A semantic communication method based on a dynamic knowledge base, electronic equipment and medium

Abstract

The application discloses a kind of based on dynamic knowledge base's semantic communication method, electronic equipment, medium, the method includes: obtaining channel decoding vector;Channel decoding vector is mapped into uniform semantic representation space;Entity in knowledge base is mapped into uniform semantic representation space and carried out entity embedding, obtains the associated entity corresponding to this entity;Entity and corresponding associated entity are carried out relationship prediction, obtain entity-relation-entity triple, so that relevant entity forms associated knowledge graph;After relationship prediction, the knowledge graph extracted is carried out knowledge fusion with channel decoding vector, and is decoded by semantic decoder.

Description

Technical Field

[0001] This invention relates to the field of semantic communication technology, and in particular to a semantic communication method, electronic device, and medium based on a dynamic knowledge base. Background Technology

[0002] With the rapid development of deep learning and natural language processing, a new communication paradigm—semantic communication—is increasingly attracting widespread attention from researchers in the field. Compared to traditional communication methods, semantic communication emphasizes the precision of semantics in information transmission, rather than merely focusing on accurate symbolic transmission. Most existing semantic communication schemes are based on an implicit assumption: that the sender and receiver share a fixed prior knowledge system. Within this framework, communication systems are jointly trained using deep learning methods to achieve efficient transmission of semantic information.

[0003] In practical applications, the prior knowledge involved in semantic communication is often dynamic, and this knowledge can change significantly over time and in different contexts. This dynamic evolution of knowledge necessitates that communication systems possess greater flexibility and adaptability to maintain excellent communication performance across various situations. Furthermore, even at the same point in time, the sender and receiver may have different knowledge backgrounds, potentially leading to ambiguity during information interpretation. Therefore, how to achieve accurate semantic information transmission despite these knowledge differences is a crucial problem that urgently needs to be solved in the field of semantic communication. Summary of the Invention

[0004] In view of this, the present invention provides a semantic communication method, electronic device, and medium based on a dynamic knowledge base.

[0005] In a first aspect, embodiments of the present invention provide a semantic communication method based on a dynamic knowledge base, the method comprising:

[0006] Obtain the channel decoding vector;

[0007] Map the channel decoding vector to a unified semantic representation space;

[0008] Entities in the knowledge base are mapped to a unified semantic representation space for entity embedding to obtain the associated entities corresponding to the entity.

[0009] Relationship prediction is performed on entities and their corresponding related entities to obtain entity-relationship-entity triples, so that related entities form a knowledge graph of association;

[0010] After relation prediction, the extracted knowledge graph is fused with the channel decoding vector and then decoded using a semantic decoder.

[0011] Secondly, embodiments of the present invention provide an electronic device, including a memory and a processor, wherein the memory is coupled to the processor; wherein the memory is used to store program data, and the processor is used to execute the program data to implement the above-described semantic communication method based on a dynamic knowledge base.

[0012] Thirdly, embodiments of the present invention provide a computer-readable storage medium having a computer program stored thereon, wherein the program, when executed by a processor, implements the above-described semantic communication method based on a dynamic knowledge base.

[0013] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0014] 1. This invention utilizes knowledge graphs to enhance the performance of semantic communication. Compared to schemes that implicitly store semantic knowledge in neural network parameters, explicit knowledge representation helps to more accurately guide the extraction and reasoning of semantic information during the decoding process, thereby increasing the reliability of the system.

[0015] 2. This invention does not impose any additional assumptions or requirements on the structure of the sending end. This makes the invention compatible with conventional semantic communication systems without requiring special modifications to the sending end. This versatility ensures that the invention can be widely applied to various types of semantic communication systems, and is not limited to specific scenarios or configurations.

[0016] 3. The dynamic knowledge base mechanism introduced in this invention enables the receiving end to continuously update the content in the knowledge base through prediction. This adaptive mechanism gives the system a better ability to cope with diverse knowledge scenarios, thereby enhancing the system's robustness. Regardless of the changes in knowledge, the system can continuously maintain efficient semantic information processing capabilities. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 A schematic diagram illustrating a semantic communication method based on a dynamic knowledge base provided in an embodiment of the present invention;

[0019] Figure 2 A schematic diagram illustrating the training process of a unified semantic representation based on contrastive learning, provided in an embodiment of the present invention;

[0020] Figure 3 This is a schematic diagram of an electronic device provided in an embodiment of the present invention. Detailed Implementation

[0021] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0022] It should be noted that, unless otherwise specified, the features in the following embodiments and implementation methods can be combined with each other.

