A Method for Constructing a 6G Knowledge System for On-Demand Services Across All Domains and Scenarios

Abstract

This invention relates to a method for constructing a 6G knowledge system for on-demand services across all domains and scenarios. The method includes: constructing a 6G knowledge base based on 6G academic literature using a combination of top-down and bottom-up approaches; performing statistical analysis on the metadata of the literature in the 6G knowledge base to predict the development of 6G academic research; processing and training the 6G corpus data in the 6G knowledge base using natural language processing and deep learning methods to achieve knowledge extraction and generation; and annotating the metadata in the 6G knowledge base with knowledge during the statistical analysis, knowledge extraction, and generation processes. The 6G knowledge system is formed by three core layers: statistical analysis, knowledge extraction and generation, and knowledge annotation. Based on the 6G knowledge base and the 6G knowledge system, the method performs 6G-related tasks to achieve on-demand applications driven by 6G knowledge. This method constructs a 6G knowledge base and knowledge system, and on this basis, enables on-demand applications of knowledge, achieving efficient knowledge closure and knowledge-driven on-demand services.

Description

Technical Field

[0001] This invention belongs to the field of communication technology, specifically relating to a method for constructing a 6G knowledge system for on-demand services across all domains and scenarios. Background Technology

[0002] In recent years, discussions and research on 6G have been increasingly prevalent, but current 6G-related concepts lack consensus and urgently require a unified understanding and definition. Simultaneously, academia and industry lack a comprehensive understanding of the overall development of 6G, making it difficult for researchers to gain a clear understanding of research progress in related fields. Furthermore, 6G aims to achieve a paradigm shift from the Internet of Things to the Intelligent Internet of Things; the introduction of knowledge and the manifestation of intelligence are also key characteristics that distinguish 6G from 5G.

[0003] The current development of 6G in the academic field faces two main problems: First, due to the short research period, the exploration of 6G-related fields lacks an overall systematic construction and contextualization, which limits in-depth research on 6G theory and technology. Second, simply applying artificial intelligence algorithms to existing communication systems cannot realize the 6G on-demand service vision of "services as desired, networks adaptable to changing needs, and resources shared as desired." Deep integration of mobile communication and artificial intelligence technologies in a new knowledge-embedded architecture has become an urgent need. Summary of the Invention

[0004] To address the aforementioned problems in existing technologies, this invention provides a method for constructing a 6G knowledge system oriented towards on-demand services across all domains and scenarios. The technical problem to be solved by this invention is achieved through the following technical solution:

[0005] This invention provides a method for constructing a 6G knowledge system oriented towards on-demand services across all domains and scenarios, including the following steps:

[0006] S1. Combining expert knowledge and natural language processing technology, a 6G knowledge base is constructed based on 6G academic literature, integrating top-down and bottom-up approaches.

[0007] S2. Perform statistical analysis on the document metadata in the 6G knowledge base to predict the development of 6G academic research.

[0008] S3. Apply natural language processing technology and deep learning methods to process and train the 6G corpus data in the 6G knowledge base to achieve knowledge extraction and generation;

[0009] S4. During the statistical analysis, knowledge extraction and generation process, knowledge annotation is performed on the metadata in the 6G knowledge base; wherein, the 6G knowledge system is formed by the three-layer kernel of statistical analysis, knowledge extraction and generation and knowledge annotation.

[0010] S5. Based on the 6G knowledge base and the 6G knowledge system, perform 6G domain requirements tasks to realize 6G knowledge-driven on-demand applications.

[0011] In one embodiment of the present invention, step S1 includes:

[0012] By combining expert knowledge and natural language processing technology, ontology and pattern information are extracted from the structured data sources of the 6G academic literature and added to the 6G knowledge base, thus achieving top-down 6G knowledge base construction. At the same time, target data patterns are obtained from the 6G academic literature using annotation and induction methods, and information with higher confidence in the target data patterns is selected and included in the 6G knowledge base, thus achieving bottom-up 6G knowledge base construction.

[0013] In one embodiment of the present invention, the 6G knowledge base includes metadata fields and extended attributes, wherein,

[0014] The metadata fields include the ID, title, abstract, field, publication year, and DOI number of the 6G academic documents; the extended attributes include the number of articles and the article attribute category.

[0015] In one embodiment of the present invention, step S2 includes:

[0016] Statistical analysis is conducted on the distribution of papers, hot topics, and hot keywords in the 6G academic literature over time to predict the development of 6G academic research.

