Subjects: Library Science，Information Science >> Information Science submitted time 2024-04-24
Abstract： Abstract: Objective This paper reviews the relevant progress of semantic novelty research at home and abroad, and summarizes the relevant technologies to provide reference for the subsequent research. Coverage Using retrieval methods such as Novelty of the literature , semantic novelty , literature novelty , semantic novelty and literature evaluation , after reading and sorting, this paper traced the source of representative relevant theoretical researches, and finally selected 70 literatures for review. Methods This paper summarizes the relevant researches on semantic novelty at home and abroad, and analyzes the development status and future trend of semantic novelty evaluation in scientific and technological literature by focusing on the definition of novelty, evaluation indexes of novelty and different evaluation methods. Results Semantic novelty evaluation has gradually attracted widespread attention from the academic community. There have been related studies on semantic content mining and evaluation, but no unified measurement index has yet been formed. Limitations The existing literature novelty is mostly evaluated from external features, while the number of literatures directly focusing on semantic novelty is small, which has limitations in supporting the review. Conclusions The evaluation of semantic novelty of scientific and technological literature is based on the novelty of content. Quantitative research has become the mainstream research method, but the calculation method of evaluation index needs to be clarified, and the development direction of novelty evaluation should be combined with qualitative and quantitative methods to achieve a scientific and reasonable comprehensive evaluation.
Subjects: Library Science，Information Science >> Information Science submitted time 2024-04-21
Abstract： Purpose/Significance As a new type of production factor, data factor, together with other factors, constitutes a total factor and provides an important driving force for global economic and social development. Data factor is a major theoretical innovation proposed for the first time in China, and it is important to systematically collect, sort out and analyze the representative results of data factor research in China, so as to build a knowledge framework system of data factor with Chinese characteristics, which will provide reference and inspiration for the theoretical research and practical exploration of data factor worldwide. Method/Process It mainly adopts network survey, literature research and content analysis method to summarize 56 representative literature contents, 14 policy reports and experts’ opinions on data element research in China, and comb and analyze them in four aspects, namely, connotation interpretation, value release, operation mechanism and governance system. Result/Conclusion China’s data elements research results are fruitful, the knowledge framework system is basically formed, and certain progress has been made in the realization of value and market cultivation, but there is still much room for improvement in the effective operation of data elements and the governance of data elements, and future research needs to focus on coordinating and promoting the construction of the governance system of data elements and promoting the combination with the application scenarios so as to play the multiplier effect.
Peer Review Status:
Awaiting Review
Subjects: Library Science，Information Science >> Information Science submitted time 2024-04-19
Abstract： While computer modeling and simulation are crucial for understanding scientometrics, their practical use in literature remains somewhat limited. In this study, we establish a joint coauthorship and citation network using preferential attachment. As papers get published, we update the coauthorship network based on each paper’s author list, representing the collaborative team behind it. This team is formed considering the number of collaborations each author has, and we introduce new authors at a fixed probability, expanding the coauthorship network. Simultaneously, as each paper cites a specific number of references, we add an equivalent number of citations to the citation network upon publication. The likelihood of a paper being cited depends on its existing citations, fitness value, and age. Then we calculate the journal impact factor and -index, using them as examples of scientific impact indicators. After thorough validation, we conduct case studies to analyze the impact of different parameters on the journal impact factor and -index. The findings reveal that increasing the reference number or reducing the paper’s lifetime significantly boosts the journal impact factor and average -index. On the other hand, enlarging the team size without introducing new authors or decreasing the probability of newcomers notably increases the average -index. In conclusion, it is evident that various parameters influence scientific impact indicators, and their interpretation can be manipulated by authors. Thus, exploring the impact of these parameters and continually refining scientific impact indicators are essential. The modeling and simulation method serves as a powerful tool in this ongoing process, and the model can be easily extended to include other scientific impact indicators and scenarios.
Peer Review Status:Awaiting Review
Subjects: Library Science，Information Science >> Information Science submitted time 2024-04-18
Abstract： Purpose/Significance The aim is to construct a mutation index to measure the innovation of academic papers, providing a reference for improving the evaluation system of academic paper innovation. Method/Process A large language model (LLM), Mistral-openorca, is utilized to extract the research questions and methods from papers. The mutation index is constructed based on two dimensions: the similarity of research questions and methods to existing studies, and the extent to which they stimulate subsequent research. Result/Conclusion Papers with high innovation scores generally have a higher citation frequency; among the top 10 ranked papers by innovation, seven introduced original methods or tools, whereas papers in the bottom 10 ranked by innovation had more mature research topics and methods at the time of publication, with lower levels of innovation; the correlation coefficient between the innovation score of the papers and their citation frequency is 0.530, and empirical research results further confirm the effectiveness of the mutation index proposed in this paper in measuring the innovation of academic papers.
