WO2026000751A1 - 语义通信方法、装置、终端、服务器、系统和存储介质 - Google Patents
语义通信方法、装置、终端、服务器、系统和存储介质Info
- Publication number
- WO2026000751A1 WO2026000751A1 PCT/CN2024/129083 CN2024129083W WO2026000751A1 WO 2026000751 A1 WO2026000751 A1 WO 2026000751A1 CN 2024129083 W CN2024129083 W CN 2024129083W WO 2026000751 A1 WO2026000751 A1 WO 2026000751A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- semantic
- information
- terminal
- communication
- server
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N5/00—Computing arrangements using knowledge-based models
- G06N5/02—Knowledge representation; Symbolic representation
- G06N5/022—Knowledge engineering; Knowledge acquisition
- G06N5/025—Extracting rules from data
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L1/00—Arrangements for detecting or preventing errors in the information received
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L67/00—Network arrangements or protocols for supporting network services or applications
- H04L67/01—Protocols
- H04L67/12—Protocols specially adapted for proprietary or special-purpose networking environments, e.g. medical networks, sensor networks, networks in vehicles or remote metering networks
Definitions
- This application relates to the field of wireless communication technology, and in particular to a semantic communication method, apparatus, terminal, server, system, and storage medium.
- This application provides a semantic communication method, apparatus, terminal, server, system, and storage medium, which can improve the communication reliability of semantic communication.
- This application provides a semantic communication method in a first aspect.
- the method includes: performing semantic processing on semantic communication information during the semantic communication process; if the semantic processing fails, sending the semantic communication information to a server so that the server can process the semantic communication information and obtain the semantic information corresponding to the semantic communication information, and returning the semantic information.
- the semantic processing includes semantic extraction and/or semantic restoration.
- the semantic processing of semantic communication information during semantic communication includes: performing semantic processing on the semantic communication information using a local semantic knowledge base model.
- the local semantic knowledge base model is the semantic knowledge base foundation model constructed by the first terminal through federated learning.
- the method further includes receiving the semantic communication information sent by a second terminal.
- the method further includes: receiving the semantic information sent by the server; wherein the semantic information is obtained by the server through semantic processing of the semantic communication information using a global model, or the semantic information is obtained by the server through semantic processing of approximate features of the semantic communication information using the global model.
- the method further includes obtaining the semantic information if the semantic processing is successful.
- the method further includes: determining whether the current semantic communication process is completed based on the semantic information; if the current semantic communication process is determined to be completed, sending semantic communication completion information to the second terminal.
- the method after sending semantic communication completion information to the second terminal, the method further includes: performing an operation corresponding to the semantic information.
- determining whether the current semantic communication process is completed based on the semantic information includes: determining whether the semantic information satisfies the communication intent or communication goal corresponding to the semantic communication information; if the semantic information satisfies the communication intent or the communication goal, then the current semantic communication process is determined to be completed.
- the method further includes: if it is determined that the current semantic communication process has not been completed, sending feedback information to the second terminal, the feedback information being used to instruct the second terminal to re-execute the semantic communication process.
- the method further includes: receiving semantic communication information retransmitted by the second terminal, and performing semantic processing on the semantic communication information retransmitted by the second terminal to obtain the semantic information.
- the method further includes: receiving updated model parameters sent by the server, wherein the updated model parameters are obtained by updating the global model based on the approximate features after the server performs semantic processing on the semantic information obtained by the server on the approximate features, and upon receiving semantic communication completion information sent by the second terminal; and updating the local semantic knowledge base model based on the updated model parameters.
- the method further includes: receiving semantic processing failure feedback sent by the server; and sending the semantic processing failure feedback to a second terminal.
- This application provides a semantic communication method in a second aspect.
- the method includes: receiving semantic communication information sent by a first terminal, and performing semantic processing on the semantic communication information; and sending the semantic information obtained from the semantic processing to the first terminal.
- the semantic processing of the semantic communication information includes: performing semantic processing on the semantic communication information according to a global model; if no semantic information is obtained, then performing information decomposition on the semantic communication information to obtain approximate features, and performing semantic processing on the approximate features through the global model.
- the method further includes: receiving semantic communication completion information sent by a second terminal, wherein the semantic communication completion information is determined by the first terminal based on the semantic information, and the semantic information is obtained by the server through semantic processing of the approximate features using the global model; updating the model parameters of the global model based on the approximate features to obtain updated model parameters; and sending the updated model parameters to the first terminal.
- the method further includes: if semantic information is not obtained during the semantic processing of the approximate features through the global model, then sending a semantic processing failure feedback to the first terminal.
- the method further includes: if multiple failure notifications are received from the second terminal, then deleting the approximate features obtained from the disassembly.
- the device includes: a semantic processing module for performing semantic processing on semantic communication information during the semantic communication process; and a sending module for sending the semantic communication information to a server if the semantic processing fails, so that the server can process the semantic communication information, obtain the semantic information corresponding to the semantic communication information, and return the semantic information.
- the device includes: a receiving module for receiving semantic communication information sent by a first terminal and performing semantic processing on the semantic communication information; and a sending module for sending the semantically processed semantic information to the first terminal.
- the terminal includes a transmitter and a processor; the processor is configured to perform semantic processing on semantic communication information during semantic communication; the transmitter is configured to send the semantic communication information to a server if semantic processing fails, so that the server can process the semantic communication information, obtain the semantic information corresponding to the semantic communication information, and return the semantic information.
- the server includes a transmitter, a processor, and a receiver: the receiver is configured to receive semantic communication information sent by a first terminal; the processor is configured to perform semantic processing on the semantic communication information; and the transmitter is configured to send the semantically processed semantic information to the first terminal.
- the semantic communication system includes a first terminal, a second terminal, and a server; the second terminal is configured to send semantic communication information to the first terminal; the first terminal is configured to perform the steps of the method described in the first aspect above; and the server is configured to perform the steps of the method described in the second aspect above.
- the second terminal is further configured to perform a knowledge base search based on the communication information to be transmitted to obtain the semantic communication information; the second terminal is further configured to, if the knowledge base search fails, send the communication information to be transmitted... The information is sent to the server, which then performs a knowledge base search based on the communication information to be transmitted to obtain the semantic communication information.
- This application also provides a computer-readable storage medium in an eighth aspect.
- the computer-readable storage medium stores a computer program that, when executed by a processor, performs the steps of the method described in the first or second aspect above.
- the chip includes programmable logic circuitry and/or program instructions that, when the chip is executed, implement the steps of the methods described in the first or second aspect above.
- the computer program product includes a computer program that, when executed by a processor, performs the steps of the method described in the first or second aspect above.
- the aforementioned semantic communication method, apparatus, terminal, server, system, and storage medium involve a first terminal performing semantic processing on semantic communication information during the semantic communication process. If semantic processing fails, the first terminal sends the semantic communication information to the server, which processes the semantic communication information, obtains the corresponding semantic information, and returns the semantic information.
- this embodiment allows the first terminal to utilize the server's capabilities when its capabilities are insufficient (i.e., when the first terminal fails to successfully process the semantic communication information).
- the server then returns the obtained semantic information to the first terminal, thereby improving the success rate of semantic processing, ensuring the smooth progress of the semantic communication process, and enhancing the reliability of semantic communication.
- Figure 1 is a diagram of the implementation environment of the semantic communication method in one embodiment of this application.
- Figure 2 is a flowchart illustrating a semantic communication method in one embodiment of this application.
- Figure 3 is a flowchart illustrating the semantic communication method in another embodiment of this application.
- Figure 4 is a flowchart illustrating the semantic communication method in another embodiment of this application.
- Figure 5 is a flowchart illustrating the process by which the first terminal determines whether the semantic communication process is complete in another embodiment of this application.
- Figure 6 is a flowchart illustrating the semantic communication method in another embodiment of this application.
- Figure 7 is a flowchart illustrating the semantic communication method in another embodiment of this application.
- Figure 8 is a flowchart illustrating the semantic communication method in another embodiment of this application.
- Figure 9 is a flowchart illustrating the semantic communication method in another embodiment of this application.
- Figure 10 is a flowchart illustrating a semantic communication method in another embodiment of this application.
- Figure 11 is a structural block diagram of a semantic communication device in one embodiment of this application.
- Figure 12 is a structural block diagram of a semantic communication device in another embodiment of this application.
- Figure 13 is a schematic diagram of the terminal structure in one embodiment of this application.
- Figure 14 is a schematic diagram of the server structure in one embodiment of this application.
- Figure 15 is a schematic structural diagram of a chip in one embodiment of this application.
- Semantic communication integrates user needs and information meaning into the communication process. Its fundamental purpose is information exchange; it doesn't pursue the fidelity of the original data or signals, but rather the accurate transmission of semantics between the communicating parties. Therefore, in semantic communication, both the sending and receiving ends possess background knowledge bases. The communicating parties perform semantic parsing based on these background knowledge bases, thereby enabling semantic transmission between them.
- point-to-point semantic communication can be divided into three categories: knowledge graph-based semantic communication, semantic base-based semantic communication, and model-based semantic communication.
- semantic communication methods suffer from poor communication reliability. For example, if the sending and receiving ends fail to perform semantic parsing based on their respective background knowledge bases (e.g., the receiving end fails to perform semantic reconstruction based on its local background knowledge base), semantic communication will fail, resulting in poor communication reliability.
- embodiments of this application provide a semantic communication method, apparatus, terminal, server, system, and storage medium, which can improve the communication reliability of semantic communication.
- Figure 1 is a schematic diagram of the application environment of a semantic communication method provided in an embodiment of this application.
- the first terminal 102 communicates with the second terminal 104 via a network
- the first terminal 102 communicates with the server 106 via a network
- the second terminal 104 communicates with the server 108 via a network.
- the first terminal 102 and the second terminal 104 can be, but are not limited to, various personal computers, laptops, smartphones, tablets, IoT devices, and portable wearable devices.
- IoT devices can include smart speakers, smart TVs, smart air conditioners, smart in-vehicle systems, projection devices, etc.
- Portable wearable devices can include smartwatches, smart bracelets, head-mounted displays, etc. Head-mounted displays can include virtual reality (VR) devices, augmented reality (AR) devices, smart glasses, etc.
- VR virtual reality
- AR augmented reality
- Server 106 and Server 108 can be the same server or different servers. Taking Server 106 as an example, Server 106 can be a standalone physical server, a server cluster or distributed system composed of multiple physical servers, an edge server, a cloud server providing cloud computing services, etc. Server 108 is similar in form to Server 106, and will not be described in detail here.
- a semantic communication method is provided. Taking the application of this method to the first terminal in FIG1 as an example, the method includes the following steps S201 to S202.
- Step S201 The first terminal performs semantic processing on the semantic communication information in the semantic communication process.
- the first terminal and the second terminal communicate semantically.
- the first terminal and the second terminal can be heterogeneous terminals, for example, the computing power, hardware capabilities, and power consumption of the first terminal and the second terminal may differ; alternatively, the relevant parameters of the first terminal and the second terminal may also be the same, without specific restrictions.
- the number of first terminals can be one or more, and the number of second terminals can also be one or more.
- the following embodiments will be described using the example that the number of first terminals and the number of second terminals are both one.
- the first terminal can be either a sender or a receiver.
- the sender is the party transmitting the information, used to modulate and transmit it;
- the receiver is the party receiving the information, used to demodulate and receive it.
- the sender extracts semantic features from the communication information to be transmitted, and sends these features to the receiver.
- the receiver Upon receiving the semantic features, the receiver performs semantic reconstruction to restore the original communication information.
- semantic processing includes semantic extraction and/or semantic reconstruction.
- the semantic communication information can be the communication information to be transmitted acquired by the first terminal.
- the semantic processing performed by the first terminal on the semantic communication information can refer to the first terminal performing semantic extraction on the semantic communication information to obtain the semantic information corresponding to the semantic communication information.
- the semantic information can be the semantic features extracted from the semantic communication information.
- the first terminal When the first terminal acts as the receiving end, it receives semantic communication information sent by the second terminal.
- This semantic communication information can be semantic features sent by the second terminal, and the first terminal performs semantic processing on the semantic communication information. This can refer to the first terminal performing semantic restoration on the semantic communication information to obtain the semantic information corresponding to the semantic communication information.
- This semantic information can be the complete communication information obtained by restoring the semantic features.
- the first terminal can utilize a knowledge graph to perform semantic processing on the semantic communication information in an attempt to obtain the semantic information corresponding to the semantic communication information.
- the first terminal can also utilize a semantic base to perform semantic processing on the semantic communication information in an attempt to obtain the semantic information corresponding to the semantic communication information.
- Other methods by which the first terminal performs semantic processing on the semantic communication information will be exemplarily described in the following embodiments.
- Step S202 If semantic processing fails, the first terminal sends the semantic communication information to the server.
- the semantic communication information is used by the server to process and obtain the semantic information corresponding to the semantic communication information, and then return the semantic information.
- the first terminal performs semantic processing on the semantic communication information in accordance with the above step S201. If the semantic processing is successful, the first terminal obtains the semantic information corresponding to the semantic communication information.
- the first terminal will send the semantic communication information to the server.
- the server After receiving the semantic communication information sent by the first terminal, the server performs semantic processing on the semantic communication information. Then, the server sends the semantic information obtained from the semantic processing back to the first terminal. The process of the server performing semantic processing on the semantic communication information will be described in the following embodiments.
- the first terminal when the processing capability of the first terminal is insufficient, that is, when the first terminal fails to successfully perform semantic processing on the semantic communication information, the first terminal can borrow the capability of the server to perform semantic processing on the semantic communication information, and the server can then return the obtained semantic information to the first terminal. This can improve the success rate of semantic processing and also improve the communication reliability of semantic communication.
- the second terminal can also utilize the server's processing power.
- the second terminal first searches the knowledge base based on the communication information to be transmitted, attempting to obtain the semantic communication information. If the knowledge base search fails, the second terminal sends the communication information to be transmitted to the server. The server then searches the knowledge base based on this communication information to obtain the semantic communication information, thus ensuring the smooth progress of the semantic communication process.
- Step S2011 The first terminal uses the local semantic knowledge base model to perform semantic processing on the semantic communication information.
- the first terminal can input semantic communication information into the local semantic knowledge base model stored locally on the first terminal, and perform semantic processing on the semantic communication information through the local semantic knowledge base model.
- the local semantic knowledge base model can perform semantic extraction on the semantic communication information to obtain the semantic information corresponding to the semantic communication information.
- This semantic information can be the semantic features extracted from the semantic communication information.
- the local semantic knowledge base model can perform semantic restoration on the semantic communication information to obtain the semantic information corresponding to the semantic communication information.
- This semantic information can be the complete communication information restored based on the semantic features.
- Federated Machine Learning enables multi-party joint modeling by exchanging locally trained model parameters.
- the server can collaborate with a first terminal and/or a second terminal to perform federated learning. After learning is complete, the server obtains a global model, and can then distribute the model parameters of the global model to the first and second terminals.
- the first terminal stores the global model as a local semantic knowledge base model.
- the first terminal can also optimize the global model using local data to obtain a local semantic knowledge base model.
- the first terminal After the first terminal obtains the semantic communication information in the semantic communication process, it can perform semantic processing on the semantic communication information through the local semantic knowledge base model, that is, use the local semantic knowledge base model to perform semantic extraction or semantic restoration to obtain semantic information.
- the first terminal will send semantic communication information to the server.
- the first terminal inputs semantic communication information into the local semantic knowledge base model but does not obtain the semantic information corresponding to the semantic communication information output by the local semantic knowledge base model, it determines that the semantic processing has failed, and the first terminal sends the semantic communication information to the server. After receiving the semantic communication information sent by the first terminal, the server performs semantic processing on the semantic communication information, and then sends the semantic information obtained from the semantic processing to the first terminal.
- the first terminal obtains the semantic information corresponding to the semantic communication information.
- the above embodiments construct a local semantic knowledge base model through federated learning.
- the parties participating in the joint modeling optimize the model by exchanging model parameters, resulting in better model performance.
- Federated learning can meet the requirement of keeping the original data locally and is applicable to more complex scenarios, while avoiding security risks such as data/model privacy leaks during the joint modeling process.
- the semantic communication method further includes step S401 shown in FIG4.
- Step S401 The first terminal receives semantic information sent by the server.
- the server If the server successfully processes the semantic communication information sent by the first terminal and obtains semantic information, the server will send the semantic information to the first terminal, so that the first terminal can receive the semantic information sent by the server.
- the semantic information is obtained by the server through semantic processing of the semantic communication information based on the global model, or the semantic information is obtained by the server through semantic processing of the approximate features of the semantic communication information based on the global model.
- the global model is the basic semantic knowledge base model constructed by the server through federated learning.
- the server can perform federated learning in conjunction with the first terminal and/or the second terminal. After the learning is completed, the server obtains the global model, and the server can distribute the model parameters of the global model to the first terminal and the second terminal.