[0017] In one embodiment of the present invention, step S3 includes:

[0018] S31. Extract a set of keywords from the 6G knowledge base and construct a matching word list for the 6G field using expert knowledge;

[0019] S32. Using a hierarchical topic detection algorithm, calculate the relevance between the keywords and the topic, and find the topic and its corresponding keyword set;

[0020] S33. Use the matching word list to perform fuzzy matching on the topic to obtain topic words;

[0021] S34. Calculate the relevance between keywords using a hierarchical topic detection algorithm and establish a topic hierarchy structure;

[0022] S35. Calculate the relevance between the keywords and papers in the topic hierarchy and find the collection of papers corresponding to the keywords;

[0023] S36. A 6G knowledge tree is constructed from the topic and its corresponding keyword set, the topic hierarchical structure, and the paper set corresponding to the topic words, thereby realizing knowledge extraction and generation.

[0024] In one embodiment of the present invention, step S32 includes:

[0025] The frequency of co-occurrence of the keywords in the text corpus of the 6G knowledge base is calculated to extract related words and obtain the topic and its corresponding keyword set;

[0026] The relevance of the keywords is ranked using mutual information, and the five most relevant keywords in the topic are selected to name the topic. The formula for mutual information is:

[0027]

[0028] Where (X; Y) represent different topics X and Y, P(X,Y) represents the joint probability density function of topics X and Y, P(Z) represents the marginal probability density function of topic X, and P(Y) represents the marginal probability density function of topic Y.

[0029] In one embodiment of the present invention, step S33 includes:

[0030] The BM25 algorithm is used to score the similarity of the topics in the matching word list, and the topic with the highest score is selected as the target topic word. The formula for scoring the similarity of the topics is:

[0031]

[0032] Where Q represents the corpus set, D represents a corpus within Q, and IDF(q) i ) indicates the keyword q i The IDF value in Q, f(q) i D) represents the keyword q i In the TF value of corpus D, k1 represents word frequency saturation, b represents field length reduction, i.e., the ratio of the corpus length of D to that of Q, |D| represents the corpus length, and avgdl represents the average length of all corpus words in Q. The formulas for calculating IDF and TF are as follows:

[0033]

[0034] Where, n i This indicates the number of times the keyword appears in the corpus. This represents the total number of occurrences of all keywords in D;

[0035]

[0036] Wherein, 1+|{j:t i ∈d j}| represents the total number of instances of the keyword in Q.

[0037] In one embodiment of the present invention, step S4 includes:

[0038] During the statistical analysis, knowledge extraction, and generation processes, the scenarios, technologies, and indicators of each article in the 6G knowledge base are specifically labeled.

[0039] In one embodiment of the present invention, step S5 includes:

[0040] The target model for the required task is trained using the knowledge extraction and generation kernel in the 6G knowledge system and the 6G corpus data in the 6G knowledge base. The trained target model is then used to perform the required task in the 6G domain, realizing 6G knowledge-driven on-demand application.

[0041] In one embodiment of the present invention, the on-demand application includes the relevance of demand, the ambiguity of demand, and the scalability of demand.

[0042] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0043] This invention constructs a 6G knowledge base and knowledge system. The 6G knowledge base provides structured storage of all 6G academic literature to date, extending knowledge dimensions based on initial fields. The 6G knowledge system, using the 6G knowledge base as its carrier, is the crucial core for knowledge growth and application. It includes statistical analysis across the entire 6G domain, extraction and generation of 6G knowledge, and annotation of specific knowledge, enabling on-demand application of knowledge. The 6G knowledge base and 6G knowledge system are the first knowledge cluster built for the entire 6G domain, possessing multi-dimensional and multi-domain scalability, and supporting a combination of top-down and bottom-up knowledge generation, enabling efficient knowledge closure and knowledge-driven on-demand services. Attached Figure Description

[0044] Figure 1 A flowchart illustrating a method for constructing a 6G knowledge system for on-demand services across all domains and scenarios, provided by an embodiment of the present invention;

[0045] Figure 2 This invention provides a schematic diagram of a 6G knowledge base and knowledge system construction process.

[0046] Figure 3 Statistical charts of 5G and 6G document distribution provided for embodiments of the present invention;

[0047] Figure 4 This is a statistical chart of the distribution of the 6G Top-20 hot words provided in an embodiment of the present invention;

[0048] Figure 5 A flowchart of 6G knowledge tree construction provided in an embodiment of the present invention;

[0049] Figure 6 A visualization result of the 6G knowledge tree provided in an embodiment of the present invention;

[0050] Figure 7 A summary diagram of the ten typical 6G network scenarios and ten key technologies provided for embodiments of the present invention;

[0051] Figure 8 This is a schematic diagram of the 6G-BERT model architecture provided in an embodiment of the present invention;

[0052] Figure 9 This is a schematic diagram of 6G hotspot recommendation based on text generation, provided in an embodiment of the present invention. Detailed Implementation

[0053] The present invention will be further described in detail below with reference to specific embodiments, but the implementation of the present invention is not limited thereto.