[0023] This invention provides a semantic communication method based on a dynamic knowledge base, aiming to intelligently update its knowledge base in real time according to the signals from the sending end. In this example, the receiving end maintains a knowledge base where knowledge is organized and stored in the form of (entity-relationship-entity) triples. These entities are mapped to a unified semantic space, which can eliminate the semantic gap that may be caused by heterogeneous data sources, thereby improving the accuracy and consistency of communication. During actual communication, the receiving end quickly retrieves and locates related entities based on the received signals, and then reveals the relationships between these entities through reasoning and prediction, thus providing powerful auxiliary information for the decoding process. At the same time, new knowledge is continuously added to the knowledge base. This adaptive mechanism further enhances the receiver's flexibility and adaptability in knowledge updates, enabling it to continuously improve its semantic information processing level in diverse knowledge scenarios.

[0024] like Figure 1 As shown, this invention proposes a semantic communication method based on a dynamic knowledge base, the method specifically including the following steps:

[0025] Step S1: Obtain the channel decoding vector y.

[0026] Specifically, the transmitting end transmits the semantically encoded signal to the receiving end through a physical channel, and then the receiving end decodes the semantically encoded signal through a channel decoder to obtain the channel decoding vector y.

[0027] Step S2: Map the channel decoding vector y to the unified semantic representation space.

[0028] Furthermore, the channel decoding vector y is mapped to a vector in a latent space, which is called the unified semantic representation space.

[0029] Specifically, in this example, the mapping process can be implemented using a neural network. Taking a linear mapping as an example, it can be represented as follows:

[0030] v y=W k y+b k

[0031] In the formula, v y W is the unified semantic representation vector obtained by mapping. k and b k These are the parameters of the neural network.

[0032] Step S3: Map the entities in the knowledge base to a unified semantic representation space for entity embedding, so as to achieve the fusion of knowledge from different sources and obtain the associated entities corresponding to the entity.

[0033] Specifically, suppose the associated entity to be extracted is ε i The corresponding embedding vector is From the unified semantic representation space, find the embedding vector v corresponding to the entity. y All distances less than the threshold λ That is, satisfy of in, Distance functions, such as cosine distance and Euclidean distance, are used to represent distances. The entities corresponding to these vectors are treated as related entities and extracted.

[0034] The knowledge in the knowledge base is organized into (entity-relation-entity) triples, where entities correspond to things and concepts in the physical world, and relations describe the connections between entities.

[0035] Step S4: Perform relationship prediction on the entity and its corresponding associated entity to obtain entity-relationship-entity triples, so that the related entities form an associated knowledge graph.

[0036] Specifically, for two entities ε i and ε j If the knowledge base already records the triple (ε) i ,r ij ,ε j ), then r ij This refers to the relationship between the two. Otherwise, the relationship is predicted using a pre-trained relation prediction neural network model, which can be viewed as a multi-classification task. Let the predicted result be... and triplet Add to the knowledge base. The training process of the relation prediction neural network model includes: based on the classification task, selecting several triples (ε... i ,r ij ,ε j ) as a sample, input ε i and ε j Train a neural network to predict r from a set of predefined relations. ij .

[0037] Step S5: After relation prediction, the extracted knowledge graph is fused with the channel decoding vector y, and then decoded by a semantic decoder.

[0038] Specifically, entities and relations are re-embedded as vectors and combined with the channel decoding vector y through concatenation or summation. The extracted knowledge graph can provide the semantic decoder with additional prior knowledge, thereby increasing the semantic decoder's ability to understand knowledge and improving the performance of semantic communication.

[0039] It should be noted that, at the same time, newly generated knowledge will be continuously added to the knowledge base. This adaptive mechanism further enhances the flexibility and adaptability of the receiving end in terms of knowledge updates.

[0040] The method also includes a training process, which specifically includes the following steps:

[0041] Step S100: Calculate a vector representation for all entities using a knowledge graph embedding algorithm, such as TransE, as the initial value for the entity embedding layer.

[0042] Step S200: Each sample contains a sentence s and several semantic triples (ε) in the representation. i ,r ij ,ε j The sentence is encoded as s at the sending end, transmitted through the channel, and then passed through the channel decoder at the receiving end to obtain the channel decoding vector y.

[0043] Step S300: Map the channel decoding vector y to the unified semantic representation vector v y Extract an associated entity ε from the sample. i Obtain it through embedding As positive samples, K unrelated entities (K is a parameter used to control the ratio of positive to negative samples) are randomly selected and embedded as negative samples.

[0044] Step S400: Calculate the infoNCE loss function, the expression of which is as follows:

[0045]

[0046] in, τ is the loss function, and τ is the temperature hyperparameter. The summation sign in the denominator indicates that the summation is calculated separately for all positive and negative samples.

[0047] Step S500: Perform backpropagation on the loss function to update the parameters of the corresponding layer in the neural network. Repeat the above process on all samples.