Peer Review Status:Awaiting Review
Subjects: Library Science，Information Science >> Information Science submitted time 2024-04-18
Abstract： Purpose/Significance In the context of Artificial Intelligence + and Digital Education , many primary and secondary school students have already started using smartphones for information searching and leisure entertainment. However, there is still a lack of extensive research on the current situation of smartphone usage among primary school students and their digital literacy. This study conducts a survey on digital habits and digital literacy among primary school students in China, comparing samples from developed and underdeveloped regions, with the aim to provide a basis for the design of information technology curricula for primary school students and to offer suggestions for improving digital literacy in the new era. Research Design/Method This study collected data on the digital habits and digital literacy of upper primary school students using a questionnaire survey method, resulting in 701 valid responses. Based on this, methods such as descriptive statistics, difference testing, clustering, and text encoding were used to describe and compare the current situation of digital habits and literacy among primary school students. Conclusion/Findings Primary school students exhibit rich and good digital habits and have a certain level of digital literacy, but there is room for improvement in digital skills, digital awareness, and digital learning intentions. Significant differences are also noted between students from underdeveloped and developed regions. Innovation/Value This research fills the gap in surveys on digital habits and literacy among primary school students and proposes suggestions for cultivating correct digital habits and improving digital education policies.
Peer Review Status:Awaiting Review
Subjects: Library Science，Information Science >> Information Science submitted time 2024-04-18
Abstract： Abstract: Purpose/Significance Examining the effect of policy signals on corporate environmental information disclosure behavior is of great significance for the government to formulate policies, for the enhancement of corporate social responsibility, and for investors to make decisions. Method/Process An econometric model was constructed by constructing an indicator system, extracting policy signals, measuring media attention, adding control variables to identify research variables, and performing descriptive statistics; The sequential test regression coefficient method was used to test the effect of policy signaling on environmental information disclosure and the mediating effect of media attention; The sample is divided into state-owned enterprises and private enterprises to test the different strengths of the role of industrial policy signals on corporate environmental disclosure under different property rights. Results/Conclusions It is concluded that policy signals drive firms’ willingness to disclose environmental information, disclosure compliance and the extent of environmental governance disclosure, but they are not related to the extent of environmental liability disclosure; The mechanism of industrial policy’s effect on corporate disclosure is influenced by the media’s attention to the firm; The nature of firms’ property rights can have a heterogeneous impact on the driving effect of policy signals.
Peer Review Status:Awaiting Review
Subjects: Library Science，Information Science >> Information Science submitted time 2024-04-18
Abstract： Abstract: Purpose/Significance The independent knowledge system is an integration of disciplinary system, academic system, and discourse system. The construction and refinement of China’s independent knowledge system in intelligence science is an inevitable requirement for the development of Chinese characteristic intelligence science and the needs of the times. Method/Process Academic journals serve as an important vehicle for the independent knowledge system of intelligence science in China. In this study, we selected 3173 intelligence science literature published in academic core journals between 2014 and 2023 as the subjects of observation. We conducted a thorough analysis of the academic core journals that published intelligence science articles, the themes set by these journals, and the frequency of keywords. We employed bibliometric and content analysis methods and utilized CiteSpace software for visualization observation to present the internal structure and general framework of China’s independent knowledge system in intelligence science. Result/Conclusion In the era of digital intelligence, the innovation of autonomous knowledge in Chinese intelligence studies focuses on exploring technology and emerging technologies as the guiding intelligence tools, researching intelligence work in the context of big data, studying intelligence application systems in the era of blockchain technology and artificial intelligence, and constructing disciplines of intelligence studies in this context.