[0054] Example 1

[0055] Faced with the rapid development of 6G technology, intelligent analysis of the overall picture of 6G research and the development trajectory of specific technologies has become a common need for many researchers and engineers. To achieve the mining of 6G knowledge and the embedding of native intelligence, this embodiment constructs a 6G knowledge base and knowledge system. The constructed 6G knowledge base and knowledge system not only include the storage and mining of data intelligence, but also aim to create an intelligent management platform for the entire lifecycle of academic knowledge. The construction of the 6G knowledge base and knowledge system is conducive to understanding the academic and industrial layout of 6G and future development hotspots, and enables on-demand knowledge application through in-depth knowledge mining.

[0056] Please see Figure 1 and Figure 2 , Figure 1 This is a flowchart illustrating a method for constructing a 6G knowledge system for on-demand services across all domains and scenarios, as provided in an embodiment of the present invention. Figure 2 This is a schematic diagram illustrating the construction process of a 6G knowledge base and knowledge system, provided as an embodiment of the present invention. The construction method includes the following steps:

[0057] S1. Combining expert knowledge and natural language processing technology, a 6G knowledge base is constructed based on 6G academic literature, integrating top-down and bottom-up approaches.

[0058] Based on structured 6G academic data, the 6G knowledge base combines expert knowledge and natural language processing technology to construct a knowledge system that integrates top-down and bottom-up approaches. Top-down construction refers to extracting ontology and schema information from structured data sources and adding them to the knowledge base; while bottom-up construction utilizes methods such as annotation and induction to obtain the required data schemas, selecting information with high confidence levels and adding it to the knowledge base.

[0059] In this embodiment, the 6G knowledge base provides structured storage for all 6G academic literature to date, expanding knowledge dimensions based on the initial fields. Specifically, this embodiment selects 1754 academic documents related to 6G, belonging to a total of 633 different fields. First, the selected 6G documents are preprocessed to form standardized text data, thereby constructing a 6G-oriented knowledge corpus. Currently, the 6G knowledge base contains initial data fields (i.e., metadata fields) for the 1754 articles, including specific IDs, titles, abstracts, fields, publication years, and DOI numbers. Simultaneously, the knowledge base supports vertical (number of articles) and horizontal (article attribute categories) expansion. Currently, multiple attribute dimensions, including scenarios, technologies, and KPIs, have been added, supporting on-demand knowledge expansion.

[0060] In summary, the 6G knowledge base includes metadata fields and extended attributes. The metadata fields include the ID, title, abstract, field, publication year, and DOI number of the 6G academic documents; the extended attributes include the number of articles and the article attribute category.

[0061] S2. Perform statistical analysis on the metadata of the documents in the 6G knowledge base to predict the development of 6G academic research.

[0062] Specifically, statistical analysis can be performed on the distribution of papers over time, hot topics, and hot keywords in the 6G academic literature to predict the development of 6G academic research. In other words, statistical analysis can be performed on the structured data such as metadata fields and extended attributes of 6G literature according to year, field, etc., to achieve an overall grasp of the current development trend of 6G academic research. Based on this, future development trends can be further predicted.

[0063] In one specific embodiment, after a combination of automated and manual cleaning and screening, the knowledge base yielded a total of 1754 6G-related documents and 36682 documents related to 5G. This embodiment provides a statistical analysis of the year distribution of these documents; please refer to [link to relevant documentation]. Figure 3 , Figure 3This invention provides a statistical chart of 5G and 6G literature distribution. The chart clearly shows that 6G academic research began to emerge in 2018, and accelerated from 2019 to 2021, with the number of articles reaching 1067 in 2021 alone. This rapid growth is expected to continue in the coming years. However, 6G academic research is still in its early stages, and data from recent years alone cannot provide a concrete picture of future 6G development. To better understand the future development trend of 6G, this embodiment uses the distribution of 5G academic literature as a reference and supporting evidence to compare, analyze, and predict the growth trend of 6G academic research. Figure 3 It can be observed that academic research on 5G began to emerge around 2010; it experienced rapid growth from 2013 to 2018; the growth rate slowed down after 2018, reaching its peak in 2020 with nearly 8,000 articles; the development cycle from emergence to peak is 8-10 years. After 2020, a downward trend has emerged, and it is expected that the number of published papers will decline from a gradual decrease to a sharp drop in the next 5 years.