- the first terminal stores the global model as a local semantic knowledge base model; optionally, the first terminal can also use local data to optimize all models, and obtain a local semantic knowledge base model after optimization.
- the server after the server receives the semantic communication information sent by the first terminal, the server first performs semantic processing on the semantic communication information according to the global model stored locally. For example, the server can input the semantic communication information into the global model and attempt to perform semantic processing on the semantic communication information through the global model to obtain semantic information.
- the global model stored locally in the server can perform semantic extraction to obtain semantic information, which can be the semantic features corresponding to the semantic communication information.
- the global model stored locally in the server can perform semantic reconstruction to obtain semantic information, which can be the complete communication information reconstructed based on the semantic features.
- the following embodiments use the first terminal as the receiver as an example.
- the server If the server performs semantic processing on the semantic communication information according to the global model and successfully obtains the semantic information, the server will return the obtained semantic information to the first terminal.
- the server will decompose the semantic communication information to obtain approximate features, and then perform semantic processing on the approximate features based on the global model.
- the server can input semantic communication information into an approximate feature decomposition model to obtain approximate features of the semantic communication information; then, the server inputs the approximate features into a global model, performs semantic processing on the approximate features through the global model, and attempts to obtain semantic information; if the semantic information is successfully obtained, the server returns the obtained semantic information to the first terminal.
- the server fails to obtain semantic information during semantic processing of approximate features based on the global model, it indicates that the server's semantic processing has failed, and the server sends a semantic processing failure feedback to the first terminal. In some embodiments, if the server fails to obtain approximate features after decomposing the semantic communication information, it also indicates that the server's semantic processing has failed, and the server sends a semantic processing failure feedback to the first terminal.
- the first terminal After receiving a semantic processing failure feedback from the server, the first terminal sends the feedback to the second terminal.
- the second terminal sends the feedback to the server, which then determines to re-process the semantics (e.g., re-decompose the information to obtain approximate features, and apply the re-decomposed information to the global model). (Semantic processing is performed on approximate features) to obtain semantic information. The process of re-performing semantic processing will be described in the following embodiments. If the server receives multiple failure notifications from the second terminal, the server deletes the approximate features obtained from the decomposition and terminates the current semantic communication process.
- the first terminal performs semantic processing on the semantic communication information
- the first terminal obtains the semantic information corresponding to the semantic communication information
- the first terminal sends the semantic communication information to the server.
- the server performs semantic processing successfully, the server returns the semantic information corresponding to the semantic communication information to the first terminal.
- the semantic communication process cannot be considered complete.
- the semantic communication method of this application embodiment further includes steps S501-S504 shown in FIG5.
- Step S501 The first terminal determines whether the current semantic communication process is complete based on the semantic information.
- the first terminal can either successfully process the semantic communication information on its own to obtain semantic information, or receive semantic information returned by the server after successfully processing the semantic communication information. Based on this semantic information, the first terminal can determine whether the current semantic communication process is complete.
- the second terminal when the second terminal sends semantic communication information to the first terminal, it may also send the communication target and/or communication intent corresponding to the current semantic communication process.
- the second terminal as the sender, senses the communication environment, obtains the communication information to be transmitted, and then searches the knowledge base based on the communication information to be transmitted to extract the semantic communication information to be transmitted (in some embodiments, this may also be the extraction of the communication target and/or communication intent) and sends it to the first terminal.
- the first terminal performs semantic processing based on the received semantic communication information (in some embodiments, this may also be combined with the sensed communication environment and the communication target and/or communication intent sent by the second terminal), or the first terminal performs semantic processing through a server to obtain semantic information.
- the first terminal determines whether the current semantic communication process is complete based on semantic information
- the first terminal can determine whether the semantic information satisfies the communication intent or communication goal corresponding to the semantic communication information. For example, the first terminal extracts the communication intent or communication goal from the semantic communication information, and then compares the extracted communication intent or communication goal with the communication goal or communication intent sent by the second terminal to determine whether the semantic information satisfies the communication intent or communication goal sent by the second terminal. If the semantic information satisfies the communication intent or communication goal, the first terminal determines that the current semantic communication process is complete.
- the first terminal determines that the semantic information can form a complete semantic meaning that conforms to the communication intent or communication goal, and thus determines that the current semantic communication process is complete. Following this method, it is possible to ensure that both communicating parties reach a consensus on the judgment standard for whether communication is complete, thus ensuring normal semantic communication.
- the first terminal may also input the semantic communication information and its corresponding semantic information into a pre-trained classification model, and the classification model shall determine whether the semantic information conforms to the communication intent or communication goal corresponding to the semantic communication information.
- Step S502 If it is determined that the current semantic communication process is completed, the first terminal sends semantic communication completion information to the second terminal.
- Step S503 The first terminal performs the operation corresponding to the semantic information.
- the first terminal sends semantic communication completion information to the second terminal.
- the semantic information instructs the first terminal to perform a certain operation
- the first terminal performs the operation corresponding to the semantic information.
- Step S504 If it is determined that the current semantic communication process is not completed, the first terminal sends feedback information to the second terminal.
- the feedback information is used to instruct the second terminal to re-execute the semantic communication process.
- the second terminal judges the context information and re-obtains the semantic communication information. For example, the second terminal re-extracts the semantics of the communication information to be transmitted, thereby extracting semantic features from the communication information to be transmitted as semantic communication information that needs to be re-sent to the first terminal. The second terminal then sends the semantic communication information to the first terminal.
- the first terminal After receiving the semantic communication information resent by the second terminal, the first terminal performs semantic processing on the resent semantic communication information to obtain semantic information.
- the first terminal first attempts to perform semantic processing on the semantic communication information during the semantic communication process. If the semantic processing is successful, the first terminal obtains the semantic information. If the semantic processing fails, the first terminal resends the semantic communication information to the server, which then performs semantic processing on the semantic communication information again.
- the server first performs semantic processing on the semantic communication information based on the locally stored global model. If the server successfully obtains the semantic information, it returns the obtained semantic information to the first terminal. If the server does not obtain the semantic information after performing semantic processing on the semantic communication information based on the global model, the server decomposes the semantic communication information to obtain approximate features and performs semantic processing on the approximate features based on the global model. If the server successfully obtains the semantic information, it returns the obtained semantic information to the first terminal. If the semantic processing fails, the server sends a semantic processing failure feedback to the first terminal.
- the first terminal and the server when the first terminal and the server perform semantic processing, they can also combine context information to search the local semantic knowledge base basic model and parse the semantic communication information to obtain semantic information.
- the server decomposes the semantic communication information sent by the first terminal to obtain approximate features.
- the server performs semantic processing on the approximate features through a global model and successfully obtains semantic information
- the server returns the obtained semantic information to the first terminal.
- the first terminal determines that the current semantic communication process is complete based on the semantic information
- the first terminal sends semantic communication completion information to the second terminal.
- the server After receiving the semantic communication completion information sent by the second terminal, the server updates the model parameters of the global model based on the approximate features to obtain the updated model parameters. That is, the server uses the approximate features as training samples to train the global model to optimize it, and obtains the latest model parameters of the global model after optimization. After updating the model parameters, the server sends the updated model parameters to the first terminal so that the first terminal can update its local semantic knowledge base model. In some embodiments, the server can also perform corresponding operations according to the instructions of the semantic communication completion information.
- the first terminal After receiving the updated model parameters from the server, the first terminal updates its local semantic knowledge base model based on the updated model parameters.
- the server and terminal in the above embodiments can update the semantic knowledge base basic model according to their own processing capabilities and communication data, maximize the coverage of various approximate features, thereby improving the accuracy of communication and ensuring communication quality.
- this application provides a semantic communication method for a server, the method comprising the following steps S601 to S602.
- Step S601 The server receives the semantic communication information sent by the first terminal and performs semantic processing on the semantic communication information.
- Step S602 The server sends the semantic information obtained through semantic processing to the first terminal.
- step S601 includes steps S701, S702 and S703 shown in FIG7.
- Step S701 The server receives semantic communication information sent by the first terminal.
- Step S702 The server performs semantic processing on the semantic communication information based on the global model.
- Step S703 If semantic information is not obtained, the server will decompose the semantic communication information to obtain approximate features, and perform semantic processing on the approximate features through a global model.
- the method further includes:
- the server receives semantic communication completion information sent by the second terminal.
- the semantic communication completion information is determined by the first terminal based on semantic information, which is obtained by the server through semantic processing of approximate features using a global model.
- the server updates the model parameters of the global model based on the approximate features, and obtains the updated model parameters.
- the server sends the updated model parameters to the first terminal.
- the method further includes:
- the server sends a semantic processing failure feedback to the first terminal.
- the method further includes:
- the server receives multiple failure notifications from the second terminal, it will delete the approximate features obtained from the disassembly.
- semantic communication method for the server please refer to the relevant description in the above description of the semantic communication method for the first terminal. Furthermore, the beneficial effects of the semantic communication method for the server are the same as those described above. The beneficial effects of the semantic communication method used for the first terminal are similar and will not be repeated here.
- a semantic communication method is provided, as shown in FIG8, including the following steps S801 to S808.
- Step S801 The first terminal receives semantic communication information sent by the second terminal.
- Step S802 The first terminal uses the local semantic knowledge base model to perform semantic processing on the semantic communication information. If the semantic processing fails, the first terminal sends the semantic communication information to the server.
- Step S803 The server performs semantic processing on the semantic communication information through the global model. If no semantic information is obtained, the server decomposes the semantic communication information to obtain approximate features and performs semantic processing on the approximate features through the global model.
- Step S804 The server sends the semantic information obtained from semantic processing to the first terminal.
- Step S805 The first terminal determines whether the current semantic communication process is complete based on the semantic information. If the current semantic communication process is complete, the first terminal sends semantic communication completion information to the second terminal.
- Step S806 The second terminal sends semantic communication completion information to the server.
- Step S807 After receiving the semantic communication completion information sent by the second terminal, the server updates the model parameters of the global model based on the approximate features to obtain the updated model parameters.
- Step S808 The server sends the updated model parameters to the first terminal and/or the second terminal.
- a semantic communication method is provided, as shown in FIG9, including the following steps S901 to S912.
- Step S901 The first terminal receives semantic communication information sent by the second terminal.
- Step S902 The first terminal uses the local semantic knowledge base model to perform semantic processing on the semantic communication information. If the semantic processing fails, the first terminal sends the semantic communication information to the server.
- Step S903 The server performs semantic processing on the semantic communication information through the global model. If no semantic information is obtained, the server decomposes the semantic communication information to obtain approximate features and performs semantic processing on the approximate features through the global model.
- Step S904 The server sends the semantic information obtained from semantic processing to the first terminal.
- Step S905 The first terminal determines whether the current semantic communication process is complete based on the semantic information. If the current semantic communication process is not complete, the first terminal sends feedback information to the second terminal.
- Step S906 The second terminal resends semantic communication information to the first terminal.
- Step S907 The second terminal sends a semantic processing failure feedback to the server.
- Step S908 The first terminal performs semantic processing on the semantic communication information resent by the second terminal. If the semantic processing fails, the first terminal resends the semantic communication information to the server.
- Step S909 The server reprocesses the semantic communication information through the global model. If no semantic information is obtained, the server decomposes the semantic communication information to obtain approximate features and processes the approximate features through the global model. The server then sends the semantic information obtained from the semantic processing to the first terminal.
- Step S910 The first terminal determines whether the current semantic communication process is completed based on the semantic information. If the current semantic communication process is not completed, the first terminal sends feedback information to the second terminal.
- Step S911 If the number of times the semantic communication process is not completed reaches the threshold, the second terminal sends multiple failure notifications to the server.
- Step S912 After receiving multiple failure notifications, the server deletes the approximate features obtained from the disassembly and retains the model parameters of the global model.
- a semantic communication method is provided, as shown in FIG10, including the following steps S1001 to S1004.
- Step S1001 The first terminal receives semantic communication information sent by the second terminal.
- Step S1002 The first terminal uses the local semantic knowledge base model to perform semantic processing on the semantic communication information. If the semantic processing is successful, the first terminal obtains the semantic information.
- Step S1003 The first terminal determines whether the current semantic communication process is complete based on the semantic information.
- Step S1004 If the current semantic communication process is completed, the first terminal sends semantic communication completion information to the second terminal.
- Example 1 There are multiple smart home terminals: terminal 1, terminal 2, terminal 3 and terminal 4 (in the case where the first terminal is the receiver and the second terminal is the transmitter, terminal 1 is equivalent to the second terminal, and terminals 2, 3 and 4 are equivalent to the first terminal).
- terminal 1 is a mobile phone
- terminal 2 is a computer
- terminal 3 is a robot vacuum cleaner
- terminal 4 is an air conditioner. All smart home terminals can be connected to the indoor home computing center (equivalent to the server mentioned above).
- Terminal 1 senses the communication environment and obtains communication scenario information: Terminal 1 is outdoors, and the current time is the time to prepare to leave work. The environment where Terminal 1 is located has little noise interference and a lot of available transmission bandwidth resources. Terminal 1 needs to set the indoor terminal to the welcome home mode.
- Terminal 1 Based on the perceived communication scenario information, Terminal 1 determines the communication target as: the indoor terminal needs to be set to the welcome home mode. Then, Terminal 1 searches the knowledge base based on the communication scenario information and extracts the semantic communication information to be transmitted (optionally, the communication intent can also be extracted). Terminal 1 sends the semantic communication information (optionally carrying the communication target and/or communication intent) to Terminal 2, Terminal 3 and Terminal 4 respectively.
- Terminal 2 senses the communication environment: Terminal 2 is in a normal position, it is now 5:00, and it is in standby mode;
- Terminal 3 senses the communication environment: Terminal 3 is in the bedroom, it is now 5:00, and it is in sweeping and mopping mode;
- Terminal 4 senses the communication environment: Terminal 4 is in a normal position, it is now 5:00, the indoor temperature is 30°C, and the humidity is 90%.
- Terminal 2 Terminal 3 and Terminal 4 receive semantic communication information from Terminal 1.
- Terminal 2 Terminal 3 and Terminal 4 perform semantic processing (semantic restoration) on the semantic communication information respectively, and attempt to restore the complete semantic information corresponding to the semantic communication information.
- terminal 2 After semantic processing, terminal 2 obtains semantic information: set to welcome mode, terminal 2 judges that the semantic information is semantically complete and conforms to the communication intent, and terminal 2 determines that the current semantic communication process is complete.
- Terminal 2 switches from standby mode to power-on mode and sends semantic communication completion information to terminal 1, ending the communication.
- terminals 3 and 4 When semantic processing of terminals 3 and 4 fails, terminals 3 and 4 will transmit the received semantic communication information to the indoor home computing center.
- the indoor home computing center performs semantic processing (semantic restoration) on semantic communication information through a global model.
- the semantic processing of the indoor home computing center is successful, and the semantic information obtained is: set to welcome home mode.
- the indoor home computing center will send the semantic information to terminal 3 and terminal 4.
- Terminal 3 determines that the semantic information is semantically complete and conforms to the communication intent, and Terminal 3 determines that the current semantic communication process is complete.
- Terminal 3 switches from working mode to standby mode and sends semantic communication completion information to terminal 1, ending the communication.
- Terminal 4 determines that the semantic information is semantically complete and conforms to the communication intent, and Terminal 4 determines that the current semantic communication process is complete.
- Terminal 4 turns on the air conditioner and sets the temperature to 23°C, then sends a semantic communication completion message to Terminal 1 to end the communication.
- Example 2 There are multiple metaverse terminals: Terminal 1, Terminal 2, Terminal 3, and Terminal 4 (where the first terminal acts as the receiver and the second terminal acts as the transmitter, Terminal 1 is equivalent to the second terminal, and Terminals 2, 3, and 4 are equivalent to the first terminal).
- Terminal 1 is a game console
- Terminal 2 is Extended Reality (ER) glasses connected to computing server 1
- Terminal 3 is Augmented Reality (AR) glasses connected to computing server 1
- Terminal 4 is a regular translation glasses connected to computing server 2.
- Terminal 1 senses the communication environment and obtains communication scenario information: Terminal 1 has abundant available transmission bandwidth resources.
- terminal 1 Based on the perceived communication scenario information, terminal 1 determines that the communication target is: the game content needs to be updated. Then, terminal 1 searches the knowledge base based on the communication scenario information and extracts the semantic communication information to be transmitted (optionally, the communication intent can also be extracted). Terminal 1 sends the semantic communication information (optionally carrying the communication target and/or communication intent) to terminal 2, terminal 3 and terminal 4 respectively.
- Terminal 2 senses the communication environment: Terminal 2 is in a normal position, connected to the Wireless Fidelity (WIFI) network, and is being worn;
- Terminal 3 senses the communication environment: Terminal 3 is in the bedroom, connected to WIFI, and is in conference mode;
- Terminal 4 senses the communication environment: Terminal 4 is in a normal position and is in power-on data mode.