Peer Review Status:Awaiting Review
Subjects: Library Science，Information Science >> Information Science submitted time 2024-04-18
Abstract： Abstract: Purpose/Significance With the rapid development of social informatization, the spread scope and influence of online public opinions are constantly expanding. The analysis of public opinion intention is helpful for government departments to monitor and manage public opinion and create a healthy network environment. This study aims to sort out the development context and frontier trends in the field of public opinion intention recognition, provide direction for future research on public opinion intention recognition, and promote the development and application of the discipline. Method/Process This study discusses the current research status in the field of public opinion intention recognition, defines the concept and sources of public opinion intention recognition, and summarizes the technical methods, application fields, challenges and development directions of public opinion intention recognition. Result/Conclusion It is found that the current intention recognition of public opinion is mainly applied to the research of opinion leaders, early warning of public emergencies, public opinion governance and education. At the same time, the false information in online public opinion, the ambiguity of network language and the limited big data resources bring challenges to the recognition of public opinion intention. In the future, the process of public opinion intention recognition should focus on false information recognition, implicit information mining, algorithm and resource improvement.
Peer Review Status:Awaiting Review
Subjects: Library Science，Information Science >> Information Science submitted time 2024-04-17
Abstract： Purpose/Significance Medical device products are the products of clinical scientific research results transformation and knowledge innovation management. As one of the main competition tracks in the science and technology industry, the strength of medical science and technology is an important aspect to measure the country’s overall scientific research ability and environment. Knowledge innovation drives industrial development, product structure reflects the level of knowledge management, and knowledge management involving frontier science and technology is closely related to national strategic planning. From the perspective of product theme mining and evolution sharing, it is of great strategic significance for enterprises and countries to look into the current situation of domain technical knowledge management. Method/Process A comprehensive approach to technical topic identification and analysis is adopted in this study. Through the application of LDA algorithm in the field of machine learning, potential technical topics are identified and clustered from the medical device product data, and then the distribution characteristics and change rules of technical topics in different periods are analyzed. The future technology trend is predicted quantitatively by using hidden Markov model. Using this combination method, the topic distribution, evolution law and future trend of innovative medical devices from 2014 to 2023 are deeply studied. Result/Conclusion 38 technical topics were identified and sorted out using domestic approved innovative medical device product data. The results show the deep development of genetic testing and implanted electrical stimulator technology. Disease diagnosis, surgical navigation and biological mesh are the focus of future technological innovation and product competition. However, from the perspective of quantitative characteristics, China’s research and development strength is relatively weak, and the innovation ability is in a weak position, which needs to attract the attention of enterprise technology research and development institutions and the government.
Peer Review Status:Awaiting Review
Subjects: Library Science，Information Science >> Information Science submitted time 2024-04-16
Abstract： The most special kind of asset in the era of digital economy is data, data as a factor of production and enterprise assets, has become the new era of China’s production of core elements, is the core engine of the digital economy, data elements not only to promote the traditional factors of production revolutionary fusion and fission and thus produce a multiplier effect, but also to improve the allocation efficiency of the traditional factors to drive the economy to accelerate the development of innovation. However, data circulation and trading as a key link in the market-oriented configuration of data elements lacks effective practice, this paper starts from the perspective of information management, takes Shanghai Data Exchange as an example, and provides a sample reference for the realization path of data asset for enterprises from the three aspects of institutional safeguard, infrastructure construction, and platform operation mechanism, through which the study aims to improve the awareness of enterprises to pay more attention to data elements and the value excavation and use of data assets. Through this study, we aim to raise the awareness of enterprises to pay more attention to data elements, pay more attention to the value excavation and use of data assets, and devote ourselves to the effective realization of resource integration and optimization, so as to effectively promote the high-quality development of our economy.
Peer Review Status:Awaiting Review
Subjects: Library Science，Information Science >> Information Science submitted time 2024-04-14
Abstract： Purpose/Significance In order to address the issues of poor interactivity and low security in traditional electronic medical record big data resources, this paper proposes a blockchain-based electronic medical record big data sharing solution that is decentralized, tamper-proof, and encrypted. Method/Process Firstly, the application of current blockchain technology in the medical field is analyzed. Then, the demands for electronic medical records in the medical market are collected through questionnaire surveys and interviews. Based on this, a five-level blockchain-based electronic medical record big data resource sharing solution is proposed, which incorporates mechanisms such as asymmetric encryption technology, consensus mechanism, and smart contracts. Results/Conclusions This solution reshapes the process of sharing medical big data resources and further improves the application of blockchain in the secure storage and sharing of medical data. It can serve as a reference for relevant companies, hospitals, and researchers.