[0064] In one specific embodiment, this embodiment extracts and analyzes high-frequency hot words appearing in the 6G knowledge base, and performs statistical analysis on the Top-20 hot words. The statistical results are as follows: Figure 4 As shown, Figure 4 This is a statistical chart showing the distribution of the 6GTop-20 hot words provided in an embodiment of the present invention. From... Figure 4 As can be seen, the top 20 trending terms, in order, are: edge computing, AI, security, IoT, terahertz, connected vehicles, MIMO, cellular networks, cloud computing, smart metasurfaces, virtualization, blockchain, satellite, non-orthogonal multiple access (NOMA), software-defined networking, beamforming, quantum communication, network slicing, and big data. These trending terms clearly correspond to typical 6G scenarios, enabling technologies, and development directions. Specifically, they represent future scenarios such as integrated air-space-ground-sea construction, connected vehicle-to-everything (V2X) transportation, intelligent cloud-edge-device collaboration, and distributed-centralized integrated deployment; in-depth exploration of emerging technologies such as terahertz, massive MIMO, and blockchain; and development concepts and new frameworks such as virtualization, software-defined networking, and on-demand services. This distribution of trending terms provides a comprehensive overview of the current development of 6G academia and industry and guides future directions.

[0065] S3. Apply natural language processing technology and deep learning methods to process and train the 6G corpus data in the 6G knowledge base to achieve knowledge extraction and generation.

[0066] Specifically, the implementation of 6G knowledge extraction and generation includes, but is not limited to: using a hierarchical topic detection algorithm to sort out the 6G context and generate a 6G knowledge tree; using the acquired 6G literature corpus to train a language model, resulting in a 6G-oriented language model—6G-BERT, which can be applied to various 6G-related downstream knowledge services in the future; and using a neural network model, this invention realizes 6G hotspot recommendation based on text generation, which can be used for fine-grained hotspot recommendation and association for various 6G sub-scenarios in the future.

[0067] For details, please see Figure 5 , Figure 5 This is a flowchart illustrating the construction process of a 6G knowledge tree according to an embodiment of the present invention. Taking the generation of a 6G knowledge tree as an example, the specific steps for knowledge extraction and generation include:

[0068] S31. Extract a set of keywords from the 6G knowledge base and construct a matching word list for the 6G field using expert knowledge.

[0069] Specifically, the 6G knowledge base stores standardized text data, from which a keyword set of 10,000 keywords is extracted. Furthermore, expert knowledge is used for screening and verification to construct a matching word list for the 6G domain. This matching word list is used for fuzzy matching with knowledge network nodes and contains 4,278 candidate words for 6G knowledge system topic nodes.

[0070] S32. Using a hierarchical topic detection algorithm, calculate the relevance between the keywords and the topic, and find the topic and its corresponding keyword set.

[0071] Specifically, this step is the first level of association, which mainly includes the extraction of related words and the determination of subject terms.

[0072] The extraction of related words is mainly determined by the co-occurrence characteristics of keywords. That is, the frequency of co-occurrence of the keywords in the text corpus of the 6G knowledge base is calculated to extract related words, thereby obtaining the topic and its corresponding keyword set. Furthermore, when the frequency of occurrence reaches a set threshold, these keywords are considered to be related, that is, they belong to the same topic.

[0073] To determine the keywords, the relevance of the keywords is ranked using mutual information I(X;Y). Then, the five most relevant keywords under a given topic are selected to name that topic. The formula for mutual information I(X;Y) is defined as follows:

[0074]

[0075] Where (X; Y) represent different topics X and Y, P(X,Y) represents the joint probability density function of topics X and Y, P(X) represents the marginal probability density function of topic X, and P(Y) represents the marginal probability density function of topic Y.

[0076] S33. Use the matching word list to perform fuzzy matching on the topic to obtain topic words.

[0077] Specifically, knowledge should be presented in a concise manner. Naming a topic with five words is somewhat redundant and not conducive to the intuitive expression of knowledge tree visualization. Therefore, this embodiment further utilizes the constructed 6G matching word list to perform fuzzy matching on the topic to obtain topic words.

[0078] Furthermore, this embodiment uses the BM25 algorithm to score the similarity of the topics in the matching word list, and selects the topic with the highest score as the target topic word. The formula for scoring the similarity of the topics is:

[0079]

[0080] Where Q represents the corpus set; D represents a corpus within Q; IDF(q) i ) indicates the keyword q i The IDF value in Q, i.e., q i In Q, the rarer the word, the higher its weight; therefore, its importance decreases as the number of words increases. i D) represents the keyword q i The TF value in corpus D, i.e., q i The importance of words in D increases with the number of words; k1 represents word frequency saturation, which is used to adjust the rate of change in saturation; b represents field length reduction, i.e., the ratio of the length of the corpus in D to that in Q; |D| represents the corpus length; avgdl represents the average length of all corpus words in Q. The formulas for calculating IDF and TF are as follows:

[0081]

[0082] Where, n i This indicates the number of times the keyword appears in the corpus. This represents the total number of occurrences of all keywords in D;

[0083]

[0084] Wherein, 1+|{j:t i ∈d j}| represents the total number of instances of the keyword in Q.