[0048] Accordingly, this application also provides an electronic device, comprising: one or more processors; a memory for storing one or more programs; and, when the one or more programs are executed by the one or more processors, causing the one or more processors to implement the semantic communication method based on a dynamic knowledge base as described above. Figure 3 The diagram shown is a hardware structure diagram of any device with data processing capabilities for the semantic communication method based on a dynamic knowledge base provided in an embodiment of the present invention, except... Figure 3 In addition to the processor, memory, and network interface shown, any data processing device in the embodiment may also include other hardware depending on the actual function of the data processing device, which will not be described in detail here.

[0049] Accordingly, this application also provides a computer-readable storage medium storing computer instructions thereon, which, when executed by a processor, implement the semantic communication method based on a dynamic knowledge base as described above. The computer-readable storage medium can be an internal storage unit of any data-processing device as described in any of the foregoing embodiments, such as a hard disk or memory. The computer-readable storage medium can also be an external storage device, such as a plug-in hard disk, smart media card (SMC), SD card, flash card, etc., equipped on the device. Furthermore, the computer-readable storage medium can include both internal storage units of any data-processing device and external storage devices. The computer-readable storage medium is used to store the computer program and other programs and data required by the data-processing device, and can also be used to temporarily store data that has been output or will be output.

[0050] Other embodiments of this application will readily occur to those skilled in the art upon consideration of the specification and practice of the disclosure herein. This application is intended to cover any variations, uses, or adaptations of this application that follow the general principles of this application and include common knowledge or customary techniques in the art not disclosed herein. The specification and embodiments are to be considered exemplary only.

[0051] It should be understood that this application is not limited to the precise structure described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope.

Claims

1. A semantic communication method based on a dynamic knowledge base, characterized in that, The method includes: Obtain the channel decoding vector; Map the channel decoding vector to a unified semantic representation space; Entities in the knowledge base are mapped to a unified semantic representation space for entity embedding to obtain the associated entities corresponding to the entity. Relationship prediction is performed on entities and their corresponding related entities to obtain entity-relation-entity triples, so that the related entities form a knowledge graph of association; After relation prediction, the extracted knowledge graph is fused with the channel decoding vector and then decoded using a semantic decoder. Mapping the channel decoding vector to the unified semantic representation space includes: A linear mapping is used to map the channel decoding vector to a unified semantic representation space, as shown in the following expression: ； In the formula, It is a unified semantic representation vector obtained through mapping. and These are the parameters of the neural network. Represents the channel decoding vector; This includes predicting relationships between entities and their corresponding associated entities to obtain entity-relationship-entity triples, thus forming a knowledge graph of related entities: The related entities to be extracted The corresponding embedding vector is denoted as ; Searching for satisfying conditions in the unified semantic representation space of ;in, Indicates the distance threshold. Represents a unified semantic representation vector. Represents the distance function; This includes predicting relationships between entities and their corresponding associated entities to obtain entity-relationship-entity triples, thus forming a knowledge graph of related entities: For two entities and If the knowledge base already records the triple ( , , ),but That is, the relationship between the two; otherwise, the relationship is predicted by a pre-trained relation prediction neural network model, and the relationship predicted by the relation prediction neural network model is denoted as... and the triplet ( , , Add to the knowledge base; The method also includes a training process, which specifically includes the following steps: A vector representation is computed for all entities using a knowledge graph embedding algorithm, which serves as the initial value for the entity embedding layer. Each sample contains a sentence and several semantic triples in representation ( , , ); for sentences The code is encoded, transmitted through the channel, and then decoded to obtain the channel decoding vector. ; Channel decoding vector Mapped to a unified semantic representation vector Extract an associated entity from the sample. Obtain it through embedding As positive samples; then randomly select... Unrelated entities are embedded as negative samples; Calculate the infoNCE loss function; Backpropagation is performed on the infoNCE loss function to update the parameters of the corresponding layer in the neural network.

2. The semantic communication method based on a dynamic knowledge base according to claim 1, characterized in that, The distance function is either cosine distance or Euclidean distance.

3. The semantic communication method based on a dynamic knowledge base according to claim 1, characterized in that, The training process of a relation prediction neural network model includes: Based on the classification task, select several triples ( , , As a sample, input and Training a relation prediction neural network to predict relation values ​​from a set of predefined relations. .

4. The semantic communication method based on a dynamic knowledge base according to claim 1, characterized in that, The knowledge fusion process between the extracted knowledge graph and the channel decoding vector includes: Entities and relations are re-embedded as vectors and then fused with the channel decoding vectors through concatenation or summation.

5. An electronic device, comprising a memory and a processor, characterized in that, The memory is coupled to the processor; wherein the memory is used to store program data, and the processor is used to execute the program data to implement the semantic communication method based on a dynamic knowledge base as described in any one of claims 1-4.

6. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the semantic communication method based on a dynamic knowledge base as described in any one of claims 1-4.

Images