Peer Review Status:Awaiting Review
Subjects: Library Science，Information Science >> Library Science Subjects: Library Science，Information Science >> Information Science submitted time 2024-04-11
Abstract：
Abstract: Research objective With the wide application of generative artificial intelligence in cyberspace, the content security risks surrounding its industrial chain have become increasingly prominent. From the model layer to the application layer, it has continuously impacted the original content supply security and content generation ecology, and has further spread from the content domain to gradually affect many fields such as politics, law and ethics. The content security in cyberspace faces multiple challenges such as traditional risks, new risks and potential risks brought by generative artificial intelligence. Therefore, content governance for generative AI is particularly important. Research methods Based on the heterogeneous construction of people and technology in the actor network theory, at this stage, we should build a governance actor network oriented to content security risks, emphasizing the core decision maker position of government departments in pre-risk prevention, regular risk supervision, immediate disposal and effective reconstruction afterwards. And bring into play the governance effectiveness of heterogeneous actors such as algorithms and institutional texts. Research Conclusion In the specific governance path, an executable multi-actor network system with government departments as the leading force, text collaboration as the path support, risk identification as the behavior orientation, context construction as the reconstruction principle, and content security risks of generative artificial intelligence as the main line of governance is formed to jointly deal with the content security challenges brought by generative artificial intelligence.
Peer Review Status:Awaiting Review
Subjects: Library Science，Information Science >> Information Science submitted time 2024-04-09
Abstract： With the advent of the era of intelligent media, traditional information transmission is changing, and information transmission is becoming more complex, iterative and uncontrollable, which also brings more online public opinion pressure for universities. This paper builds a model of the influence of time distance and social distance on the re-transmission intention of online public opinions in universities in the age of intelligent media, and verifies the moderating effect of self-efficacy. The results show that: time distance and social distance positively affect the re-transmission intention of online public opinion in universities. Self-efficacy plays a significant role in regulating the relationship between time distance, social distance and the re-transmission intention of online public opinions in universities.
Peer Review Status:Awaiting Review
Subjects: Library Science，Information Science >> Information Science submitted time 2024-04-03 Cooperative journals: 《农业图书情报学报》
Abstract： Purpose/Significance With the popularity and development of social media, Weibo has become an important information transmission platform. Understanding the law of emergency information transmission based on the characteristics of Weibo users is of great significance for grasping the development trend of events, evaluating the influence of information, and formulating effective response strategies. Methods/Process Selecting typical cases of four types of emergencies (natural disasters, accidents, public health events, and social security events) as the research objects, this paper crawled the user data of the information transmission on the Weibo platform and analyzed the distributions of user gender, user geographical distribution, user attention and fan number and user type of the information transmission for these events, in order to reveal the differences and similarities of the user characteristics and transmission laws for the four types of emergencies. Results/Conclusions As for the similarities of the user characteristics and transmission laws, first, the audiences of the four types of emergencies have obvious regional characteristics, and the users in economically developed provinces generally have a higher attention rate. Second, the transmission users of all types of emergencies are generally concentrated on active users and grassroot users with strong and weak propagation power respectively, accompanied by a small number of authoritative users with strong influence. However, there are differences in the broadness and gender distribution of the audiences for different types of emergencies. First, natural disasters have a wider spread of the audiences and relatively higher attentiveness; accidents and public health events tend to be local events with relatively low attentiveness, whose audiences may be more concentrated in and around the affected areas; the audience spread of social security events is scattered depending on the nature of the event, the scope of influence as well as the communication channels. Second, female users pay more attention to natural disasters, public health events and social security events, while male users pay more attention to accidents, which can be attributed to the different emotional orientations and psychological characteristics associated with users of different genders. These results provide insights for the formulation of targeted guidance strategies for information dissemination. In future studies, we will collect data on other social media platforms, obtain more information on user characteristics through different channels, and introduce more in-depth analysis methods and indicators to comprehensively reveal the dynamic mechanisms of emergency information dissemination, thus improving the accuracy and effectiveness of the research.