[0085] S34. Calculate the relevance between keywords using a hierarchical topic detection algorithm and establish a topic hierarchy structure.

[0086] Specifically, based on the first-level association, the obtained first-level topics are used as keywords to conduct the above association again. By repeating this process continuously, the knowledge tree hierarchy structure in the second-level association, namely the topic hierarchy structure, can be obtained.

[0087] S35. Calculate the relevance between the keywords and papers in the topic hierarchy and find the collection of papers corresponding to the keywords.

[0088] Specifically, after obtaining the topics and their hierarchical structure, a similarity match is performed between the topics and the corpus of papers, thereby recommending a set of papers related to the topics. In one specific embodiment, Elasticsearch methods can be used to perform similarity matching between topics and the corpus of papers.

[0089] S36. A 6G knowledge tree is constructed from the topic and its corresponding keyword set, the topic hierarchical structure, and the paper set corresponding to the topic words, thereby realizing knowledge extraction and generation.

[0090] Please see Figure 6 , Figure 6 A visualization of the 6G knowledge tree provided in an embodiment of the present invention. Figure 6 In the 6G knowledge tree, there are 6 levels and a total of 1453 topics, with different colors representing different node levels.

[0091] Overall, the 6G knowledge tree covers all aspects of 6G, especially corresponding to the ten typical 6G scenarios and ten key technologies. Specifically, for example... Figure 6 As shown in the enlarged section, the IoT nodes include Industrial IoT and Vehicle-to-Everything (V2X). Industrial IoT involves sensor fusion technology, equipment monitoring, and data security, while V2X covers sub-fields such as Vehicle-to-Vehicle (V2V), Vehicle-to-Cloud (V2C), and Vehicle-to-Infrastructure (V2I). The air-space-ground-sea nodes include space-based networks, air-based networks, and land-based networks. Space-based networks include various satellite communications, air-based networks cover high-altitude and near-Earth space, involving UAV communication and related technologies, and land-based networks cover cellular networks, Wi-Fi, and Device-to-Device (D2D) communication. In short, the 6G topic tree provides an overview of the entire 6G academic field while ensuring the accuracy and reliability of the knowledge. Furthermore, accessing these nodes provides recommendations for relevant papers, facilitating knowledge searches in specific areas.

[0092] This embodiment utilizes a hierarchical topic detection algorithm to generate a 6G knowledge tree from massive academic data, realizing a three-layer association between keywords, topics, and papers, and achieving the extraction of 6G full-domain knowledge structure.

[0093] S4. During the statistical analysis, knowledge extraction and generation process, knowledge annotation is performed on the metadata in the 6G knowledge base; wherein, the 6G knowledge system is formed by the three-layer kernel of statistical analysis, knowledge extraction and generation and knowledge annotation.

[0094] In addition to statistical analysis and knowledge extraction of metadata, this embodiment also performs rule-based knowledge annotation. Knowledge annotation refers to the structured labeling of text or other forms of data so that computer programs can better understand and utilize it. These labels are usually based on predefined categories and rules and can be used to identify important information such as entities, relationships, and events in the text. This embodiment mainly focuses on the targeted annotation of important attributes such as typical scenarios, enabling technologies, and key KPIs for each article in the 6G knowledge base. The annotated data can be applied to a wide range of on-demand knowledge services, currently mainly involving scene recognition, technology association, and KPI clustering analysis, and will be applied to more scientific research and application needs in the future.

[0095] Furthermore, a 6G knowledge system is formed by three core layers: statistical analysis, knowledge extraction and generation, and knowledge annotation. The 6G knowledge system, carried by a 6G knowledge base, is a crucial core for the knowledge base to achieve knowledge growth and application. It includes statistical analysis across the entire 6G domain, extraction and generation of 6G knowledge, and annotation of specific knowledge, enabling on-demand knowledge application.

[0096] The three core layers of the 6G knowledge system also interact with each other: knowledge annotation provides data samples for knowledge generation, driving related model training; knowledge generation provides knowledge dimensions to be annotated for knowledge annotation; knowledge generation provides statistically significant data dimensions for statistical analysis; and the results of statistical analysis can guide the extraction and generation of specific knowledge. These three core layers drive the 6G knowledge system to achieve knowledge distillation for specific 6G domains, with its output further fed back to the 6G database, realizing the cyclical operation of knowledge, i.e., a knowledge loop. Therefore, the 6G knowledge base has strong scalability; knowledge is not limited to predefined rules but rather achieves reasonable reasoning and discovery within a specific domain. Furthermore, utilizing AI technology, it is possible to present specific knowledge concepts, mine service needs, and recommend decisions. These results will continue to serve as input to the knowledge base, achieving true knowledge growth and a knowledge loop. Therefore, by constructing a 6G knowledge base and knowledge system, a closed-loop management system for the entire lifecycle of 6G academic knowledge, from extraction and mining to on-demand application, is achieved.