- WIFI Wireless Fidelity
- Terminal 2 Terminal 3 and Terminal 4 receive semantic communication information from Terminal 1.
- Terminal 2 Terminal 3 and Terminal 4 perform semantic processing (semantic restoration) on the semantic communication information respectively, and attempt to restore the complete semantic information corresponding to the semantic communication information.
- terminal 2 After semantic processing, terminal 2 obtains semantic information: skin update. Terminal 2 determines that the semantic information is semantically complete and conforms to the communication intent. Terminal 2 then determines that the current semantic communication process is complete.
- Terminal 2 updates the game skin and sends semantic communication completion information to Terminal 1, ending the communication.
- terminal 3 transmits the received semantic communication information to computing server 1, and terminal 4 transmits the received semantic communication information to computing server 2.
- the computing server 1 performs semantic processing (semantic restoration) on the semantic communication information through the global model, and obtains the semantic information: update skin status. The computing server 1 then sends the semantic information to the terminal 3.
- semantic processing semantic restoration
- the terminal 3 determines that the semantic information is semantically complete and conforms to the communication intent, the terminal 3 determines that the current semantic communication process is complete and performs the game skin update; if the semantic information does not conform to the communication intent or communication goal, the game skin update will not be performed.
- Terminal 3 sends semantic communication completion information to Terminal 1, ending the communication.
- Computing server 2 failed to perform semantic processing (semantic restoration) on semantic communication information through the global model.
- the computing server 2 decomposes the semantic communication information to obtain approximate features: lens color (at this time, the approximate features obtained by the computing server 2 are inaccurate).
- the computing server 2 inputs the approximate features into the global model and returns the extraction results (semantic information) to the terminal 4.
- terminal 4 determines that the current semantic communication process is not completed, it will send feedback information to terminal 1 and restart the communication.
- terminal 1 Based on the feedback information from terminal 4, terminal 1 judges the context information, searches the knowledge base again, and obtains semantic communication information again: updates the game character skin and sends it to terminal 4.
- Terminal 4 judges the context information again based on the received semantic communication information and performs semantic processing (semantic restoration) again. If the semantic information cannot be restored, terminal 4 sends the semantic communication information to computing server 2.
- the computing server 2 decomposes the semantic communication information but fails to extract approximate features.
- the computing server 2 sends the semantic processing failure feedback to the terminal 4, deletes the relevant data, and retains the original model parameters without updating.
- Terminal 4 feedback to Terminal 1 Semantic processing failure feedback.
- steps in the flowchart above are shown sequentially as indicated by the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowchart above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages in other steps.
- a semantic communication device disposed in a first terminal, comprising:
- the semantic processing module 1101 is used to perform semantic processing on the semantic communication information during the semantic communication process
- the sending module 1102 is used to send the semantic communication information to the server if the semantic processing fails, so that the server can process the semantic communication information, obtain the semantic information corresponding to the semantic communication information, and return the semantic information.
- semantic processing includes semantic extraction and/or semantic restoration.
- the semantic processing module 1101 is specifically used to perform semantic processing on semantic communication information using a local semantic knowledge base model.
- the local semantic knowledge base model is the semantic knowledge base constructed by the first terminal through federated learning. Basic model.
- the above-mentioned apparatus further includes:
- the receiving module is used to receive semantic communication information sent by the second terminal.
- the receiving module is further configured to receive semantic information sent by the server.
- the semantic information is obtained by the server through semantic processing of the semantic communication information using a global model, or the semantic information is obtained by the server through semantic processing of the approximate features of the semantic communication information using the global model.
- the above-mentioned apparatus further includes:
- the acquisition module is used to obtain semantic information if semantic processing is successful.
- the above-mentioned apparatus further includes:
- the determination module is used to determine whether the current semantic communication process is complete based on semantic information.
- the sending module 1102 is also used to send semantic communication completion information to the second terminal if the current semantic communication process is completed.
- the above-mentioned apparatus further includes:
- the execution module is used to perform operations corresponding to semantic information.
- the determining module is specifically used to determine whether the semantic information satisfies the communication intent or communication goal corresponding to the semantic communication information; if the semantic information satisfies the communication intent or communication goal, then the current semantic communication process is determined to be complete.
- the sending module 1102 is further configured to send feedback information to the second terminal if the current semantic communication process is not completed, the feedback information being used to instruct the second terminal to re-execute the semantic communication process.
- the receiving module is further configured to receive semantic communication information retransmitted by the second terminal, and perform semantic processing on the semantic communication information retransmitted by the second terminal to obtain semantic information.
- the receiving module is further configured to receive updated model parameters sent by the server. These updated model parameters are obtained by updating the global model based on the approximate features after the server performs semantic processing on the approximate features to obtain semantic information. This is done when the server receives semantic communication completion information sent by the second terminal.
- the above-mentioned device also includes:
- the update module is used to update the local semantic knowledge base model based on the updated model parameters.
- the receiving module is also used to receive semantic processing failure feedback sent by the server.
- the sending module 1102 is also used to send semantic processing failure feedback to the second terminal.
- Each module in the aforementioned semantic communication device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in the first terminal in hardware form, or stored in the memory of the first terminal in software form, so that the processor can call and execute the operations corresponding to each module.
- a semantic communication device disposed in a second terminal, the device comprising:
- the receiving module 1201 is used to receive semantic communication information sent by the first terminal and perform semantic processing on the semantic communication information;
- the sending module 1202 is used to send the semantic information obtained from semantic processing to the first terminal.
- the receiving module 1201 is specifically used to perform semantic processing on the semantic communication information through a global model; if no semantic information is obtained, the semantic communication information is decomposed to obtain approximate features, and the approximate features are then semantically processed through a global model.
- the receiving module 1201 is further configured to receive semantic communication completion information sent by the second terminal, which is determined by the first terminal based on semantic information, which is obtained by the server through semantic processing of approximate features using a global model.
- the above-mentioned device also includes:
- the update module is used to update the model parameters of the global model based on approximate features, and obtain the updated model parameters.
- the sending module 1202 is also used to send the updated model parameters to the first terminal.
- the sending module 1202 is further configured to send a semantic processing failure feedback to the first terminal if semantic information is not obtained during the semantic processing of approximate features through the global model.
- the above-mentioned apparatus further includes:
- the deletion module is used to delete the approximate features obtained from the disassembly if multiple failure notifications are received from the second terminal.
- Each module in the aforementioned semantic communication device can be implemented entirely or partially through software, hardware, or a combination thereof.
- These modules can be embedded in or independent of the server's processor in hardware form, or stored in the server's memory in software form, so that the processor can call and execute the operations corresponding to each module.
- FIG. 13 is a schematic diagram of a terminal structure provided in an embodiment of this application.
- the terminal 1300 shown in Figure 13 includes: at least one processor 1301, a memory 1302, at least one network interface 1304, and a user interface 1303.
- the various components in the terminal 1300 are coupled together via a bus system 1305.
- the bus system 1305 is used to implement communication between these components.
- the bus system 1305 also includes a power bus, a control bus, and a status signal bus. However, for clarity, all buses are labeled as bus system 1305 in Figure 1.
- the terminal 1300 also includes a transceiver 1306.
- the transceiver 1306 may be multiple elements, including a transmitter and a receiver, providing a unit for communicating with various other devices over a transmission medium.
- the user interface 1303 may include a display, keyboard, or clicking device (e.g., mouse, trackball, touchpad, or touchscreen).
- clicking device e.g., mouse, trackball, touchpad, or touchscreen.
- the memory 1302 in the embodiments of this application can be volatile memory or non-volatile memory, or may include both volatile and non-volatile memory.
- the non-volatile memory can be read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), or flash memory.
- the volatile memory can be random access memory (RAM), which is used as an external cache.
- RAM Static Random Access Memory
- DRAM Dynamic Random Access Memory
- SDRAM Synchronous DRAM
- DDRSDRAM Double Data Rate SDRAM
- ESDRAM Enhanced Synchronous DRAM
- SLDRAM Synchlink DRAM
- DRRAM Direct Rambus RAM
- memory 1302 stores elements, executable modules or data structures, or subsets thereof, or extended sets thereof: operating system 13021 and application program 13022.
- the operating system 13021 includes various system programs, such as a framework layer, a core library layer, and a driver layer, used to implement various basic business functions and handle hardware-based tasks.
- the application program 13022 includes various applications, such as a media player and a browser, used to implement various application functions. Programs implementing the methods of this application embodiment can be included in application program 13022.
- the program or instructions stored in memory 1302 are invoked; specifically, these may be programs or instructions stored in application program 13022.
- the processor 1301 is used to perform semantic processing on semantic communication information during the semantic communication process.
- the sender is used to send semantic communication information to the server if semantic processing fails, so that the server can process the semantic communication information, obtain the corresponding semantic information, and return the semantic information.
- the methods disclosed in the above embodiments of this application can also be applied to, implemented by, or cooperating with other components (e.g., a transceiver).
- the processor 1301 may be an integrated circuit chip with signal processing capabilities. During implementation, the steps of the above methods can be... The process is accomplished through integrated logic circuits in the hardware or through software instructions in the processor 1301.
- the processor 1301 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, or discrete hardware components. It can implement or execute the methods, steps, and logic block diagrams disclosed in the embodiments of this application.
- the general-purpose processor can be a microprocessor or any conventional processor.
- the steps of the methods disclosed in the embodiments of this application can be directly implemented by a hardware decoding processor, or by a combination of hardware and software modules in the decoding processor.
- the software modules can be located in random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, registers, or other mature storage media in the art. This storage medium is located in memory 1302.
- the processor 1301 reads the information in memory 1302 and, in conjunction with its hardware, completes the steps of the above methods.
- the embodiments described in this application can be implemented using hardware, software, firmware, middleware, microcode, or a combination thereof.
- 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 this application, or combinations thereof.
- ASICs application-specific integrated circuits
- DSPs digital signal processors
- DSPDs digital signal processing devices
- PLDs programmable logic devices
- FPGAs field-programmable gate arrays
- general-purpose processors controllers, microcontrollers, microprocessors, other electronic units for performing the functions described in this application, or combinations thereof.
- the technology described in the embodiments of this application can be implemented by modules (e.g., procedures, functions, etc.) that perform the functions described in the embodiments of this application.
- the software code can be stored in memory and executed by processor 1301.
- the memory can be implemented in processor 1301 or external to processor 1301.
- semantic processing includes semantic extraction and/or semantic restoration.
- the processor 1301 is specifically used to perform semantic processing on semantic communication information using a local semantic knowledge base model.
- the local semantic knowledge base model is the basic semantic knowledge base model built by the first terminal through federated learning.
- a receiver is used to receive semantic communication information sent by a second terminal.
- the receiver is also used to receive semantic information sent by the server;
- the semantic information is obtained by the server through semantic processing of the semantic communication information using a global model, or the semantic information is obtained by the server through semantic processing of the approximate features of the semantic communication information using a global model.
- the processor 1301 is further configured to obtain semantic information if the semantic processing is successful.
- the processor 1301 is further configured to determine, based on semantic information, whether the current semantic communication process has been completed;
- the transmitter is also used to send semantic communication completion information to the second terminal if the current semantic communication process is completed.
- the processor 1301 is also configured to perform operations corresponding to semantic information.
- the processor 1301 is specifically used to determine whether the semantic information satisfies the communication intent or communication goal corresponding to the semantic communication information; if the semantic information satisfies the communication intent or communication goal, then it is determined that the current semantic communication process is completed.
- the transmitter is further configured to send feedback information to the second terminal if the current semantic communication process is not completed, the feedback information being used to instruct the second terminal to re-execute the semantic communication process.
- the receiver is further configured to receive semantic communication information retransmitted by the second terminal, and to perform semantic processing on the semantic communication information retransmitted by the second terminal to obtain semantic information.
- the receiver is further configured to receive updated model parameters sent by the server. These updated model parameters are obtained after the server performs semantic processing on the approximate features to obtain the semantic information, and when the server receives the semantic communication completion information sent by the second terminal, the global model is updated with model parameters based on the approximate features.
- the processor is also used to update the local semantic knowledge base model based on the updated model parameters.
- the receiver is also used to receive semantic processing failure feedback sent by the server;
- the transmitter is also used to send semantic processing failure feedback to a second terminal.
- FIG 14 is a schematic diagram of a server structure provided in an embodiment of this application.
- the server 1400 shown in Figure 14 includes at least one processor 1401, a memory 1402, and at least one network interface 1404.
- the various components in the server 1400 are coupled together via a bus system 1405.
- the bus system 1405 is used to implement communication between these components.
- the bus system 1405 also includes a power bus, a control bus, and a status signal bus.
- the server also includes a transceiver 1406.
- the transceiver can be multiple elements, including a transmitter and a receiver, providing a unit for communicating with various other devices over a transmission medium.
- the memory 1402 in the embodiments of this application can be volatile memory or non-volatile memory, or may include both volatile and non-volatile memory.
- the non-volatile memory can be read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), or flash memory.
- the volatile memory can be random access memory (RAM), which is used as an external cache.
- RAM Static Random Access Memory
- DRAM Dynamic Random Access Memory
- SDRAM Synchronous DRAM
- DDRSDRAM Double Data Rate SDRAM
- ESDRAM Enhanced Synchronous DRAM
- SLDRAM Synchlink DRAM
- DRRAM Direct Rambus RAM
- memory 1402 stores elements, executable modules or data structures, or subsets thereof, or extended sets thereof, such as operating system 14021.
- Operating system 14021 includes various system programs, such as a framework layer, core library layer, and driver layer, used to implement various basic business functions and handle hardware-based tasks.
- the receiver is used to receive semantic communication information sent by the first terminal; the processor 1401 is used to perform semantic processing on the semantic communication information; and the transmitter is used to send the semantic information obtained from the semantic processing to the first terminal.
- Processor 1401 may be an integrated circuit chip with signal processing capabilities. In the implementation process, each step of the above method can be completed by the integrated logic circuit of the hardware in processor 1401 or by instructions in the form of software.
- the processor 1401 may 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, or discrete hardware components. It can implement or execute the methods, steps, and logic block diagrams disclosed in the embodiments of this application.
- the general-purpose processor may be a microprocessor or any conventional processor.
- the steps of the methods disclosed in the embodiments of this application can be directly embodied in the execution of a hardware decoding processor, or executed by a combination of hardware and software modules in the decoding processor.
- the software module can reside in a mature storage medium in the art, such as random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, or registers. This storage medium is located in memory 1402.
- Processor 1401 reads the information in memory 1402 and, in conjunction with its hardware, completes the steps of the above method.
- the embodiments described in this application can be implemented using hardware, software, firmware, middleware, microcode, or a combination thereof.
- 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, etc.
- ASICs application-specific integrated circuits
- DSPs digital signal processors
- DSPDs digital signal processing devices
- PLDs programmable logic devices
- FPGAs field-programmable gate arrays
- controllers etc.
- microcontrollers microcontrollers, microprocessors, other electronic units for performing the functions described in this application, or combinations thereof.
- the technology described in the embodiments of this application can be implemented by modules (e.g., procedures, functions, etc.) that perform the functions described in the embodiments of this application.
- the software code can be stored in memory and executed by processor 1401.
- the memory can be implemented in processor 1401 or external to processor 1401.
- the processor 1401 is specifically used to perform semantic processing on semantic communication information through a global model; if semantic information is not obtained, the semantic communication information is decomposed to obtain approximate features, and the approximate features are then semantically processed through a global model.
- the receiver is further configured to receive semantic communication completion information sent by the second terminal, which is determined by the first terminal based on semantic information, which is obtained by the server through semantic processing of approximate features using a global model.
- the processor 1401 is also used to update the model parameters of the global model based on the approximate features, so as to obtain the updated model parameters.
- the transmitter is also used to send updated model parameters to the first terminal.
- the transmitter is further configured to send a semantic processing failure feedback to the first terminal if semantic information is not obtained during the semantic processing of approximate features through a global model.
- the processor 1401 is also used to delete the approximate features obtained from the disassembly.
- a semantic communication system which, as shown in FIG1, includes a first terminal, a second terminal, and a server.
- the second terminal is used to send semantic communication information to the first terminal.
- the first terminal is used to execute the steps of the semantic communication method for the first terminal in the above embodiments.
- the server is used to execute the steps of the semantic communication method for the server in the above embodiments.
- the second terminal is further configured to perform a knowledge base search based on the communication information to be transmitted in order to obtain semantic communication information.
- the second terminal is also used to send the communication information to be transmitted to the server if the knowledge base search fails, so that the server can perform a knowledge base search based on the communication information to be transmitted in order to obtain semantic communication information.