Subjects: Library Science，Information Science >> Information Science submitted time 2024-04-03 Cooperative journals: 《农业图书情报学报》
Abstract： Purpose/Significance With the development of the Internet and the popularization of smart devices, the scale of social media users around the world is increasing, and social media has become the center of people’s communication, but it also produces a series of negative problems. Digital stress refers to the painful subjective experience of users caused by different stressors in the process of heavy use of social media. At present, most studies on social media focus on users’ social media behaviors. In recent years, digital stress has gradually attracted the attention of researchers. However, most studies discuss the impact of single variables on results from the perspective of linear regression, and few studies focus on the complex mechanism behind digital stress. This study aims to explore the influencing factors and mechanism of social media users’ digital stress in order to further understand the path of digital stress and provide reference for relieving users’ digital stress. Method/Process This paper adopted a mixed research method combining grounded theory and qualitative comparative analysis. First, 21 users were selected as interview objects, interview data were coded, and theoretical models were constructed. Second, a questionnaire was designed according to the theoretical model and 262 questionnaires were collected. The fuzzy set qualitative comparative analysis (fsQCA) method was used to analyze the questionnaire data and explore the formation path of social media users’ digital stress. Results/Conclusions There were 9 influencing factors of social media users’ digital stress: overuse, technology intrusion, information overload, communication load, fear of missing out, approval anxiety, social expectation, digital coping skills and self-control. They were classified into four dimensions: technology, social, environment and personal characteristics. There are six paths that trigger digital stress, among which overuse and fear of missing out are the key factors leading to high-level digital stress, technology intrusion and information overload are the important conditions for digital stress, and digital coping skills and self-control have no significant effects. At the same time, this paper has some limitations. First, the end point and result variable of this paper is the digital pressure of social media users, focusing on the factors that affect the generation of digital pressure and how the combination of factors and variables ultimately leads to the generation of results. Less attention is paid to user behavior. Although some attention is paid in the interview, antecedent variables are emphasized in the analysis. Future studies can take users’ social media usage behavior into consideration to explore the whole process formation mechanism of influencing factors-digital pressure-user behavior. In addition, most of the research samples are young users of social media, and users of other age groups are less involved. In the future, the research samples should be further expanded to improve the universality of the research conclusions.
Subjects: Library Science，Information Science >> Information Science submitted time 2024-04-03 Cooperative journals: 《农业图书情报学报》
Abstract： Purpose/Significance Today, the methods of global technology cooperation have become diversified, while the patterns of knowledge and resource flows have become networked. The way knowledge flows between science and technology and promotes innovation is a hot topic in informatics research. Under such circumstances, the transfer and flow of technical knowledge has become increasingly important for the creation of new technologies and the generation of new knowledge. It can provide potential solutions for solving technical problems and generating technological competitive advantages. This study reveals the structural characteristics of the knowledge flow network between science and technology from a micro perspective, and provides new ideas for discovering new knowledge flow and communication channels, understanding the links between science and technology, and exploring innovation breakthroughs. Method/Process This paper adopts a multi-perspective approach and exames technical knowledge citation, scientific knowledge citation, and the intersection of scientific knowledge. Specifically, the study focuses on patent-patent citation, patent-paper citation, and paper-discipline co-occurrence, and uses a multi-layer network method to develop a science-technology (S&T) knowledge flow micro-network analysis framework. Through this framework, knowledge flow patterns between science and technology can be comprehensively explored. An empirical study was conducted in the field of genetically modified corn to investigate the knowledge flow patterns between science and technology. Using a multi-layered network model, this study aims to explore the relationship between science and technology from three different perspectives. Results/Conclusions The results of the study suggest that technology nodes that absorb multidisciplinary knowledge are more likely to engage in knowledge diffusion or absorption. Additionally, knowledge flow is more likely to occur between the technology nodes corresponding to two interdisciplinary subjects. The analytical framework proposed in our study can reveal the characteristics of knowledge flow from a micro perspective, and can identify two types of knowledge flow structures in the empirical case of genetically modified corn, which extends and enriches existing research on knowledge flow. In the future, the S&T knowledge flow research framework proposed in this study has the potential to be further extended and widely applied by incorporating different types of knowledge subjects, enriching knowledge attribute factors, and expanding knowledge flow paths. For instance, the research framework can be extended to broader fields to explore the relationship between knowledge flow patterns and the characteristics of these fields. It can also be used to explore knowledge flow patterns and potential technological breakthroughs at different flow levels and innovation themes. Such research can serve as a basis for innovation policy and provide theoretical support for the development of strategies and technology research directions.