[0097] Please see Figure 7 , Figure 7 This embodiment of the invention provides an overview of ten typical 6G network scenarios and ten key technologies. After multiple rounds of rule cleaning and manual screening of the knowledge base, this embodiment summarizes and defines ten typical 6G scenarios, namely: multi-sensory immersive communication, three-dimensional multi-amplitude transportation, twin virtual interaction, fully functional and fully automated green industry, integrated communication and computing network, smart city and life, full-coverage cross-domain spatial communication, ubiquitous intelligent on-demand interaction, anti-interference secure and reliable network, and disaster-adaptive network. Simultaneously, this embodiment also summarizes ten core 6G technologies, namely: holographic communication, terahertz technology, visible light technology, digital twin, intelligent metasurface, knowledge graph, intent-driven, massive MIMO, blockchain, and big data. The typical 6G scenarios and core technologies are closely related, complementing each other and jointly depicting the blueprint and direction for the future development of 6G.

[0098] S5. Based on the 6G knowledge base and the 6G knowledge system, perform 6G domain requirements tasks to realize 6G knowledge-driven on-demand applications.

[0099] Specifically, the purpose of building a 6G knowledge system is to drive related knowledge applications to empower on-demand services across all domains and scenarios. Understandably, building a 6G knowledge base and system, based on specific fields, can automatically analyze corresponding scenarios, technologies, and service requirements, achieving truly on-demand services across all scenarios.

[0100] This embodiment first clarifies the logical connections between related concepts: "On-demand service" falls under the category of 6G services, such as communication, access, autonomous driving, and XR, aiming to provide customized services for users with different subjects, environments, and needs within 6G; "Knowledge-driven" is a new technology dedicated to making on-demand services more accurate, efficient, and low-consumption; "Knowledge application" refers to some programmatic applications that the 6G knowledge system can provide for knowledge-driven on-demand services, which can be realized by calling specific applications or methods; "Empowering 'Knowledge-Driven 6G On-Demand Service'" means that for known or unknown, standardized or customized service needs, based on the extraction of corresponding data patterns and entity attributes, knowledge analysis, generation, reasoning, and recommendation can be implemented to meet service needs, and the knowledge service management capabilities throughout the entire lifecycle can be achieved. Taking scenario cognition as an example, the implementation process of "knowledge-driven on-demand service" is illustrated as follows: First, the elements of the 6G full scenario are decomposed to construct an ontology structure that includes environment, subject, resources and services; second, knowledge graphs are used to represent and cognize the relationships between different entities and instances, and on this basis, a multi-dimensional and multi-granular rapid resource perception scheme is formed; finally, based on the perception and transfer of network intent, accurate demand identification and strategy generation are achieved in different scenarios.

[0101] To further illustrate the on-demand application capabilities of the 6G knowledge system, this embodiment summarizes three key aspects of on-demand: First, the relevance of the demand, meaning that the knowledge itself must contribute to the analysis and understanding of the demand, so that the services provided truly reflect the demand; second, the ambiguity of the demand, meaning that the demand does not have fixed granularity and characteristics, and knowledge can provide a certain degree of choice for it; and third, the scalability of the demand, meaning that knowledge services need to have the ability to handle new or unknown demands.

[0102] To achieve the integration of knowledge and on-demand services, the 6G knowledge base and knowledge system constructed in this embodiment provide rich knowledge applications. Specifically, the 6G knowledge base contains abundant corpus data. By training on these texts, numerous 6G-related tasks can be performed, such as scene recognition, technology association, and KPI clustering. This enables the application of the 6G knowledge system. Specifically, the knowledge extraction and generation kernel within the 6G knowledge system is used to train the target model for the required tasks on the 6G corpus data in the 6G knowledge base. The trained target model is then used to perform 6G-related tasks, realizing 6G knowledge-driven on-demand applications. Furthermore, 6G knowledge-driven on-demand applications include, but are not limited to: scene recognition, technology association, KPI clustering, knowledge graph generation, knowledge completion and reasoning, and on-demand knowledge recommendation.

[0103] This embodiment takes text-based 6G hotspot recommendations as an example, and uses theoretical tools such as deep learning and neural networks to realize knowledge application for 6G on-demand services.