- a computer-readable storage medium on which a computer program is stored, and when the computer program is executed by a processor, it performs the following steps:
- semantic communication information is sent to the server so that the server can process the semantic communication information, obtain the corresponding semantic information, and return the semantic information.
- semantic processing includes semantic extraction and/or semantic restoration.
- a computer program when executed by a processor, it specifically implements the following steps:
- Semantic processing of semantic communication information is performed using a local semantic knowledge base model.
- the local semantic knowledge base model is the basic semantic knowledge base model built by the first terminal through federated learning.
- the computer program when executed by a processor, it also performs the following steps:
- the computer program when executed by a processor, it also performs the following steps:
- the semantic information is obtained by the server through semantic processing of the semantic communication information using a global model, or the semantic information is obtained by the server through semantic processing of the approximate features of the semantic communication information using a global model.
- the computer program when executed by a processor, it also performs the following steps:
- semantic information is obtained.
- the computer program when executed by a processor, it also performs the following steps:
- the computer program when executed by a processor, it also performs the following steps:
- a computer program when executed by a processor, it specifically implements the following steps:
- the computer program when executed by a processor, it also performs the following steps:
- a feedback message is sent to the second terminal, which instructs the second terminal to re-execute the semantic communication process.
- the computer program when executed by a processor, it also performs the following steps:
- the system receives semantic communication information retransmitted by the second terminal and performs semantic processing on the retransmitted semantic communication information to obtain semantic information.
- the computer program when executed by a processor, it also performs the following steps:
- the server receives updated model parameters sent by the server. These updated model parameters are obtained by updating the global model parameters based on the approximate features after the server performs semantic processing on the approximate features.
- the server receives the semantic communication completion information sent by the second terminal, it updates the model parameters based on the approximate features.
- the computer program when executed by a processor, it also performs the following steps:
- the semantic processing failure feedback is sent to the second terminal.
- a computer-readable storage medium on which a computer program is stored, and when the computer program is executed by a processor, it performs the following steps:
- the semantic information obtained from semantic processing is sent to the first terminal.
- a computer program when executed by a processor, it specifically implements the following steps:
- Semantic processing of semantic communication information is performed through a global model
- semantic communication information is decomposed to obtain approximate features, and the approximate features are semantically processed through a global model.
- the computer program when executed by a processor, it also performs the following steps:
- the model parameters of the global model are updated to obtain the updated model parameters
- the updated model parameters are sent to the first terminal.
- the computer program when executed by a processor, it also performs the following steps:
- the computer program when executed by a processor, it also performs the following steps:
- Figure 15 is a schematic structural diagram of a chip according to an embodiment of this application.
- the chip 1500 shown in Figure 15 includes a processor 1510, which can call and run computer programs from memory to implement the methods in the embodiments of this application.
- chip 1500 may further include memory 1520.
- Processor 1510 may retrieve and run computer programs from memory 1520 to implement the methods in the embodiments of this application.
- the memory 1520 can be a separate device independent of the processor 1510, or it can be integrated into the processor 1510.
- the chip 1500 may further include an input interface 1530.
- the processor 1510 can control the input interface 1530 to communicate with other devices or chips; specifically, it can acquire information or data sent by other devices or chips.
- the chip 1500 may further include an output interface 1540.
- the processor 1510 can control the output interface 1540 to communicate with other devices or chips, specifically, to output information or data to other devices or chips.
- the chip 1500 can be applied to network functions or base stations in the embodiments of this application, and the chip 1500 can implement the corresponding processes implemented in the various methods of the embodiments of this application. For the sake of brevity, it will not be described in detail here.
- chip 1500 mentioned in the embodiments of this application may also be referred to as a system-on-a-chip, system chip, chip system, or system-on-a-chip, etc.
- This application also provides a computer program product containing instructions, which, when executed by a processor, performs the following steps:
- semantic communication information is sent to the server so that the server can process the semantic communication information, obtain the corresponding semantic information, and return the semantic information.
- semantic processing includes semantic extraction and/or semantic restoration.
- a computer program when executed by a processor, it specifically implements the following steps:
- Semantic processing of semantic communication information is performed using a local semantic knowledge base model.
- the local semantic knowledge base model is the basic semantic knowledge base model built by the first terminal through federated learning.
- the computer program when executed by a processor, it also performs the following steps:
- the computer program when executed by a processor, it also performs the following steps:
- the semantic information is obtained by the server through semantic processing of the semantic communication information using a global model, or the semantic information is obtained by the server through semantic processing of the approximate features of the semantic communication information using a global model.
- the computer program when executed by a processor, it also performs the following steps:
- semantic information is obtained.
- the computer program when executed by a processor, it also performs the following steps:
- the computer program when executed by a processor, it also performs the following steps:
- a computer program when executed by a processor, it specifically implements the following steps:
- the computer program when executed by a processor, it also performs the following steps:
- the computer program when executed by a processor, it also performs the following steps:
- the system receives semantic communication information retransmitted by the second terminal and performs semantic processing on the retransmitted semantic communication information to obtain semantic information.
- the computer program when executed by a processor, it also performs the following steps:
- the system receives updated model parameters from the server. These updated model parameters are semantic information obtained by the server after semantic processing of the approximate features.
- the system receives semantic communication completion information from the second terminal, it bases its analysis on the approximate features. This is obtained after updating the model parameters of the global model.
- the computer program when executed by a processor, it also performs the following steps:
- the semantic processing failure feedback is sent to the second terminal.
- This application also provides a computer program product containing instructions, which, when executed by a processor, performs the following steps:
- the semantic information obtained from semantic processing is sent to the first terminal.
- a computer program when executed by a processor, it specifically implements the following steps:
- Semantic processing is performed on semantic communication information based on the global model
- semantic communication information is decomposed to obtain approximate features, and the approximate features are semantically processed through a global model.
- the computer program when executed by a processor, it also performs the following steps:
- model parameters of the global model are updated to obtain the updated model parameters
- the updated model parameters are sent to the first terminal.
- the computer program when executed by a processor, it also performs the following steps:
- the computer program when executed by a processor, it also performs the following steps:
- Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, or optical storage, etc.
- Volatile memory can include random access memory (RAM) or external cache memory.
- RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM), etc.
Landscapes
- Engineering & Computer Science (AREA)
- Computer Networks & Wireless Communication (AREA)
- Signal Processing (AREA)
- General Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Computing Systems (AREA)
- Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Data Mining & Analysis (AREA)
- Computational Linguistics (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Artificial Intelligence (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Computer And Data Communications (AREA)
Abstract
本申请涉及一种语义通信方法、装置、终端、服务器、系统和存储介质。方法包括:第一终端对语义通信过程中的语义通信信息进行语义处理;若语义处理失败,第一终端则将所述语义通信信息发送至服务器,以由所述服务器处理所述语义通信信息并获得所述语义通信信息对应的语义信息,并返回所述语义信息。采用本方法能够提升语义通信的通信可靠性。
Description
相关申请
本申请要求2024年6月26日申请的,申请号为202410832278.9,名称为“语义通信方法、装置、终端、服务器、系统和存储介质”的中国专利申请的优先权,在此将其全文引入作为参考。
本申请涉及无线通信技术领域,尤其涉及一种语义通信方法、装置、终端、服务器、系统和存储介质。
随着云计算、大数据、物联网、人工智能等技术的快速发展,信息量(数据量)的增长趋势也越来越显著,面对海量的信息传输需求,出现了语义通信(Semantic Communication)的方式,以解决传输带宽受限等问题。
在语义通信中,发送端以及接收端都具有背景知识库,从而能够实现通信双方之间的语义传输。但是,目前的语义通信方式存在通信可靠性差的问题。
发明内容
本申请实施例提供一种语义通信方法、装置、终端、服务器、系统和存储介质,能够提升语义通信的通信可靠性。
本申请在第一方面提供了一种语义通信方法所述方法包括:对语义通信过程中的语义通信信息进行语义处理;若语义处理失败,则将所述语义通信信息发送至服务器,以由所述服务器处理所述语义通信信息并获得所述语义通信信息对应的语义信息,并返回所述语义信息。
在其中一些实施例中,所述语义处理包括语义提取和/或语义还原。
在其中一些实施例中,所述对语义通信过程中的语义通信信息进行语义处理,包括:利用本地语义知识库模型对所述语义通信信息进行语义处理。
在其中一些实施例中,所述本地语义知识库模型为所述第一终端通过联邦学习构建的语义知识库基础模型。
在其中一些实施例中,所述方法还包括:接收第二终端发送的所述语义通信信息。
在其中一些实施例中,所述方法还包括:接收所述服务器下发的所述语义信息;其中,所述语义信息为所述服务器通过全局模型对所述语义通信信息进行语义处理得到的,或者,所述语义信息为所述服务器通过所述全局模型对所述语义通信信息的近似特征进行语义处理得到的。
在其中一些实施例中,所述方法还包括:若语义处理成功,则得到所述语义信息。
在其中一些实施例中,所述方法还包括:根据所述语义信息,确定当前的语义通信过程是否完成;若确定当前的语义通信过程完成,则向所述第二终端发送语义通信完成信息。
在其中一些实施例中,所述向所述第二终端发送语义通信完成信息之后,所述方法还包括:执行与所述语义信息对应的操作。
在其中一些实施例中,所述根据所述语义信息,确定当前的语义通信过程是否完成,包括:确定所述语义信息是否满足所述语义通信信息对应的通信意图或者通信目标;若确定所述语义信息满足所述通信意图或者所述通信目标,则确定当前的语义通信过程完成。
在其中一些实施例中,所述方法还包括:若确定当前的语义通信过程未完成,则向所述第二终端发送反馈信息,所述反馈信息用于指示所述第二终端重新执行所述语义通信过程。
在其中一些实施例中,所述方法还包括:接收所述第二终端重新发送的语义通信信息,并对所述第二终端重新发送的语义通信信息进行语义处理,以得到所述语义信息。
在其中一些实施例中,所述方法还包括:接收所述服务器发送的更新的模型参数,所述更新的模型参数是所述服务器对所述近似特征进行语义处理得到的所述语义信息之后,在接收到所述第二终端发送的语义通信完成信息时,根据所述近似特征对所述全局模型进行模型参数更新后得到的;基于所述更新模型参数,更新所述本地语义知识库模型。
在其中一些实施例中,所述方法还包括:接收所述服务器发送的语义处理失败反馈;将所述语义处理失败反馈发送至第二终端。
本申请在第二方面提供了一种语义通信方法所述方法包括:接收第一终端发送的语义通信信息,并对所述语义通信信息进行语义处理;将语义处理得到的语义信息发送至所述第一终端。
在其中一些实施例中,所述对所述语义通信信息进行语义处理,包括:根据全局模型对所述语义通信信息进行语义处理;若未获得语义信息,则对所述语义通信信息进行信息拆解得到近似特征,并通过所述全局模型对所述近似特征进行语义处理。
在其中一些实施例中,所述方法还包括:接收第二终端发送的语义通信完成信息,其中所述语义通信完成信息是所述第一终端根据所述语义信息确定的,所述语义信息为服务器通过所述全局模型对所述近似特征进行语义处理得到的;根据所述近似特征,对所述全局模型进行模型参数更新,得到更新的模型参数;将所述更新模型参数发送至所述第一终端。
在其中一些实施例中,所述方法还包括:若通过所述全局模型对所述近似特征进行语义处理的过程中未获得语义信息,则向所述第一终端发送语义处理失败反馈。
在其中一些实施例中,所述方法还包括:若接收到所述第二终端发送的多次失败通知,则删除拆解得到的近似特征。
本申请在第三方面提供了一种语义通信装置所述装置包括:语义处理模块,用于对语义通信过程中的语义通信信息进行语义处理;发送模块,用于若语义处理失败,则将所述语义通信信息发送至服务器,以由所述服务器处理所述语义通信信息并获得所述语义通信信息对应的语义信息,并返回所述语义信息。
本申请在第四方面提供了一种语义通信装置所述装置包括:接收模块,用于接收第一终端发送的语义通信信息,并对所述语义通信信息进行语义处理;发送模块,用于将语义处理得到的语义信息发送至所述第一终端。
本申请在第五方面提供了一种终端。所述终端包括发送器和处理器;所述处理器,用于对语义通信过程中的语义通信信息进行语义处理;所述发送器,用于若语义处理失败,则将所述语义通信信息发送至服务器,以由所述服务器处理所述语义通信信息并获得所述语义通信信息对应的语义信息,并返回所述语义信息。
本申请在第六方面提供了一种服务器。所述服务器包括发送器、处理器和接收器:所述接收器,用于接收第一终端发送的语义通信信息;所述处理器,用于对所述语义通信信息进行语义处理;所述发送器,用于将语义处理得到的语义信息发送至所述第一终端。
本申请在第七方面提供了一种语义通信系统。所述语义通信系统包括第一终端、第二终端以及服务器;所述第二终端,用于向所述第一终端发送语义通信信息;所述第一终端,用于执行如上述第一方面所述的方法的步骤;所述服务器,用于执行如上述第二方面所述的方法的步骤。
在其中一些实施例中,所述第二终端还用于根据待传输通信信息进行知识库搜索,以得到所述语义通信信息;所述第二终端还用于若知识库搜索失败,则将所述待传输通信信
息发送至所述服务器,以由所述服务器根据所述待传输通信信息进行知识库搜索,以得到所述语义通信信息。
本申请在第八方面还提供了一种计算机可读存储介质。所述计算机可读存储介质上存储有计算机程序,所述计算机程序被处理器执行时实现如上述第一方面或者第二方面所述的方法的步骤。
本申请在第九方面还提供了一种芯片。所述芯片包括可编程逻辑电路和/或程序指令,当所述芯片运行时实现如上述第一方面或者第二方面所述方法的步骤。
本申请在第十方面还提供了一种计算机程序产品。所述计算机程序产品包括计算机程序,所述计算机程序被处理器执行时实现如上述第一方面或者第二方面所述的方法的步骤。
上述语义通信方法、装置、终端、服务器、系统和存储介质,第一终端对语义通信过程中的语义通信信息进行语义处理,若语义处理失败,第一终端则将语义通信信息发送至服务器,以由服务器处理语义通信信息并获得语义通信信息对应的语义信息,并返回语义信息,这样,相较于传统技术中发送端和接收端仅依靠本地的背景知识库进行语义解析的方式而言,本申请实施例在第一终端能力不足的场景下,即第一终端未能成功对语义通信信息进行语义处理的情况下,第一终端可以借用服务器的能力,通过服务器对语义通信信息进行语义处理,服务器再将获得的语义信息返回至第一终端,由此可以提升语义处理的成功率,保障语义通信过程的顺利进行,以及提升语义通信的通信可靠性。
为了更清楚地说明本申请实施例或相关技术中的技术方案,下面将对实施例或相关技术描述中所需要使用的附图作简单地介绍。显而易见地,下面描述中的附图仅仅是本申请的一些实施例,对于本领域普通技术人员来讲,在不付出创造性劳动的前提下,还可以根据这些附图获得其他的附图。
图1为本申请一实施例中语义通信方法的实施环境图。
图2为本申请一实施例中语义通信方法的流程示意图。
图3为本申请另一实施例中语义通信方法的流程示意图。
图4为本申请另一实施例中语义通信方法的流程示意图。
图5为本申请另一实施例中第一终端确定语义通信过程是否完成的流程示意图。
图6为本申请另一实施例中语义通信方法的流程示意图。
图7为本申请另一实施例中语义通信方法的流程示意图。
图8为本申请另一实施例中语义通信方法的流程示意图。
图9为本申请另一实施例中语义通信方法的流程示意图。
图10为本申请另一实施例中语义通信方法的流程示意图。
图11为本申请一实施例中语义通信装置的结构框图。
图12为本申请另一实施例中语义通信装置的结构框图。
图13为本申请一实施例中终端的结构示意图。
图14为本申请一实施例中服务器的结构示意图。
图15为本申请一实施例中芯片的示意性结构图。
为了使本申请的目的、技术方案及优点更加清楚明白,以下结合附图及实施例,对本申请进行进一步详细说明。应当理解,此处描述的具体实施例仅仅用以解释本申请,并不用于限定本申请。
随着云计算、大数据、物联网、人工智能等技术的快速发展,信息量(数据量)的增长趋势也越来越显著,面对海量的信息传输需求,出现了语义通信(Semantic Communication)
的方式,以解决传输带宽受限等问题。
语义通信可以将用户的需求和信息含义融入通信过程,语义通信的根本目的在于信息交换,它不追求对原始数据或者信号的保真,而是追求在通信双方之间进行语义的准确传输。因此,在语义通信中,发送端以及接收端都具有背景知识库,通信双方依据背景知识库进行语义解析,从而能够实现通信双方之间的语义传输。根据背景知识库的不同,可以将点对点语义通信分为三类:基于知识图谱的语义通信、基于语义基的语义通信、基于基础模型的语义通信。
然而,目前的语义通信方式存在通信可靠性差的问题。例如,发送端以及接收端依据各自背景知识库无法成功进行语义解析(如接收端依据本地的背景知识库无法成功进行语义还原),则会导致语义通信失败,进而导致语义通信的通信可靠性差。
鉴于此,本申请实施例提供了一种语义通信方法、装置、终端、服务器、系统和存储介质,可以提升语义通信的通信可靠性。
需要说明的是,本申请实施例所带来的有益效果或者所解决的技术问题并不限定于这一个,还可以是其它隐含或者关联的问题,具体可以参见下述实施例的描述。
图1为本申请实施例提供的一种语义通信方法的应用环境示意图。其中,第一终端102通过网络与第二终端104进行通信,第一终端102通过网络与服务器106进行通信,第二终端104通过网络与服务器108进行通信。
其中,第一终端102和第二终端104可以但不限于是各种个人计算机、笔记本电脑、智能手机、平板电脑、物联网设备和便携式可穿戴设备,物联网设备可为智能音箱、智能电视、智能空调、智能车载设备、投影设备等。便携式可穿戴设备可为智能手表、智能手环、头戴设备等。头戴设备可以为虚拟现实(Virtual Reality,VR)设备、增强现实(Augmented Reality,AR)设备、智能眼镜等。
服务器106和服务器108可以是同一服务器,也可以是不同的服务器。以服务器106为例,服务器106可以是独立的物理服务器,也可以是多个物理服务器构成的服务器集群或分布式系统,还可以是边缘服务器、提供云计算服务的云服务器等。服务器108的形式与服务器106类似,在此不再赘述。
在本申请一个实施例中,如图2所示,提供了一种语义通信方法,以该方法应用于图1中第一终端为例进行说明,包括以下步骤S201至S202。
步骤S201:第一终端对语义通信过程中的语义通信信息进行语义处理。
第一终端与第二终端进行语义通信。具体地,第一终端和第二终端可以是异构终端,例如,第一终端和第二终端的算力、硬件能力、电量等存在不同之处;可选地,第一终端和第二终端的相关参数也可以相同,在此不做具体限制。
本申请实施例中,第一终端的数量可以是一个或者多个,第二终端的数量也可以是一个或者多个,为了描述方便,下文实施例均以第一终端的数量和第二终端的数量均为一个为例进行说明。
在语义通信的过程中,第一终端可以是发送端,也可以是接收端。发送端是指进行语义通信过程中的发射方,其用于调制发送信息;接收端是指语义通信过程中的接收方,用于解调接收信息。以语义通信过程中一次单向的信息传输过程为例,发送端对待传输的通信信息进行语义提取,从而可以从待传输的通信信息中提取语义特征,发送端将该语义特征发送至接收端;接收端收到语义特征后对语义特征进行语义还原,还原出通信信息。
本申请实施例中,语义处理包括语义提取和/或语义还原。在第一终端作为发送端的情况下,该语义通信信息则可以是第一终端获取的待传输的通信信息,第一终端对该语义通信信息进行语义处理则可以是指第一终端对该语义通信信息进行语义提取,以得到语义通信信息对应的语义信息,该语义信息可以是从语义通信信息中提取的语义特征。
在第一终端作为接收端的情况下,第一终端接收第二终端发送的语义通信信息,该语义通信信息则可以是第二终端发送的语义特征,第一终端对该语义通信信息进行语义处理
则可以是指第一终端对该语义通信信息进行语义还原,以得到语义通信信息对应的语义信息,该语义信息可以是根据语义特征还原得到的完整通信信息。
本申请实施例中,第一终端可以利用知识图谱对语义通信信息进行语义处理,尝试获取语义通信信息对应的语义信息;可选地,第一终端还可以利用语义基对语义通信信息进行语义处理,尝试获取语义通信信息对应的语义信息。关于其他第一终端对语义通信信息进行语义处理的方式,将在下文实施例进行示例性地说明。
步骤S202:若语义处理失败,第一终端则将语义通信信息发送至服务器。
其中,该语义通信信息用于服务器处理并获得语义通信信息对应的语义信息,并返回语义信息。
第一终端按照上述步骤S201的方式,对语义通信信息进行语义处理,若语义处理成功,第一终端则得到语义通信信息对应的语义信息。
而若语义处理失败,即第一终端并未成功获得语义通信信息对应的语义信息,第一终端则将该语义通信信息发送至服务器。
服务器接收第一终端发送的语义通信信息后,由服务器对语义通信信息进行语义处理,而后,服务器将语义处理得到的语义信息发送至第一终端,关于服务器对语义通信信息进行语义处理的过程将在下文实施例进行说明。
这样,上述实施例在第一终端的处理能力不足的情况下,即第一终端未能成功对语义通信信息进行语义处理的情况下,第一终端可以借用服务器的能力,通过服务器对语义通信信息进行语义处理,服务器再将获得的语义信息返回至第一终端,这样可以提升语义处理的成功率,同时也提升语义通信的通信可靠性。
需要说明的是,在第一终端为接收端、第二终端为发送端的情况下,除了第一终端可以借用服务器的处理能力以通过服务器对语义通信信息进行语义处理之外,第二终端也可以借用服务器的处理能力。第二终端在获取语义通信信息时,首先根据待传输通信信息进行知识库搜索,尝试得到语义通信信息,若知识库搜索失败,第二终端则将待传输通信信息发送至服务器,服务器根据该待传输通信信息进行知识库搜索,以得到语义通信信息,这样可以保障语义通信过程的顺利进行。
在一个实施例中,基于图2所示的实施例并结合图3所示实施例,本实施例将示例性地介绍一种第一终端对语义通信信息进行语义处理的方式。如图3所示,本实施例中,步骤S201包括图3所示的步骤S2011。
步骤S2011:第一终端利用本地语义知识库模型对语义通信信息进行语义处理。
第一终端可以将语义通信信息输入至第一终端本地存储的本地语义知识库模型中,通过本地语义知识库模型对语义通信信息进行语义处理。
如上文所述,在第一终端作为发送端的情况下,该本地语义知识库模型可以对该语义通信信息进行语义提取,以得到语义通信信息对应的语义信息,该语义信息可以是从语义通信信息中提取的语义特征。在第一终端作为接收端的情况下,该本地语义知识库模型可以对该语义通信信息进行语义还原,以得到语义通信信息对应的语义信息,该语义信息可以是根据语义特征还原得到的完整通信信息。
其中,本地语义知识库模型为第一终端通过联邦学习构建的语义知识库基础模型。
联邦学习(Federated Machine Learning)是通过交换本地训练的模型参数实现多方联合建模。本申请实施例中,服务器可以联合第一终端和/或第二终端进行联邦学习,学习完成后服务器获得全局模型,服务器可以将全局模型的模型参数下发至第一终端以及第二终端;可选地,第一终端存储该全局模型作为本地语义知识库模型;可选地,第一终端还可利用本地数据对全局模型进行模型优化,优化后得到本地语义知识库模型。
第一终端获取到语义通信过程中的语义通信信息之后,则可以通过该本地语义知识库模型对语义通信信息进行语义处理,即利用本地语义知识库模型进行语义提取或者语义还原,以得到语义信息。
请继续参见图3,如上文所述,若语义处理失败,第一终端则将语义通信信息发送至服务器。
