be performed locally, then semantic features will be extracted based on the local semantic knowledge base; If it is determined that semantic feature extraction will be performed in the cloud, then an instruction message is sent to the server; the instruction message is used to instruct the server to extract semantic features based on a public semantic knowledge base.

2. The method of claim 1, wherein, The step of determining whether to perform semantic feature extraction locally or in the cloud based on available local resources includes: Based on the available local resources and the resources required for semantic feature extraction, confirm whether the local system has the capability to complete semantic feature extraction. If the local machine has the capability to perform semantic feature extraction, then the performance metrics of performing semantic feature extraction locally and performing semantic feature extraction offloaded to the cloud are compared, and based on the comparison results, it is determined whether to perform semantic feature extraction locally or in the cloud.

3. The method of claim 1, wherein, If it is determined that semantic feature extraction will be performed locally, then semantic features will be extracted based on the local semantic knowledge base, including: Based on the local semantic knowledge base, semantic features are extracted; Receive optimization suggestions from the server, and optimize the semantic features according to the optimization suggestions.

4. The method of claim 3, wherein, Before the step of receiving optimization suggestions from the server and optimizing the semantic features according to the optimization suggestions, the method further includes: Based on the semantic communication objective and the semantic features, prompt words are created; The prompt words and necessary data are transmitted to the server to instruct the server to generate optimization suggestions for the semantic features based on the public semantic knowledge base, and then transmit the optimization suggestions to the sending end.

5. The method of claim 1, wherein, The instruction information includes prompts and necessary data.

6. The method of claim 5, wherein, If it is determined that semantic feature extraction will be performed in the cloud, then sending an instruction message to the server includes: The prompt words are created based on the semantic communication objective and the local semantic knowledge base; The prompt word and the necessary data are transmitted to the server, instructing the server to extract semantic features based on the public semantic knowledge base, and then transmit the semantic features to the sending end.

7. The method of claim 2, wherein, The step of confirming whether the local system has the capability to complete semantic feature extraction based on the available local resources and the resources required for semantic feature extraction includes: The data to be processed is evaluated, and the data volume information of the data to be processed is obtained; Based on the data volume information, the resource requirements for running the local semantic feature extraction algorithm are assessed, and the resource requirements are taken as the resources required for the semantic feature extraction. If the available local resources are greater than the resources required for semantic feature extraction, then it is determined that the local system has the capability to complete semantic feature extraction.

8. The method of claim 7, wherein, The data volume information includes at least one of data size and data type; the resource requirements include at least one of computing requirements and memory requirements.

9. The method of claim 7, wherein, The resources required for semantic feature extraction include the computational requirements and memory requirements for semantic feature extraction. If the available local resources are greater than the resources required for semantic feature extraction, then determining that the local system has the capability to complete semantic feature extraction includes: Obtain the locally available resources; the locally available resources include locally available computing resources and locally available memory space; If the available local computing resources are greater than the semantic feature extraction computing requirements, and the available local memory space is greater than the semantic feature extraction memory requirements, then it is determined that the local system has the capability to complete semantic feature extraction.

10. The method of claim 7, wherein, The performance metrics for completing semantic feature extraction locally include local completion time and local completion energy consumption; the performance metrics for completing semantic feature extraction on the cloud include cloud completion time, and data transmission time and data transmission energy consumption for transmitting the data to be processed to the server. The comparison of performance metrics for semantic feature extraction performed locally and performance metrics for semantic feature extraction performed offloaded to the cloud, and the determination of whether to perform semantic feature extraction locally or in the cloud based on the comparison results, includes: The data volume information is transmitted to the server; the data volume information is used to instruct the server to calculate and report back the cloud completion time, the data transmission time, and the data transmission energy consumption. If the local completion time is less than the sum of the data transmission time and the cloud completion time, and the local completion energy consumption is less than the data transmission energy consumption, then it is determined that semantic feature extraction will be performed locally. If the local completion time is greater than or equal to the sum of the data transmission time and the cloud completion time, and / or the local completion energy consumption is greater than or equal to the data transmission energy consumption, then semantic feature extraction is determined to be performed in the cloud.

11. The method of claim 1, wherein, The local semantic knowledge base includes one or more of the following: user information, device information, target domain knowledge, and contextual information.

12. The method according to claim 11, characterized in that, The user information includes one or more of user profiles, user preferences, user behavior patterns, and historical interaction records; the device information includes one or more of hardware configuration, software configuration, performance parameters, and device operating status; the target domain knowledge includes one or more of concept definitions, entity relationships, rule logic, and domain models; and the context information includes one or more of environmental state, situational information, and time information.

13. The method according to claim 1, characterized in that, The public semantic knowledge base includes one or more of the following: general knowledge, domain-specific knowledge, device interaction knowledge, and model library.

14. The method according to claim 13, characterized in that, The general knowledge includes one or more of common sense, conceptual definitions, and general rules; the domain expertise includes one or more of professional terminology, cases, regulations, and industry standards; the device interaction knowledge includes one or more of device types, interfaces, communication protocols, and operating instructions; and the model library includes one or more of statistical models, machine learning models, and deep learning models.