[0104] Specifically, common text generation methods include language model-based text generation and deep learning-based text generation. The 6G knowledge system integrates and applies both methods. For the former, the 6G knowledge system trained a 6G-BERT language model using 6G of literature corpus, and its model architecture is as follows: Figure 8 As shown, Figure 8 This is a schematic diagram of the 6G-BERT model architecture provided in this embodiment of the invention. With the continuous expansion of knowledge and corpus, 6G-BERT is becoming increasingly complete and will be applied to a large number of downstream applications based on language models in the future. For the latter, considering the sequence order and contextual association of natural language, this embodiment achieves automatic generation of text sequences by training an LSTM model.

[0105] Text generation is essentially a multi-class classification problem. For character-level text generation, the classes to be distinguished are the types of characters in the text. This paper divides a 6G academic corpus into segments of 100 characters each, resulting in 2,022,120 samples. These samples are then used as the training and validation sets in a 4:1 ratio for model training. The corresponding parameter settings are shown in Table 1.

[0106] Table 2 Relevant Parameter Settings

[0107]

[0108]

[0109] In the prediction phase, the test text "future sixth-generation" is input into the model, and the output is processed through a Softmax layer to obtain a 148-dimensional probability result. Finally, Top-k sampling is used to sample the probabilities of the top five probabilities according to their distribution, and the corresponding characters are output. Some generated results are shown below. Figure 9 As shown, Figure 9 This is a schematic diagram of 6G hotspot recommendation based on text generation, provided in an embodiment of the present invention.

[0110] Overall, the 6G knowledge system and knowledge base model effectively generate a wealth of 6G-related information, encompassing 6G scenarios, technologies, attributes, and characteristics, such as artificial intelligence, network security, cellular networks, access networks, mobile edge computing, antenna technology, data transmission technology, non-orthogonal technologies, latency, capacity, and energy efficiency. This enables relevant and consistent hot topic recommendations based on the input text. Building upon this knowledge system application, introducing more artificial intelligence theories and technologies can improve the fluency and accuracy of text generation, further enhancing the capabilities of on-demand knowledge services. Qualitatively, the knowledge application for 6G hot topic recommendations aligns with the three core aspects of on-demand services. Its recommendation results are highly relevant to the requested content and encompass multiple dimensions, including scenarios, technologies, features, and attributes, satisfying both relevance and ambiguity of the requirements. Furthermore, its input can be any characters and sentences, addressing emerging or unknown needs and fully satisfying scalability. In addition, the 6G knowledge base and knowledge system provide various types of on-demand services and knowledge applications. In the future, with further improvements in demand analysis and granular segmentation, it is expected to achieve intelligent scenario connectivity and cross-domain services.

[0111] The 6G knowledge base and knowledge system constructed in this embodiment enable on-demand application of knowledge based on the extraction and summarization of knowledge across the entire 6G domain. The analysis of current 6G academic knowledge is beneficial for guiding the strategic layout and future development of various 6G fields. Simultaneously, the introduction of 6G knowledge enables comprehensive perception, decision-making inference, and dynamic adjustment of service demands and their management. For example, long-accumulated knowledge in the network and communication fields can be used to empower related network management and optimization. The proposed 6G knowledge base and knowledge system is the first knowledge cluster built for the entire 6G domain, which is of great significance for providing an overview of 6G and enabling on-demand services across all scenarios.

[0112] In summary, this embodiment constructs a 6G knowledge base and knowledge system, aiming to realize the 6G vision of "knowledge-driven" and "on-demand service." The 6G knowledge base provides structured storage of all current 6G academic literature, extending knowledge dimensions based on initial fields. The 6G knowledge system, using the 6G knowledge base as its carrier, is the crucial core for knowledge growth and application. It includes statistical analysis across the entire 6G domain, extraction and generation of 6G knowledge, and annotation of specific knowledge, enabling on-demand application of knowledge. The 6G knowledge base and knowledge system possess multi-dimensional and multi-domain scalability, supporting a combination of top-down and bottom-up knowledge generation, achieving efficient knowledge closure and knowledge-driven on-demand service. In the future, based on achieving fine-grained demand perception and knowledge expansion, a 6G on-demand service knowledge platform can be built, providing more intelligent on-demand knowledge services.

[0113] The above description, in conjunction with specific preferred embodiments, provides a further detailed explanation of the present invention. It should not be construed that the specific implementation of the present invention is limited to these descriptions. For those skilled in the art, various simple deductions or substitutions can be made without departing from the concept of the present invention, and all such modifications and substitutions should be considered within the scope of protection of the present invention.