例如,第一终端将语义通信信息输入至本地语义知识库模型后,并未获得本地语义知识库模型输出的语义通信信息对应的语义信息,则确定语义处理失败,第一终端则将语义通信信息发送至服务器;服务器接收第一终端发送的语义通信信息后,由服务器对语义通信信息进行语义处理,而后,服务器将语义处理得到的语义信息发送至第一终端。
若语义处理成功,第一终端则得到语义通信信息对应的语义信息。
上述实施例通过联邦学习构建本地语义知识库模型,参与联合建模的各方在交换模型参数的过程中实现模型优化,获得更好的模型性能。联邦学习能满足原始数据不出本地且适用更多复杂场景的联合建模,在联合建模过程中避免数据/模型隐私泄露等安全风险。
在一个实施例中,基于图2所示的实施例并结合图4所示实施例,本实施例中,语义通信方法还包括图4所示的步骤S401。
步骤S401:第一终端接收服务器下发的语义信息。
若服务器成功对第一终端发送的语义通信信息进行语义处理,得到语义信息,服务器则将语义信息发送至第一终端,从而第一终端可以接收到服务器下发的语义信息。
其中,语义信息为服务器根据全局模型对语义通信信息进行语义处理得到的,或者,语义信息为服务器根据全局模型对语义通信信息的近似特征进行语义处理得到的。
本申请实施例中,全局模型是服务器通过联邦学习构建的语义知识库基础模型,如上文所述,服务器可以联合第一终端和/或第二终端进行联邦学习,学习完成后服务器获得全局模型,服务器可以将全局模型的模型参数下发至第一终端以及第二终端。第一终端存储该全局模型作为本地语义知识库模型;可选地,第一终端还可利用本地数据对全部模型进行模型优化,优化后得到本地语义知识库模型。
因此,服务器接收到第一终端发送的语义通信信息后,服务器首先根据本地存储的全局模型对语义通信信息进行语义处理,示例性地,服务器可以将语义通信信息输入至全局模型中,尝试通过全局模型对语义通信信息进行语义处理,以获得语义信息。
可以理解的是,与第一终端中本地语义知识库模型的作用类似,在第一终端作为发送端的情况下,服务器中本地存储的全局模型可以进行语义提取,以得到语义信息,该语义信息可以是语义通信信息对应的语义特征;在第一终端作为接收端的情况下,服务器中本地存储的全局模型可以进行语义还原,以得到语义信息,该语义信息可以是根据语义特征还原得到的完整通信信息。下文实施例若无特别说明,均以第一终端作为接收端为例进行示例。
服务器若根据全局模型对语义通信信息进行语义处理,且成功获得语义信息,服务器则将获得的语义信息返回至第一终端。
而若服务器根据全局模型对语义通信信息进行语义处理时未获得语义信息,服务器则对语义通信信息进行信息拆解得到近似特征,并根据全局模型对近似特征进行语义处理。
作为一种实施方式,服务器可以将语义通信信息输入至近似特征拆解模型中,得到语义通信信息的近似特征;而后,服务器将近似特征输入至全局模型中,通过全局模型对近似特征进行语义处理,尝试获得语义信息;若成功获得语义信息,服务器则将获得的语义信息返回至第一终端。
在一些实施例中,若服务器根据全局模型对近似特征进行语义处理的过程中未获得语义信息,则表示服务器语义处理失败,服务器则向第一终端发送语义处理失败反馈。在一些实施例中,若服务器对该语义通信信息进行信息拆解未得到近似特征,同样表示服务器语义处理失败,服务器则向第一终端发送语义处理失败反馈。
第一终端在接收到服务器发送的语义处理失败反馈后,将语义处理失败反馈发送至第二终端;在一些实施例中,第二终端将语义处理失败反馈发送至服务器,服务器则确定重新进行语义处理(例如,重新进行信息拆解得到近似特征,并根据全局模型对重新拆解的
近似特征进行语义处理),以获得语义信息。关于重新进行语义处理的过程将在下文实施例说明。而若服务器接收到第二终端发送的多次失败通知,服务器则删除拆解得到的近似特征,结束当前语义通信过程。
这样,第一终端对语义通信信息进行语义处理时,若第一终端语义处理成功,第一终端则获得语义通信信息对应的语义信息;若第一终端语义处理失败,第一终端则将语义通信信息发送至服务器,若服务器进行语义处理成功后,服务器则向第一终端返回语义通信信息对应的语义信息。
可以理解的是,若第一终端语义处理成功得到了语义信息,或者,服务器进行语义处理成功得到了语义信息,但是,该语义信息不准确,这种情况下,也不能判定为语义通信过程完成。
鉴于此,参见图5,本申请实施例语义通信方法还包括图5所示的步骤S501-S504。
步骤S501:第一终端根据语义信息,确定当前的语义通信过程是否完成。
第一终端自行成功对语义通信信息进行语义处理得到语义信息,或者,第一终端接收到服务器成功对语义通信信息进行语义处理后返回的语义信息,第一终端根据该语义信息,确定当前的语义通信过程是否完成。
在一种可行的实施方式中,第二终端向第一终端发送语义通信信息时,还可以一并发送当前语义通信过程对应的通信目标和/或通信意图;第二终端作为发送端,感知通信环境,获得待传输的通信信息,然后根据待传输的通信信息,搜索知识库,提取需要传输的语义通信信息(在一些实施例中,还可以是提取通信目标和/或通信意图)并发送至第一终端;第一终端根据接收到的语义通信信息(在一些实施例中,还可以结合感知的通信环境以及第二终端发送的通信目标和/或通信意图),进行语义处理,或者,第一终端通过服务器进行语义处理,得到语义信息。这样,第一终端根据语义信息确定当前的语义通信过程是否完成时,具体地,第一终端则可以确定语义信息是否满足语义通信信息对应的该通信意图或者该通信目标,例如,第一终端从语义通信信息中提取出通信意图或者通信目标,然后将提取的通信意图或者通信目标与第二终端发送的通信目标或者通信意图进行对比,以确定语义信息是否满足第二终端发送的通信意图或者通信目标,若语义信息满足通信意图或者通信目标,第一终端则确定当前的语义通信过程完成,即第一终端确定语义信息能形成符合通信意图或者通信目标的完整语义,则确定当前的语义通信过程完成。按照这种方式,可以保障通信双方对是否完成通信达成一致的判断标准,保障正常的语义通信进行。
在另一种可行的实施方式中,第一终端也可以将语义通信信息以及其对应的语义信息输入预先训练的分类模型中,由分类模型判断该语义信息是否符合语义通信信息对应的通信意图或者通信目标。
步骤S502:若确定当前的语义通信过程完成,第一终端则向第二终端发送语义通信完成信息。
步骤S503:第一终端执行与语义信息对应的操作。
若确定当前的语义通信过程完成,第一终端则向第二终端发送语义通信完成信息。在一些实施例中,若语义信息指示第一终端执行某种操作,第一终端则执行与语义信息对应的操作。
步骤S504:若确定当前的语义通信过程未完成,第一终端则向第二终端发送反馈信息。
具体地,反馈信息用于指示第二终端重新执行语义通信过程。
第二终端根据第一终端的反馈信息,判断上下文信息,重新获得语义通信信息,例如,第二终端重新对待传输的通信信息进行语义提取,从而可以从待传输的通信信息中提取语义特征作为需要重新向第一终端发送的语义通信信息,第二终端将该语义通信信息发送至第一终端。
第一终端接收第二终端重新发送的语义通信信息后,对第二终端重新发送的语义通信信息进行语义处理,以得到语义信息。
与上述图2所示的实施例类似,第一终端首先尝试自行对语义通信过程中的语义通信信息进行语义处理,若语义处理成功,第一终端则获得语义信息;若语义处理失败,第一终端则重新将语义通信信息发送至服务器,由服务器重新对语义通信信息进行语义处理;服务器首先根据本地存储的全局模型对语义通信信息进行语义处理,若成功获得语义信息,服务器则将获得的语义信息返回至第一终端;若服务器根据全局模型对语义通信信息进行语义处理后未获得语义信息,服务器则对该语义通信信息进行信息拆解得到近似特征,并根据全局模型对近似特征进行语义处理,若成功获得语义信息,服务器则将获得的语义信息返回至第一终端,若语义处理失败,服务器则向第一终端发送语义处理失败反馈。
本实施例中,第一终端和服务器进行语义处理时,还可以结合上下文信息搜索本地的语义知识库基础模型,解析语义通信信息,以得到语义信息。
上述实施例中,服务器对第一终端发送的语义通信信息进行信息拆解得到近似特征,当服务器通过全局模型对近似特征进行语义处理,且成功获得语义信息时,服务器将获得的语义信息返回至第一终端。当第一终端根据该语义信息判断当前的语义通信过程完成时,第一终端则向第二终端发送语义通信完成信息。
服务器接收第二终端发送的语义通信完成信息后,服务器根据近似特征,对全局模型进行模型参数更新,得到更新的模型参数,即服务器将近似特征作为训练样本对全局模型进行训练,以优化全局模型,优化后得到全局模型最新的模型参数:更新模型参数后,服务器将更新的模型参数发送至第一终端,供第一终端更新本地语义知识库模型;在一些实施例中,服务器还可以按照语义通信完成信息的指示,执行对应的操作。
第一终端接收到服务器发送的更新的模型参数后,第一终端则基于更新的模型参数,更新本地语义知识库模型。
这样,上述实施例服务器和终端可以根据自身处理能力和通信数据进行语义知识库基础模型的更新,最大化覆盖各类近似特征,由此提高通信的准确性,以及保障通信质量。
在一个实施例中,如图6所示,本申请提供了一种语义通信方法,用于服务器,该方法包括如下步骤S601至S602。
步骤S601:服务器接收第一终端发送的语义通信信息,并对语义通信信息进行语义处理。
步骤S602:服务器将通过语义处理得到的语义信息发送至第一终端。
在一个实施例中,基于图6所示的实施例并结合图7所示实施例,步骤S601包括图7所示的步骤S701、步骤S702以及步骤S703。
步骤S701:服务器接收第一终端发送的语义通信信息。
步骤S702:服务器根据全局模型对语义通信信息进行语义处理。
步骤S703:若未获得语义信息,服务器则对语义通信信息进行信息拆解得到近似特征,并通过全局模型对近似特征进行语义处理。
在一个实施例中,该方法还包括:
服务器接收第二终端发送的语义通信完成信息,其中语义通信完成信息是第一终端根据语义信息确定的,语义信息为服务器通过全局模型对近似特征进行语义处理得到的;
服务器根据近似特征,对全局模型进行模型参数更新,得到更新的模型参数;
服务器将更新的模型参数发送至第一终端。
在一个实施例中,该方法还包括:
若通过全局模型对近似特征进行语义处理的过程中未获得语义信息,服务器则向第一终端发送语义处理失败反馈。
在一个实施例中,该方法还包括:
若接收到第二终端发送的多次失败通知,服务器则删除拆解得到的近似特征。
关于用于服务器的语义通信方法的实施方式,可以参见上文中用于第一终端的语义通信方法的实施方式中的相关描述。另外,用于服务器的语义通信方法的有益效果与上文中
用于第一终端的语义通信方法的有益效果类似,在此不再赘述。
在本申请一个实施例中,提供了一种语义通信方法,如图8所示,包括如下步骤S801至S808。
步骤S801:第一终端接收第二终端发送的语义通信信息。
步骤S802:第一终端利用本地语义知识库模型对语义通信信息进行语义处理,若语义处理失败,第一终端则将语义通信信息发送至服务器。
步骤S803:服务器通过全局模型对语义通信信息进行语义处理,若未获得语义信息,服务器则对语义通信信息进行信息拆解得到近似特征,并通过全局模型对近似特征进行语义处理。
步骤S804:服务器将语义处理获得的语义信息发送至第一终端。
步骤S805:第一终端根据语义信息,确定当前的语义通信过程是否完成,若当前的语义通信过程完成,第一终端则向第二终端发送语义通信完成信息。
步骤S806:第二终端向服务器发送语义通信完成信息。
步骤S807:服务器接收到第二终端发送的语义通信完成信息之后,服务器根据近似特征,对全局模型进行模型参数更新,得到更新的模型参数。
步骤S808:服务器将更新的模型参数发送至第一终端和/或第二终端。
在本申请一个实施例中,提供了一种语义通信方法,如图9所示,包括如下步骤S901至S912。
步骤S901:第一终端接收第二终端发送的语义通信信息。
步骤S902:第一终端利用本地语义知识库模型对语义通信信息进行语义处理,若语义处理失败,第一终端则将语义通信信息发送至服务器。
步骤S903:服务器通过全局模型对语义通信信息进行语义处理,若未获得语义信息,服务器则对语义通信信息进行信息拆解得到近似特征,并通过全局模型对近似特征进行语义处理。
步骤S904:服务器将语义处理得到的语义信息发送至第一终端。
步骤S905:第一终端根据语义信息,确定当前的语义通信过程是否完成,若当前的语义通信过程未完成,第一终端则向第二终端发送反馈信息。
步骤S906:第二终端向第一终端重新发送语义通信信息。
步骤S907:第二终端向服务器发送语义处理失败反馈。
步骤S908:第一终端对第二终端重新发送的语义通信信息进行语义处理,若语义处理失败,第一终端则将语义通信信息重新发送至服务器。
步骤S909:服务器重新通过全局模型对语义通信信息进行语义处理,若未获得语义信息,服务器则对语义通信信息重新进行信息拆解得到近似特征,并通过全局模型对近似特征进行语义处理,服务器将语义处理得到的语义信息发送至第一终端。
步骤S910:第一终端根据语义信息,确定当前的语义通信过程是否完成,若当前的语义通信过程未完成,第一终端则向第二终端发送反馈信息。
步骤S911:若语义通信过程未完成的次数达到阈值,第二终端向服务器发送多次失败通知。
步骤S912:服务器收到多次失败通知后删除拆解得到的近似特征,并保留全局模型的模型参数。
在本申请一个实施例中,提供了一种语义通信方法,如图10所示,包括如下步骤S1001至S1004。
步骤S1001:第一终端接收第二终端发送的语义通信信息。
步骤S1002:第一终端利用本地语义知识库模型对语义通信信息进行语义处理,若语义处理成功,第一终端则得到语义信息。
步骤S1003:第一终端根据语义信息,确定当前的语义通信过程是否完成。
步骤S1004:若当前的语义通信过程完成,第一终端则向第二终端发送语义通信完成信息。
以下,结合两个示例,对本申请语义通信方法进行举例说明。
示例一:存在多个智慧家庭终端:终端1、终端2、终端3以及终端4(在上述第一终端作为接收端、第二终端作为发送端的情况下,终端1相当于上述的第二终端,终端2、终端3以及终端4相当于上述第一终端),其中,终端1为手机,终端2为电脑,终端3为扫地机器人,终端4为空调,多个智慧家庭终端均可连接在室内家庭算力中心(相当于上述的服务器)。
1)终端1感知通信环境,获得通信场景信息:终端1处于室外,当前时刻为准备下班时间,终端1所处的环境噪声干扰小、可用传输带宽资源多,终端1需要将室内终端设置到欢迎回家模式。
2)终端1根据感知的通信场景信息,确定通信目标为:需要将室内终端设置到欢迎回家模式,而后,终端1根据通信场景信息搜索知识库,提取需要传输的语义通信信息(可选地,还可以提取通信意图),终端1将语义通信信息(可选地,携带通信目标和/或通信意图)分别发送至终端2、终端3以及终端4。
3)终端2感知通信环境:终端2处于常规位置,现在是5:00,待机状态;终端3感知通信环境:终端3处于卧室,现在是5:00,扫拖状态;终端4感知通信环境:终端4处于常规位置,现在是5:00,室内温度30°,湿度90%。
4)终端2、终端3以及终端4接收到来自于终端1的语义通信信息,终端2、终端3以及终端4分别对语义通信信息进行语义处理(语义还原),尝试还原出语义通信信息对应的完整的语义信息。
5)终端2语义处理后得到语义信息:设置为欢迎模式,终端2判断该语义信息的语义完整且符合通信意图,终端2确定当前的语义通信过程完成。
终端2从待机模式转为开机模式,并将语义通信完成信息发送给终端1,结束通信。
6)终端3和终端4语义处理失败,终端3和终端4分别将接收到的语义通信信息传输至室内家庭算力中心。
7)室内家庭算力中心通过全局模型,对语义通信信息进行语义处理(语义还原)。
8)室内家庭算力中心语义处理成功,得到语义信息为:设置为欢迎回家模式,室内家庭算力中心将语义信息下发给终端3和终端4。
9)终端3判断该语义信息的语义完整且符合通信意图,终端3确定当前的语义通信过程完成。
终端3从工作模式转为待机模式,并将语义通信完成信息发送给终端1,结束通信。
10)终端4判断该语义信息的语义完整且符合通信意图,终端4确定当前的语义通信过程完成。
终端4将空调打开并设置温度为23°,并将语义通信完成信息发送给终端1,结束通信。
示例二:存在多个元宇宙终端:终端1、终端2、终端3以及终端4(在上述第一终端作为接收端、第二终端作为发送端的情况下,终端1相当于上述的第二终端,终端2、终端3以及终端4相当于上述第一终端),其中,终端1为游戏主机,终端2为扩展现实(Extended Reality,ER)眼镜,终端2连接在算力服务器1;终端3为增强现实(Augmented Reality,AR)眼镜,终端3连接在算力服务器1;终端4为普通翻译眼镜,终端4连接在算力服务器2。
1)终端1感知通信环境,获得通信场景信息:终端1可用传输带宽资源多。
2)终端1根据感知的通信场景信息,确定通信目标为:需要更新游戏内容,而后,终端1根据通信场景信息搜索知识库,提取需要传输的语义通信信息(可选地,还可以提取通信意图),终端1将语义通信信息(可选地,携带通信目标和/或通信意图)分别发送至终端2、终端3以及终端4。
3)终端2感知通信环境:终端2处于常规位置,已连接无线保真(Wireless Fidelity,WIFI)网络,佩戴状态;终端3感知通信环境:终端3处于卧室,已连接WIFI,会议状态;终端4感知通信环境:终端4处于常规位置,开机数据模式。
4)终端2、终端3以及终端4接收到来自于终端1的语义通信信息,终端2、终端3以及终端4分别对语义通信信息进行语义处理(语义还原),尝试还原出语义通信信息对应的完整的语义信息。
5)终端2语义处理后得到语义信息:皮肤更新,终端2判断该语义信息的语义完整且符合通信意图,终端2确定当前的语义通信过程完成。
终端2将游戏皮肤更新,并将语义通信完成信息发送给终端1,结束通信。
6)终端3和终端4语义处理失败,终端3将接收到的语义通信信息传输至算力服务器1,终端4将接收到的语义通信信息传输至算力服务器2。
7)算力服务器1通过全局模型,对语义通信信息进行语义处理(语义还原),得到语义信息为:更新皮肤状态,算力服务器1将语义信息下发给终端3。
8)终端3判断该语义信息的语义完整且符合通信意图,终端3确定当前的语义通信过程完成,进行游戏皮肤更新;若语义信息不符合通信意图或者通信目标,则不进行游戏皮肤更新。
终端3语义通信完成信息发送给终端1,结束通信。
9)算力服务器2通过全局模型,未能成功对语义通信信息进行语义处理(语义还原)。
10)算力服务器2对语义通信信息进行信息拆解得到近似特征:镜片颜色(此时算力服务器2拆解得到的近似特征不准确),算力服务器2将近似特征输入全局模型,得到提取结果(语义信息)返回给终端4。
11)终端4根据该语义信息,确定当前的语义通信过程未完成,则向终端1发送反馈信息,重新进行通信。
12)终端1根据终端4的反馈信息,判断上下文信息,重新搜索知识库,重新获得语义通信信息:更新游戏角色皮肤,并发给终端4。
13)终端4再次根据接收到的语义通信信息,判断上下文信息,重新进行语义处理(语义还原),未能还原出语义信息,终端4将语义通信信息发送给算力服务器2。
14)算力服务器2对该语义通信信息进行信息拆解,未能提取到近似特征,算力服务器2将语义处理失败反馈发送给终端4,算力服务器2删除相关数据,保留原模型参数不更新。
15)终端4反馈终端1:语义处理失败反馈。
应该理解的是,虽然上述流程图中的各个步骤按照箭头的指示依次显示,但是这些步骤并不是必然按照箭头指示的顺序依次执行。除非本文中有明确的说明,这些步骤的执行并没有严格的顺序限制,这些步骤可以以其它的顺序执行。而且,上述流程图中的至少一部分步骤可以包括多个步骤或者多个阶段,这些步骤或者阶段并不必然是在同一时刻执行完成,而是可以在不同的时刻执行,这些步骤或者阶段的执行顺序也不必然是依次进行,而是可以与其它步骤或者其它步骤中的步骤或者阶段的至少一部分轮流或者交替地执行。
在本申请一个实施例中,如图11所示,还提供了一种语义通信装置,设置于第一终端,包括:
语义处理模块1101,用于对语义通信过程中的语义通信信息进行语义处理;
发送模块1102,用于若语义处理失败,则将所述语义通信信息发送至服务器,以由服务器处理该语义通信信息并获得该语义通信信息对应的语义信息,并返回该语义信息。
在一个实施例中,语义处理包括语义提取和/或语义还原。
在一个实施例中,语义处理模块1101具体用于利用本地语义知识库模型对语义通信信息进行语义处理。