15. The method according to claim 3, characterized in that, The semantic features include one or more of entity recognition, sentiment analysis, topic classification, and semantic role labeling.

16. The method according to claim 3, characterized in that, The optimization suggestions include one or more of the following: maintain, expand, reduce, merge, split, refine, and simplify.

17. The method according to claim 4, characterized in that, The necessary data includes one or more of the semantic features, raw data, data type, data format, data creation time, and data version.

18. The method according to claim 4, characterized in that, The prompt words include one or more of the following: a description of the communication target, a summary of semantic features, and the sender's intent.

19. The method according to claim 18, characterized in that, The prompts can be presented in one or more of the following formats: text, vector, and structured data.

20. The method according to claim 5, characterized in that, The necessary data includes at least one of the following: original data, data type, data format, data creation time, and data version.

21. The method according to claim 5, characterized in that, The prompt words include one or more of the following: a description of the communication target, a description of the communication scenario, the sender's intent, and a personalized tag.

22. The method according to claim 5, characterized in that, The prompts can be presented in one or more of the following formats: text, vector, and structured data.

23. A semantic feature extraction method, characterized in that, Applied to a server, the method includes: Receive indication information from the sender; Based on the indicated information, semantic features are extracted using a public semantic knowledge base.

24. The method according to claim 23, characterized in that, The instruction information is sent by the sending end when it determines that semantic feature extraction will be performed in the cloud based on available local resources; the instruction information includes prompt words and necessary data.

25. The method according to claim 24, characterized in that, The step of extracting semantic features based on a public semantic knowledge base according to the indicated information includes: In response to receiving the prompt word and the necessary data, semantic features are extracted based on the public semantic knowledge base; The semantic features are transmitted to the sending end.

26. The method according to claim 23, characterized in that, The sending end determines whether to perform semantic feature extraction locally based on available local resources, and then extracts semantic features based on the local semantic knowledge base.

27. The method according to claim 26, characterized in that, The method further includes: An optimization suggestion is sent to the sending end; the optimization suggestion is used to instruct the sending end to optimize the semantic features extracted based on the local semantic knowledge base.

28. The method according to claim 27, characterized in that, Before sending the optimization suggestions to the sending end, the process includes: Receive prompts and necessary data from the sending end; Based on the prompt words and the necessary data, optimization suggestions for the semantic features are generated using the public semantic knowledge base.

29. The method according to claim 28, characterized in that, The prompt word is created by the sending end based on the semantic communication goal and the semantic features.

30. The method according to claim 26, characterized in that, The method further includes: The performance metrics for semantic feature extraction performed on the cloud are transmitted to the sending end. The performance metrics for semantic feature extraction performed on the cloud are used to instruct the sending end, when it is determined that the local end has the capability to perform semantic feature extraction based on the available local resources and the resources required for semantic feature extraction, to compare the performance metrics for semantic feature extraction performed locally with those for semantic feature extraction performed on the cloud, and to determine whether to perform semantic feature extraction locally or in the cloud based on the comparison result.

31. The method according to claim 30, characterized in that, The performance metrics for offloading to the cloud to complete semantic feature extraction include cloud completion time, data transmission time and energy consumption for transmitting the data to be processed to the server; the method further includes: Receive data volume information of the data to be processed from the sending end; Based on the data volume information, the cloud completion time, the data transmission time, and the data transmission energy consumption are calculated, and the cloud completion time, the data transmission time, and the data transmission energy consumption are transmitted to the sending end.

32. The method according to claim 31, characterized in that, The data volume information includes at least one of the data size and data type.

33. A semantic feature extraction device, characterized in that, Applied to the transmitting end, the device includes: The extraction method determination module is used to determine whether to perform semantic feature extraction locally or in the cloud, based on available local resources. The local extraction module is used to extract semantic features based on the local semantic knowledge base if it is determined that semantic feature extraction will be performed locally. The instruction module is used to send instruction information to the server if it is determined that semantic feature extraction will be performed in the cloud; the instruction information is used to instruct the server to extract semantic features based on a public semantic knowledge base.

34. A semantic feature extraction device, characterized in that, Applied to a server, the device includes: The information receiving module is used to receive indication information from the sending end; The feature extraction module is used to extract semantic features based on a public semantic knowledge base according to the indicated information.

35. A transmitter, characterized in that, include: Transmitter and processor; The processor is used to determine whether to perform semantic feature extraction locally or in the cloud, based on available local resources. If it is determined that semantic feature extraction will be performed locally, then semantic features will be extracted based on the local semantic knowledge base; And if it is determined that semantic feature extraction is performed in the cloud, the transmitter is controlled to send instruction information to the server; the instruction information is used to instruct the server to extract semantic features based on a public semantic knowledge base.

36. A server, characterized in that, include: Receiver and processor; The receiver is used to receive indication information from the sending end; The processor is configured to extract semantic features based on a public semantic knowledge base according to the instruction information.

37. A communication system, characterized in that, It includes the sending end as described in claim 35 and the server as described in claim 36.

38. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 32.

39. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 32.

Images