Claims

1. A method for constructing a 6G knowledge system oriented towards on-demand services across all domains and scenarios, characterized in that, Including the following steps: S1. Combining expert knowledge and natural language processing technology, a 6G knowledge base is constructed based on 6G academic literature, integrating top-down and bottom-up approaches. S2. Perform statistical analysis on the document metadata in the 6G knowledge base to predict the development of 6G academic research. S3. Apply natural language processing technology and deep learning methods to process and train the 6G corpus data in the 6G knowledge base to achieve knowledge extraction and generation; S4. During the statistical analysis, knowledge extraction and generation process, knowledge annotation is performed on the metadata in the 6G knowledge base; wherein, the 6G knowledge system is formed by the three-layer kernel of statistical analysis, knowledge extraction and generation and knowledge annotation. S5. Based on the 6G knowledge base and the 6G knowledge system, perform 6G domain requirements tasks to realize 6G knowledge-driven on-demand applications; Step S3 includes: S31. Extract a set of keywords from the 6G knowledge base and construct a matching word list for the 6G field using expert knowledge; S32. Using a hierarchical topic detection algorithm, calculate the relevance between the keywords and the topic, and find the topic and its corresponding keyword set; S33. Use the matching word list to perform fuzzy matching on the topic to obtain topic words; S34. Calculate the relevance between keywords using a hierarchical topic detection algorithm and establish a topic hierarchy structure; S35. Calculate the relevance between the keywords and papers in the topic hierarchy and find the collection of papers corresponding to the keywords; S36. A 6G knowledge tree is constructed from the topic and its corresponding keyword set, the topic hierarchical structure, and the paper set corresponding to the topic words to realize knowledge extraction and generation. Step S32 includes: The frequency of co-occurrence of the keywords in the text corpus of the 6G knowledge base is calculated to extract related words and obtain the topic and its corresponding keyword set; The relevance of the keywords is ranked using mutual information, and the five most relevant keywords in the topic are selected to name the topic. The formula for mutual information is: Where (X; Y) represent different topics X and topics Y, P(X,Y) represents the joint probability density function of topics X and topics Y, P(X) represents the marginal probability density function of topic X, and P(Y) represents the marginal probability density function of topic Y; Step S33 includes: The BM25 algorithm is used to score the similarity of the topics in the matching word list, and the topic with the highest score is selected as the target topic word. The formula for scoring the similarity of the topics is: Where Q represents the corpus set, D represents a corpus within Q, and IDF(q) i ) indicates the keyword q i The IDF value in Q, f(q) i D) represents the keyword q i In the TF value of corpus D, k1 represents word frequency saturation, b represents field length reduction, i.e., the ratio of the corpus length of D to that of Q, |D| represents the corpus length, and avgdl represents the average length of all corpus words in Q. The formulas for calculating IDF and TF are as follows: Where, n i This indicates the number of times the keyword appears in the corpus. This represents the total number of occurrences of all keywords in D; Wherein, 1+|{j:t i ∈d j }| represents the total number of instances of the keyword in Q.

2. The method for constructing a 6G knowledge system for on-demand services across all domains and scenarios as described in claim 1, characterized in that, Step S1 includes: By combining expert knowledge and natural language processing technology, ontology and pattern information are extracted from the structured data sources of the 6G academic literature and added to the 6G knowledge base, thus achieving top-down 6G knowledge base construction. At the same time, target data patterns are obtained from the 6G academic literature using annotation and induction methods, and information with higher confidence in the target data patterns is selected and included in the 6G knowledge base, thus achieving bottom-up 6G knowledge base construction.

3. The method for constructing a 6G knowledge system oriented towards on-demand services across all domains and scenarios as described in claim 1, characterized in that, The 6G knowledge base includes metadata fields and extended attributes, among which, The metadata fields include the ID, title, abstract, field, publication year, and DOI number of the 6G academic documents; the extended attributes include the number of articles and the article attribute category.

4. The method for constructing a 6G knowledge system for on-demand services across all domains and scenarios as described in claim 1, characterized in that, Step S2 includes: Statistical analysis is conducted on the distribution of papers, hot topics, and hot keywords in the 6G academic literature over time to predict the development of 6G academic research.

5. The method for constructing a 6G knowledge system for on-demand services across all domains and scenarios as described in claim 1, characterized in that, Step S4 includes: During the statistical analysis, knowledge extraction, and generation processes, the scenarios, technologies, and indicators of each article in the 6G knowledge base are specifically labeled.

6. The method for constructing a 6G knowledge system for on-demand services across all domains and scenarios as described in claim 1, characterized in that, Step S5 includes: The target model for the required task is trained using the knowledge extraction and generation kernel in the 6G knowledge system and the 6G corpus data in the 6G knowledge base. The trained target model is then used to perform the required task in the 6G domain, realizing 6G knowledge-driven on-demand application.

7. The method for constructing a 6G knowledge system for on-demand services across all domains and scenarios as described in claim 1, characterized in that, The on-demand application includes the relevance of demand, the ambiguity of demand, and the scalability of demand.

Images