在一个实施例中,本地语义知识库模型为第一终端通过联邦学习构建的语义知识库基
础模型。
在一个实施例中,上述装置还包括:
接收模块,用于接收第二终端发送的语义通信信息。
在一个实施例中,接收模块还用于接收服务器下发的语义信息。
其中,语义信息为服务器通过全局模型对语义通信信息进行语义处理得到的,或者,语义信息为服务器通过所述全局模型对语义通信信息的近似特征进行语义处理得到的。
在一个实施例中,上述装置还包括:
获取模块,用于若语义处理成功,则得到语义信息。
在一个实施例中,上述装置还包括:
确定模块,用于根据语义信息,确定当前的语义通信过程是否完成。
发送模块1102,还用于若当前的语义通信过程完成,则向第二终端发送语义通信完成信息。
在一个实施例中,上述装置还包括:
执行模块,用于执行与语义信息对应的操作。
在一个实施例中,确定模块具体用于确定语义信息是否满足语义通信信息对应的通信意图或者通信目标;若语义信息满足通信意图或者通信目标,则确定当前的语义通信过程完成。
在一个实施例中,发送模块1102,还用于若当前的语义通信过程未完成,则向第二终端发送反馈信息,该反馈信息用于指示第二终端重新执行语义通信过程。
在一个实施例中,接收模块还用于接收第二终端重新发送的语义通信信息,并对该第二终端重新发送的语义通信信息进行语义处理,以得到语义信息。
在一个实施例中,接收模块还用于接收服务器发送的更新的模型参数,该更新的模型参数是服务器对近似特征进行语义处理得到的语义信息之后,当接收到第二终端发送的语义通信完成信息时,根据该近似特征对全局模型进行模型参数更新后得到的。
上述装置还包括:
更新模块,用于基于更新的模型参数,更新本地语义知识库模型。
在一个实施例中,接收模块还用于接收服务器发送的语义处理失败反馈。
发送模块1102,还用于将语义处理失败反馈发送至第二终端。
关于设置于第一终端中的语义通信装置的具体限定,可以参见上文中对于用于第一终端中的语义通信方法的限定,在此不再赘述。上述语义通信装置中的各个模块可全部或部分通过软件、硬件及其组合来实现。上述各模块可以硬件形式内嵌于或独立于第一终端中的处理器中,也可以以软件形式存储于第一终端中的存储器中,以便于处理器调用执行以上各个模块对应的操作。
在本申请一个实施例中,如图12所示,还提供了一种语义通信装置,设置于第二终端,该装置包括:
接收模块1201,用于接收第一终端发送的语义通信信息,并对该语义通信信息进行语义处理;
发送模块1202,用于将语义处理得到的语义信息发送至所述第一终端。
在一个实施例中,接收模块1201具体用于通过全局模型对所述语义通信信息进行语义处理;若未获得语义信息,则对语义通信信息进行信息拆解得到近似特征,并通过全局模型对该近似特征进行语义处理。
在一个实施例中,接收模块1201还用于接收第二终端发送的语义通信完成信息,该语义通信完成信息是由第一终端根据语义信息确定的,该语义信息是服务器通过全局模型对近似特征进行语义处理得到的。
上述装置还包括:
更新模块,用于根据近似特征,对全局模型进行模型参数更新,得到更新的模型参数;
发送模块1202,还用于将更新的模型参数发送至第一终端。
在一个实施例中,发送模块1202还用于若通过全局模型对近似特征进行语义处理的过程中未获得语义信息,则向第一终端发送语义处理失败反馈。
在一个实施例中,上述装置还包括:
删除模块,用于若接收到所述第二终端发送的多次失败通知,则删除拆解得到的近似特征。
关于设置于服务器的语义通信装置的具体限定可以参见上文中对于用于服务器中的语义通信方法的限定,在此不再赘述。上述语义通信装置中的各个模块可全部或部分通过软件、硬件及其组合来实现。上述各模块可以硬件形式内嵌于或独立于服务器中的处理器中,也可以以软件形式存储于服务器中的存储器中,以便于处理器调用执行以上各个模块对应的操作。
图13为本申请实施例提供的一种终端的结构示意图。图13所示的终端1300包括:至少一个处理器1301、存储器1302、至少一个网络接口1304和用户接口1303。终端1300中的各个组件通过总线系统1305耦合在一起。可理解,总线系统1305用于实现这些组件之间的连接通信。总线系统1305除包括数据总线之外,还包括电源总线、控制总线和状态信号总线。但是为了清楚说明起见,在图1中将各种总线都标为总线系统1305。另外,本申请实施例中,上述终端1300还包括收发器1306,收发器1306可以是多个元件,即包括发送器和接收器,提供用于在传输介质上与各种其他装置通信的单元。
其中,用户接口1303可以包括显示器、键盘或者点击设备(例如,鼠标,轨迹球(trackball)、触感板或者触摸屏等。
可以理解,本申请实施例中的存储器1302可以是易失性存储器或非易失性存储器,或可包括易失性和非易失性存储器两者。其中,非易失性存储器可以是只读存储器(Read-OnlyMemory,ROM)、可编程只读存储器(Programmable ROM,PROM)、可擦除可编程只读存储器(ErasablePROM,EPROM)、电可擦除可编程只读存储器(ElectricallyEPROM,EEPROM)或闪存。易失性存储器可以是随机存取存储器(RandomAccessMemory,RAM),其用作外部高速缓存。通过示例性但不是限制性说明,许多形式的RAM可用,例如静态随机存取存储器(StaticRAM,SRAM)、动态随机存取存储器(DynamicRAM,DRAM)、同步动态随机存取存储器(SynchronousDRAM,SDRAM)、双倍数据速率同步动态随机存取存储器(DoubleDataRate SDRAM,DDRSDRAM)、增强型同步动态随机存取存储器(Enhanced SDRAM,ESDRAM)、同步连接动态随机存取存储器(SynchlinkDRAM,SLDRAM)和直接内存总线随机存取存储器(DirectRambusRAM,DRRAM)。本申请实施例描述的系统和方法的存储器1302旨在包括但不限于这些和任意其它适合类型的存储器。
在一些实施方式中,存储器1302存储了如下的元素、可执行模块或者数据结构、或者他们的子集、或者他们的扩展集:操作系统13021和应用程序13022。
其中,操作系统13021,包含各种系统程序,例如框架层、核心库层、驱动层等,用于实现各种基础业务以及处理基于硬件的任务。应用程序13022,包含各种应用程序,例如媒体播放器(MediaPlayer)、浏览器(Browser)等,用于实现各种应用业务。实现本申请实施例方法的程序可以包含在应用程序13022中。
在本申请实施例中,通过调用存储器1302存储的程序或指令,具体地,可以是应用程序13022中存储的程序或指令。其中,处理器1301用于对语义通信过程中的语义通信信息进行语义处理。
发送器用于若语义处理失败,则将语义通信信息发送至服务器,以由服务器处理语义通信信息并获得语义通信信息对应的语义信息,并返回语义信息。
上述本申请实施例揭示的部分或者全部方法还可以应用于处理器1301中,或者由处理器1301实现,或者由处理器1301与其他元件(例如收发机)配合实现。处理器1301可能是一种集成电路芯片,具有信号的处理能力。在实现过程中,上述方法的各步骤可以
通过处理器1301中的硬件的集成逻辑电路或者软件形式的指令完成。上述的处理器1301可以是通用处理器、数字信号处理器(DigitalSignalProcessor,DSP)、专用集成电路(ApplicationSpecific IntegratedCircuit,ASIC)、现成可编程门阵列(FieldProgrammableGateArray,FPGA)或者其他可编程逻辑器件、分立门或者晶体管逻辑器件、分立硬件组件。可以实现或者执行本申请实施例中的公开的各方法、步骤及逻辑框图。通用处理器可以是微处理器或者该处理器也可以是任何常规的处理器等。结合本申请实施例所公开的方法的步骤可以直接体现为硬件译码处理器执行完成,或者用译码处理器中的硬件及软件模块组合执行完成。软件模块可以位于随机存储器,闪存、只读存储器,可编程只读存储器或者电可擦写可编程存储器、寄存器等本领域成熟的存储介质中。该存储介质位于存储器1302,处理器1301读取存储器1302中的信息,结合其硬件完成上述方法的步骤。
可以理解的是,本申请实施例描述的这些实施例可以用硬件、软件、固件、中间件、微码或其组合来实现。对于硬件实现,处理单元可以实现在一个或多个专用集成电路(ApplicationSpecificIntegratedCircuits,ASIC)、数字信号处理器(DigitalSignalProcessing,DSP)、数字信号处理设备(DSPDevice,DSPD)、可编程逻辑设备(ProgrammableLogicDevice,PLD)、现场可编程门阵列(Field-ProgrammableGateArray,FPGA)、通用处理器、控制器、微控制器、微处理器、用于执行本申请所述功能的其它电子单元或其组合中。
对于软件实现,可通过执行本申请实施例所述功能的模块(例如过程、函数等)来实现本申请实施例所述的技术。软件代码可存储在存储器中并通过处理器1301执行。存储器可以在处理器1301中或在处理器1301外部实现。
在一个实施例中,语义处理包括语义提取和/或语义还原。
在一个实施例中,处理器1301具体用于利用本地语义知识库模型对语义通信信息进行语义处理。
在一个实施例中,本地语义知识库模型为第一终端通过联邦学习构建的语义知识库基础模型。
在一个实施例中,接收器,用于接收第二终端发送的语义通信信息。
在一个实施例中,接收器,还用于接收服务器下发的语义信息;
其中,语义信息为服务器通过全局模型对语义通信信息进行语义处理得到的,或者,语义信息为服务器通过全局模型对语义通信信息的近似特征进行语义处理得到的。
在一个实施例中,处理器1301还用于若语义处理成功,则得到语义信息。
在一个实施例中,处理器1301还用于根据语义信息,确定当前的语义通信过程是否完成;
发送器,还用于若当前的语义通信过程完成,则向第二终端发送语义通信完成信息。
在一个实施例中,处理器1301还用于执行与语义信息对应的操作。
在一个实施例中,处理器1301具体用于确定语义信息是否满足语义通信信息对应的通信意图或者通信目标;若语义信息满足通信意图或者通信目标,则确定当前的语义通信过程完成。
在一个实施例中,发送器还用于若当前的语义通信过程未完成,则向第二终端发送反馈信息,该反馈信息用于指示第二终端重新执行语义通信过程。
在一个实施例中,接收器还用于接收第二终端重新发送的语义通信信息,并对该第二终端重新发送的语义通信信息进行语义处理,以得到语义信息。
在一个实施例中,接收器还用于接收服务器发送的更新的模型参数,该更新的模型参数是在服务器对近似特征进行语义处理得到的所述语义信息之后,当接收到第二终端发送的语义通信完成信息时,根据近似特征对全局模型进行模型参数更新后得到的;
处理器还用于基于更新的模型参数,更新本地语义知识库模型。
在一个实施例中,接收器还用于接收服务器发送的语义处理失败反馈;
发送器还用于将语义处理失败反馈发送至第二终端。
图14为本申请实施例提供的一种服务器的结构示意图。图14所示的服务器1400包括:至少一个处理器1401、存储器1402、至少一个网络接口1404。服务器1400中的各个组件通过总线系统1405耦合在一起。可理解,总线系统1405用于实现这些组件之间的连接通信。总线系统1405除包括数据总线之外,还包括电源总线、控制总线和状态信号总线。但是为了清楚说明起见,在图14中将各种总线都标为总线系统1405。另外,本申请实施例中,上述服务器还包括收发器1406,收发器可以是多个元件,即包括发送器和接收器,提供用于在传输介质上与各种其他装置通信的单元。
可以理解,本申请实施例中的存储器1402可以是易失性存储器或非易失性存储器,或可包括易失性和非易失性存储器两者。其中,非易失性存储器可以是只读存储器(Read-OnlyMemory,ROM)、可编程只读存储器(ProgrammableROM,PROM)、可擦除可编程只读存储器(ErasablePROM,EPROM)、电可擦除可编程只读存储器(ElectricallyEPROM,EEPROM)或闪存。易失性存储器可以是随机存取存储器(RandomAccessMemory,RAM),其用作外部高速缓存。通过示例性但不是限制性说明,许多形式的RAM可用,例如静态随机存取存储器(StaticRAM,SRAM)、动态随机存取存储器(DynamicRAM,DRAM)、同步动态随机存取存储器(SynchronousDRAM,SDRAM)、双倍数据速率同步动态随机存取存储器(DoubleDataRate SDRAM,DDRSDRAM)、增强型同步动态随机存取存储器(Enhanced SDRAM,ESDRAM)、同步连接动态随机存取存储器(SynchlinkDRAM,SLDRAM)和直接内存总线随机存取存储器(DirectRambusRAM,DRRAM)。本申请实施例描述的系统和方法的存储器1402旨在包括但不限于这些和任意其它适合类型的存储器。
在一些实施方式中,存储器1402存储了如下的元素、可执行模块或者数据结构、或者他们的子集、或者他们的扩展集:操作系统14021。其中,操作系统14021,包含各种系统程序,例如框架层、核心库层、驱动层等,用于实现各种基础业务以及处理基于硬件的任务。
在本申请实施例中,通过调用存储器1402存储的程序或指令,使得接收器用于接收第一终端发送的语义通信信息;处理器1401用于对语义通信信息进行语义处理;发送器用于将语义处理得到的语义信息发送至所述第一终端。
上述本申请实施例揭示的部分或者全部方法还可以应用于处理器1401中,或者由处理器1401实现,或者由处理器1401与其他元件(例如收发机)配合实现。处理器1401可能是一种集成电路芯片,具有信号的处理能力。在实现过程中,上述方法的各步骤可以通过处理器1401中的硬件的集成逻辑电路或者软件形式的指令完成。上述的处理器1401可以是通用处理器、数字信号处理器(DigitalSignalProcessor,DSP)、专用集成电路(ApplicationSpecific IntegratedCircuit,ASIC)、现成可编程门阵列(FieldProgrammableGateArray,FPGA)或者其他可编程逻辑器件、分立门或者晶体管逻辑器件、分立硬件组件。可以实现或者执行本申请实施例中的公开的各方法、步骤及逻辑框图。通用处理器可以是微处理器或者该处理器也可以是任何常规的处理器等。结合本申请实施例所公开的方法的步骤可以直接体现为硬件译码处理器执行完成,或者用译码处理器中的硬件及软件模块组合执行完成。软件模块可以位于随机存储器,闪存、只读存储器,可编程只读存储器或者电可擦写可编程存储器、寄存器等本领域成熟的存储介质中。该存储介质位于存储器1402,处理器1401读取存储器1402中的信息,结合其硬件完成上述方法的步骤。
可以理解的是,本申请实施例描述的这些实施例可以用硬件、软件、固件、中间件、微码或其组合来实现。对于硬件实现,处理单元可以实现在一个或多个专用集成电路(ApplicationSpecificIntegratedCircuits,ASIC)、数字信号处理器(DigitalSignalProcessing,DSP)、数字信号处理设备(DSPDevice,DSPD)、可编程逻辑设备(ProgrammableLogicDevice,PLD)、现场可编程门阵列(Field-ProgrammableGateArray,FPGA)、通用处理器、控制器、
微控制器、微处理器、用于执行本申请所述功能的其它电子单元或其组合中。
对于软件实现,可通过执行本申请实施例所述功能的模块(例如过程、函数等)来实现本申请实施例所述的技术。软件代码可存储在存储器中并通过处理器1401执行。存储器可以在处理器1401中或在处理器1401外部实现。
在一个实施例中,处理器1401具体用于通过全局模型对语义通信信息进行语义处理;若未获得语义信息,则对语义通信信息进行信息拆解得到近似特征,并通过全局模型对近似特征进行语义处理。
在一个实施例中,接收器还用于接收第二终端发送的语义通信完成信息,该语义通信完成信息是由第一终端根据语义信息确定的,语义信息是服务器通过全局模型对近似特征进行语义处理得到的。
处理器1401还用于根据近似特征,对全局模型进行模型参数更新,得到更新的模型参数。
发送器还用于将更新的模型参数发送至第一终端。
在一个实施例中,发送器还用于若通过全局模型对近似特征进行语义处理的过程中未获得语义信息,则向第一终端发送语义处理失败反馈。
在一个实施例中,若接收器接收到第二终端发送的多次失败通知,处理器1401还用于删除拆解得到的近似特征。
在本申请一个实施例中,还提供了一种语义通信系统,可以结合图1所示,该语义通信系统包括第一终端、第二终端以及服务器。
第二终端用于向第一终端发送语义通信信息。
第一终端用于执行上述实施例中用于第一终端的语义通信方法的步骤。
服务器用于执行上述实施例中用于服务器的语义通信方法的步骤。
在一个实施例中,第二终端还用于根据待传输通信信息进行知识库搜索,以得到语义通信信息。
第二终端还用于若知识库搜索失败,则将待传输通信信息发送至服务器,以由服务器根据该待传输通信信息进行知识库搜索,以得到语义通信信息。
关于语义通信系统的实施方式以及有益效果,可以参见上文实施例的相关描述,在此不再赘述。
在本申请一个实施例中,还提供了一种计算机可读存储介质,其上存储有计算机程序,计算机程序被处理器执行时实现以下步骤:
对语义通信过程中的语义通信信息进行语义处理;
若语义处理失败,则将语义通信信息发送至服务器,以由服务器处理语义通信信息并获得语义通信信息对应的语义信息,并返回语义信息。
在一个实施例中,语义处理包括语义提取和/或语义还原。
在一个实施例中,计算机程序被处理器执行时具体实现以下步骤:
利用本地语义知识库模型对语义通信信息进行语义处理。
在一个实施例中,本地语义知识库模型为第一终端通过联邦学习构建的语义知识库基础模型。
在一个实施例中,计算机程序被处理器执行时还实现以下步骤:
接收第二终端发送的语义通信信息。
在一个实施例中,计算机程序被处理器执行时还实现以下步骤:
接收服务器下发的语义信息;
其中,语义信息为服务器通过全局模型对语义通信信息进行语义处理得到的,或者,语义信息为服务器通过全局模型对语义通信信息的近似特征进行语义处理得到的。
在一个实施例中,计算机程序被处理器执行时还实现以下步骤:
若语义处理成功,则得到语义信息。
在一个实施例中,计算机程序被处理器执行时还实现以下步骤:
根据所述语义信息,确定当前的语义通信过程是否完成;
若当前的语义通信过程完成,则向第二终端发送语义通信完成信息。
在一个实施例中,计算机程序被处理器执行时还实现以下步骤:
执行与语义信息对应的操作。
在一个实施例中,计算机程序被处理器执行时具体实现以下步骤:
确定语义信息是否满足语义通信信息对应的通信意图或者通信目标;
若语义信息满足通信意图或者通信目标,则确定当前的语义通信过程完成。
在一个实施例中,计算机程序被处理器执行时还实现以下步骤:
若当前的语义通信过程未完成,则向第二终端发送反馈信息,反馈信息用于指示第二终端重新执行语义通信过程。
在一个实施例中,计算机程序被处理器执行时还实现以下步骤:
接收第二终端重新发送的语义通信信息,并对第二终端重新发送的语义通信信息进行语义处理,以得到语义信息。
在一个实施例中,计算机程序被处理器执行时还实现以下步骤:
接收服务器发送的更新的模型参数,该更新模型参数是服务器对近似特征进行语义处理得到的语义信息之后,当接收到第二终端发送的语义通信完成信息时,根据近似特征对全局模型进行模型参数更新后得到的;
基于更新的模型参数,更新本地语义知识库模型。
在一个实施例中,计算机程序被处理器执行时还实现以下步骤:
接收服务器发送的语义处理失败反馈;
将语义处理失败反馈发送至第二终端。
在本申请一个实施例中,还提供了一种计算机可读存储介质,其上存储有计算机程序,计算机程序被处理器执行时实现以下步骤:
接收第一终端发送的语义通信信息,并对该语义通信信息进行语义处理;
将语义处理得到的语义信息发送至所述第一终端。
在一个实施例中,计算机程序被处理器执行时具体实现以下步骤:
通过全局模型对语义通信信息进行语义处理;
若未获得语义信息,则对语义通信信息进行信息拆解得到近似特征,并通过全局模型对近似特征进行语义处理。
在一个实施例中,计算机程序被处理器执行时还实现以下步骤:
接收第二终端发送的语义通信完成信息,其中语义通信完成信息是第一终端根据语义信息确定的,语义信息为服务器通过全局模型对近似特征进行语义处理得到的;
根据近似特征,对全局模型进行模型参数更新,得到更新的模型参数;
将更新的模型参数发送至所述第一终端。
在一个实施例中,计算机程序被处理器执行时还实现以下步骤:
若通过全局模型对近似特征进行语义处理的过程中未获得语义信息,则向第一终端发送语义处理失败反馈。
在一个实施例中,计算机程序被处理器执行时还实现以下步骤:
若接收到第二终端发送的多次失败通知,则删除拆解得到的近似特征。
图15是本申请实施例的芯片的示意性结构图。图15所示的芯片1500包括处理器1510,处理器1510可以从存储器中调用并运行计算机程序,以实现本申请实施例中的方法。
可选地,如图15所示,芯片1500还可以包括存储器1520。其中,处理器1510可以从存储器1520中调用并运行计算机程序,以实现本申请实施例中的方法。
其中,存储器1520可以是独立于处理器1510的一个单独的器件,也可以集成在处理器1510中。
在一些实施例中,该芯片1500还可以包括输入接口1530。其中,处理器1510可以控制该输入接口1530与其他设备或芯片进行通信,具体地,可以获取其他设备或芯片发送的信息或数据。
在一些实施例中,该芯片1500还可以包括输出接口1540。其中,处理器1510可以控制该输出接口1540与其他设备或芯片进行通信,具体地,可以向其他设备或芯片输出信息或数据。
在一些实施例中,该芯片1500可应用于本申请实施例中的网络功能或者基站,并且该芯片1500可以实现本申请实施例的各个方法中实现的相应流程,为了简洁,在此不再赘述。
应理解,本申请实施例提到的芯片1500还可以称为系统级芯片、系统芯片、芯片系统或片上系统芯片等。
本申请实施例还提供了一种包含指令的计算机程序产品,计算机程序被处理器执行时实现以下步骤:
对语义通信过程中的语义通信信息进行语义处理;
若语义处理失败,则将语义通信信息发送至服务器,以由服务器处理语义通信信息并获得语义通信信息对应的语义信息,并返回语义信息。
在一个实施例中,语义处理包括语义提取和/或语义还原。
在一个实施例中,计算机程序被处理器执行时具体实现以下步骤:
利用本地语义知识库模型对语义通信信息进行语义处理。
在一个实施例中,本地语义知识库模型为第一终端通过联邦学习构建的语义知识库基础模型。
在一个实施例中,计算机程序被处理器执行时还实现以下步骤:
接收第二终端发送的语义通信信息。
在一个实施例中,计算机程序被处理器执行时还实现以下步骤:
接收服务器下发的语义信息;
其中,语义信息为服务器通过全局模型对语义通信信息进行语义处理得到的,或者,语义信息为服务器通过全局模型对语义通信信息的近似特征进行语义处理得到的。
在一个实施例中,计算机程序被处理器执行时还实现以下步骤:
若语义处理成功,则得到语义信息。
在一个实施例中,计算机程序被处理器执行时还实现以下步骤:
根据语义信息,确定当前的语义通信过程是否完成;
若当前的语义通信过程完成,则向第二终端发送语义通信完成信息。
在一个实施例中,计算机程序被处理器执行时还实现以下步骤:
执行与语义信息对应的操作。
在一个实施例中,计算机程序被处理器执行时具体实现以下步骤:
确定语义信息是否满足语义通信信息对应的通信意图或者通信目标;
若语义信息满足通信意图或者通信目标,则确定当前的语义通信过程完成。
在一个实施例中,计算机程序被处理器执行时还实现以下步骤:
若当前的语义通信过程未完成,则向第二终端发送反馈信息,其中反馈信息用于指示第二终端重新执行语义通信过程。
在一个实施例中,计算机程序被处理器执行时还实现以下步骤:
接收第二终端重新发送的语义通信信息,并对第二终端重新发送的语义通信信息进行语义处理,以得到语义信息。
在一个实施例中,计算机程序被处理器执行时还实现以下步骤:
接收服务器发送的更新的模型参数,该更新的模型参数是服务器对近似特征进行语义处理得到的语义信息之后,当接收到第二终端发送的语义通信完成信息时,根据近似特征
对全局模型进行模型参数更新后得到的;
基于更新的模型参数,更新本地语义知识库模型。
在一个实施例中,计算机程序被处理器执行时还实现以下步骤:
接收服务器发送的语义处理失败反馈;
将语义处理失败反馈发送至第二终端。
本申请实施例还提供了一种包含指令的计算机程序产品,计算机程序被处理器执行时实现以下步骤:
接收第一终端发送的语义通信信息,并对语义通信信息进行语义处理;
将语义处理得到的语义信息发送至所述第一终端。
在一个实施例中,计算机程序被处理器执行时具体实现以下步骤:
根据全局模型对语义通信信息进行语义处理;
若未获得语义信息,则对语义通信信息进行信息拆解得到近似特征,并通过全局模型对近似特征进行语义处理。
在一个实施例中,计算机程序被处理器执行时还实现以下步骤:
接收第二终端发送的语义通信完成信息,其中语义通信完成信息是第一终端根据语义信息确定的,语义信息为服务器通过全局模型对近似特征进行语义处理得到的;
根据近似特征,对全局模型进行模型参数更新,得到更新的模型参数;
将更新的模型参数发送至第一终端。
在一个实施例中,计算机程序被处理器执行时还实现以下步骤:
若通过全局模型对近似特征进行语义处理的过程中未获得语义信息,则向第一终端发送语义处理失败反馈。
在一个实施例中,计算机程序被处理器执行时还实现以下步骤:
若接收到第二终端发送的多次失败通知,则删除拆解得到的近似特征。
本领域普通技术人员可以理解实现上述实施例方法中的全部或部分流程,是可以通过计算机程序来指令相关的硬件来完成,所述的计算机程序可存储于一非易失性计算机可读取存储介质中,该计算机程序在执行时,可包括如上述各方法的实施例的流程。其中,本申请所提供的各实施例中所使用的对存储器、存储、数据库或其它介质的任何引用,均可包括非易失性和易失性存储器中的至少一种。非易失性存储器可包括只读存储器(Read-Only Memory,ROM)、磁带、软盘、闪存或光存储器等。易失性存储器可包括随机存取存储器(Random Access Memory,RAM)或外部高速缓冲存储器。作为说明而非局限,RAM可以是多种形式,比如静态随机存取存储器(Static Random Access Memory,SRAM)或动态随机存取存储器(Dynamic Random Access Memory,DRAM)等。
以上实施例的各技术特征可以进行任意的组合,为使描述简洁,未对上述实施例中的各个技术特征所有可能的组合都进行描述,然而,只要这些技术特征的组合不存在矛盾,都应当认为是本说明书记载的范围。
以上所述实施例仅表达了本申请的几种实施方式,其描述较为具体和详细,但并不能因此而理解为对本申请专利范围的限制。应当指出的是,对于本领域的普通技术人员来说,在不脱离本申请构思的前提下,还可以做出若干变形和改进,这些都属于本申请的保护范围。因此,本申请专利的保护范围应以所附权利要求为准。
Claims (27)
- 一种语义通信方法,包括:对语义通信过程中的语义通信信息进行语义处理;若语义处理失败,则将所述语义通信信息发送至服务器,以由所述服务器处理所述语义通信信息并获得所述语义通信信息对应的语义信息,并返回所述语义信息。
- 根据权利要求1所述的方法,其中所述语义处理包括语义提取和/或语义还原。
- 根据权利要求1所述的方法,其中所述对语义通信过程中的语义通信信息进行语义处理,包括:利用本地语义知识库模型对所述语义通信信息进行语义处理。
- 根据权利要求3所述的方法,其中所述本地语义知识库模型为第一终端通过联邦学习构建的语义知识库基础模型。
- 根据权利要求3所述的方法,所述方法还包括:接收第二终端发送的所述语义通信信息。
- 根据权利要求5所述的方法,所述方法还包括:接收所述服务器下发的所述语义信息;其中,所述语义信息为所述服务器通过全局模型对所述语义通信信息进行语义处理得到的,或者,所述语义信息为所述服务器通过所述全局模型对所述语义通信信息的近似特征进行语义处理得到的。
- 根据权利要求5所述的方法,所述方法还包括:若语义处理成功,则得到所述语义信息。
- 根据权利要求6或7所述的方法,所述方法还包括:根据所述语义信息,确定当前的语义通信过程是否完成;若确定当前的语义通信过程完成,则向所述第二终端发送语义通信完成信息。
- 根据权利要求8所述的方法,其中所述向所述第二终端发送语义通信完成信息之后,所述方法还包括:执行与所述语义信息对应的操作。
- 根据权利要求8所述的方法,其中所述根据所述语义信息,确定当前的语义通信过程是否完成,包括:确定所述语义信息是否满足所述语义通信信息对应的通信意图或者通信目标;若确定所述语义信息满足所述通信意图或者所述通信目标,则确定当前的语义通信过程完成。
- 根据权利要求8所述的方法,所述方法还包括:若确定当前的语义通信过程未完成,则向所述第二终端发送反馈信息,所述反馈信息用于指示所述第二终端重新执行所述语义通信过程。
- 根据权利要求11所述的方法,所述方法还包括:接收所述第二终端重新发送的语义通信信息,并对所述第二终端重新发送的语义通信信息进行语义处理,以得到所述语义信息。
- 根据权利要求6所述的方法,所述方法还包括:接收所述服务器发送的更新的模型参数,所述更新的模型参数是所述服务器对所述近似特征进行语义处理得到的所述语义信息之后,在接收到所述第二终端发送的语义通信完成信息时,根据所述近似特征对所述全局模型进行模型参数更新后得到的;基于所述更新的模型参数,更新所述本地语义知识库模型。
- 根据权利要求1所述的方法,所述方法还包括:接收所述服务器发送的语义处理失败反馈;将所述语义处理失败反馈发送至第二终端。
- 一种语义通信方法,包括:接收第一终端发送的语义通信信息,并对所述语义通信信息进行语义处理;将语义处理得到的语义信息发送至所述第一终端。
- 根据权利要求15所述的方法,其中所述对所述语义通信信息进行语义处理,包括:根据全局模型对所述语义通信信息进行语义处理;若未获得语义信息,则对所述语义通信信息进行信息拆解得到近似特征,并通过所述全局模型对所述近似特征进行语义处理。
- 根据权利要求16所述的方法,所述方法还包括:接收第二终端发送的语义通信完成信息,其中所述语义通信完成信息是所述第一终端根据所述语义信息确定的,所述语义信息为服务器通过所述全局模型对所述近似特征进行语义处理得到的;根据所述近似特征,对所述全局模型进行模型参数更新,得到更新的模型参数;将所述更新的模型参数发送至所述第一终端。
- 根据权利要求16所述的方法,所述方法还包括:若通过所述全局模型对所述近似特征进行语义处理的过程中未获得语义信息,则向所述第一终端发送语义处理失败反馈。
- 根据权利要求18所述的方法,所述方法还包括:若接收到所述第二终端发送的多次失败通知,则删除拆解得到的近似特征。
- 一种语义通信装置,包括:语义处理模块,用于对语义通信过程中的语义通信信息进行语义处理;发送模块,用于若语义处理失败,则将所述语义通信信息发送至服务器,以由所述服务器处理所述语义通信信息并获得所述语义通信信息对应的语义信息,并返回所述语义信息。
- 一种语义通信装置,包括:接收模块,用于接收第一终端发送的语义通信信息,并对所述语义通信信息进行语义处理;发送模块,用于将语义处理得到的语义信息发送至所述第一终端。
- 一种终端,包括发送器和处理器;所述处理器,用于对语义通信过程中的语义通信信息进行语义处理;所述发送器,用于若语义处理失败,则将所述语义通信信息发送至服务器,以由所述服务器处理所述语义通信信息并获得所述语义通信信息对应的语义信息,并返回所述语义信息。
- 一种服务器,包括发送器、处理器和接收器:所述接收器,用于接收第一终端发送的语义通信信息;所述处理器,用于对所述语义通信信息进行语义处理;所述发送器,用于将语义处理得到的语义信息发送至所述第一终端。
- 一种语义通信系统,包括第一终端、第二终端以及服务器;所述第二终端,用于向所述第一终端发送语义通信信息;所述第一终端,用于执行如权利要求1-14中任一项所述的方法的步骤;所述服务器,用于执行如权利要求15-19中任一项所述的方法的步骤。
- 根据权利要求24所述的语义通信系统,其中所述第二终端还用于根据待传输通信信息进行知识库搜索,以得到所述语义通信信息;所述第二终端还用于若知识库搜索失败,则将所述待传输通信信息发送至所述服务器, 以由所述服务器根据所述待传输通信信息进行知识库搜索,以得到所述语义通信信息。
- 一种计算机可读存储介质,其上存储有计算机程序,其中所述计算机程序被处理器执行时实现权利要求1至19中任一项所述的方法的步骤。
- 一种计算机程序产品,包括计算机程序,其中所述计算机程序被处理器执行时实现权利要求1至19中任一项所述的方法的步骤。
Applications Claiming Priority (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410832278.9A CN121217746A (zh) | 2024-06-26 | 2024-06-26 | 语义通信方法、装置、终端、服务器、系统和存储介质 |
| CN202410832278.9 | 2024-06-26 |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| WO2026000751A1 true WO2026000751A1 (zh) | 2026-01-02 |
Family
ID=98110263
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| PCT/CN2024/129083 Pending WO2026000751A1 (zh) | 2024-06-26 | 2024-10-31 | 语义通信方法、装置、终端、服务器、系统和存储介质 |
Country Status (2)
| Country | Link |
|---|---|
| CN (1) | CN121217746A (zh) |
| WO (1) | WO2026000751A1 (zh) |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113127609A (zh) * | 2019-12-31 | 2021-07-16 | 华为技术有限公司 | 语音控制方法、装置、服务器、终端设备及存储介质 |
| CN116611441A (zh) * | 2022-02-09 | 2023-08-18 | 北京邮电大学 | 信息传输方法、装置、电子设备及存储介质 |
| CN116935864A (zh) * | 2023-07-28 | 2023-10-24 | 中国移动通信集团设计院有限公司 | 一种语义通信方法、装置和系统 |
| CN117311344A (zh) * | 2023-09-01 | 2023-12-29 | 北京佰才邦技术股份有限公司 | 基于语义通信自动驾驶方法、装置、设备及存储介质 |
| CN118200141A (zh) * | 2024-05-15 | 2024-06-14 | 鹏城实验室 | 语义通信资源优化方法、装置、设备及存储介质 |
-
2024
- 2024-06-26 CN CN202410832278.9A patent/CN121217746A/zh active Pending
- 2024-10-31 WO PCT/CN2024/129083 patent/WO2026000751A1/zh active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113127609A (zh) * | 2019-12-31 | 2021-07-16 | 华为技术有限公司 | 语音控制方法、装置、服务器、终端设备及存储介质 |
| CN116611441A (zh) * | 2022-02-09 | 2023-08-18 | 北京邮电大学 | 信息传输方法、装置、电子设备及存储介质 |
| CN116935864A (zh) * | 2023-07-28 | 2023-10-24 | 中国移动通信集团设计院有限公司 | 一种语义通信方法、装置和系统 |
| CN117311344A (zh) * | 2023-09-01 | 2023-12-29 | 北京佰才邦技术股份有限公司 | 基于语义通信自动驾驶方法、装置、设备及存储介质 |
| CN118200141A (zh) * | 2024-05-15 | 2024-06-14 | 鹏城实验室 | 语义通信资源优化方法、装置、设备及存储介质 |
Also Published As
| Publication number | Publication date |
|---|---|
| CN121217746A (zh) | 2025-12-26 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112800247B (zh) | 基于知识图谱共享的语义编/解码方法、设备和通信系统 | |
| CN107491560B (zh) | 页面图片显示方法、装置、存储介质、服务器及移动终端 | |
| CN112262556B (zh) | 模型文件的管理方法和终端设备 | |
| WO2024120355A1 (zh) | 关联信息展示方法、装置、设备、存储介质和程序产品 | |
| US20170277526A1 (en) | Software categorization method and electronic device | |
| CN108881227B (zh) | 远程白板系统的操作控制方法和装置、远程白板系统 | |
| CN117062185A (zh) | 联邦学习节点发现、注册方法及装置、存储介质及设备 | |
| WO2025260813A1 (zh) | 视频生成方法、电子设备及计算机可读存储介质 | |
| CN113722228A (zh) | 软件系统测试方法及装置 | |
| CN108933968A (zh) | 一种消息格式的转换方法、装置、存储介质及安卓终端 | |
| WO2024199409A1 (zh) | 一种数据处理方法及其装置 | |
| CN102932520B (zh) | 移动设备之间的信息交互方法和系统 | |
| WO2026000751A1 (zh) | 语义通信方法、装置、终端、服务器、系统和存储介质 | |
| CN116319644A (zh) | 会话菜单的处理方法、装置及电子设备 | |
| CN113515509A (zh) | HBase数据清理方法及装置 | |
| CN112363932A (zh) | 待测试业务对象的测试方法及装置、电子装置 | |
| WO2025065933A1 (zh) | 算力网络场景下网络控制方法及系统、介质和设备 | |
| CN120067285A (zh) | 一种信息查询方法、装置及系统 | |
| CN117082135A (zh) | 业务传输方法、装置、计算机设备和存储介质 | |
| CN105429981A (zh) | 游戏语音传输方法、终端、语音服务模块及游戏系统 | |
| WO2024093700A1 (zh) | 服务流转方法、装置及存储介质 | |
| CN117149777A (zh) | 一种数据查询方法、装置、设备及存储介质 | |
| CN107888951A (zh) | 片单管理方法、装置及服务器 | |
| CN115687340A (zh) | 业务查询方法、装置、设备及存储介质 | |
| CN118606035B (zh) | 数据库节点连接切换方法、装置、计算机设备及存储介质 |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| 121 | Ep: the epo has been informed by wipo that ep was designated in this application |
Ref document number: 24944319 Country of ref document: EP Kind code of ref document: A1 |