Subjects: Library Science，Information Science >> Information Science submitted time 2024-04-01
Abstract： Purpose/Significance With the rapid ascent of large language models, AIGC have become ubiquitous in our daily lives. In order to mitigate potential misuse of AIGC, and to address issues such as the proliferation of false information, academic misconduct, and deceptive commentary, it is imperative to consolidate and forecast advancements in natural language processing techniques aimed at empowering AIGC discernment.［Method/Process］Firstly, it is essential to clarify that AIGC recognition constitutes a binary classification problem, with the aim of discerning whether a given piece of content is generated by artificial intelligence. Subsequently, employing a systematic review methodology, we have delineated the principal research outcomes in the domain of AIGC recognition.［Result/Conclusion］The research identifies the critical significance of comprehensive and high-quality datasets in constructing classifiers for AIGC recognition. Simultaneously, it explores the limitations and developmental objectives of currently popular datasets, as well as potential datasets. Additionally, the paper analyzes paradigms of various classifiers, presents challenges across multiple domains such as multi-domain recognition tasks, cross-lingual recognition tasks, and data ambiguity issues. Finally, it summarizes the prospective development pathways for the future of AIGC recognition. This study aims to provide relevant researchers with a clear introduction and constructive suggestions for constructing more stable and efficient classifiers.
Peer Review Status:Awaiting Review
Subjects: Library Science，Information Science >> Information Science submitted time 2024-03-11
Abstract： Purpose/Significance The establishment of public data authorization operation policies are crucial prerequisites for regulating data utilization and fostering data factorization reform. Quantitative evaluations of these policies can significantly enhance their quality. Method/Process This study analyzes the content of 274 relevant policies with the help of NVivo software, and constructs an evaluation index system using the PMC index model to quantitatively evaluate 24 specialized policies. Result/Conclusion The research shows that among the 24 specialized policies, 9 are good, 10 are passing, 5 are failing, and the policies as a whole are at the passing level. The policies’ contents have weak rewards and punishments, neglect of talent reserves, insufficient financial support, lack of publicity and guidance, and ambiguous work objectives in terms of management mechanisms, institutional safeguards, operational construction, and operational assessment. In addition, the policies show geographical and chronological differences, and the early policies pay more attention to the construction of operation services. Based on the results of the analysis, improvement suggestions are made in three aspects: policy promulgation, policy content and platform construction.
Peer Review Status:Awaiting Review
Subjects: Other Disciplines >> Synthetic discipline Subjects: Library Science，Information Science >> Information Science Subjects: Management Science >> Business Management of Department submitted time 2024-03-04
Abstract： Purpose/significance Tunisia is famous for its handicrafts. A thematic study on how Tunisia develops its handicrafts could provide useful lessons for our country. Method/process This paper summarizes Tunisia’s experience in the development of handicrafts, based mainly on foreign research data and using literature review methods. Result/conclusion Tunisia has five main successful experiences in the development of handicrafts. First, the Tunisian government and all sectors of society have attached great importance to handicrafts and have jointly created a favorable environment for their development. Second, all parts of Tunisia have adapted their crafts to local conditions and have always maintained their crafts, which have been handed down from generation to generation. Third, the Tunisian government and all sectors of society have worked together to establish a relatively mature technical and vocational education and training system. Fourth, Tunisia has vigorously promoted the export of handicrafts to generate income. Fifth, Tunisian handicrafts have found their own way between tradition and modernity.
Peer Review Status:Awaiting Review
Subjects: Library Science，Information Science >> Information Science submitted time 2024-03-01
Abstract： Objective This paper aims to improve the accuracy of automatic extraction of key technical words and corresponding function words from patent. Methods ChatGPT was used as the Teacher-model, and ChatGLM3 was used as the Student-model. Through knowledge distillation method, the training data generated by ChatGPT was used to fine-tune ChatGLM3, and multiple technical word extraction models and a function word extraction model were obtained. The technical words are extracted from the abstract, the first claim and the technical function paragraph, respectively, by using multiple technical word extraction models, and the function words are extracted from the technical function paragraph by using the function words extraction model. Results Compared with ChatGPT, the fine-tuned multiple technical word extraction models and function word extraction model show higher accuracy and lower recall rate, when extracting technical words and function words. The ChatGLM3 fine-tuning model of the first claim has the highest accuracy and F1 values of 0.734 and 0.724 respectively. Moreover, The accuracy of the function words extracted by the function word extraction model is 0.649, which is higher than the accuracy of the function words labeled by the commercial tool, which is is 0.53. Limitations The technical field and patent language of this research are single, the amount of patent verification data is small, and the data cleaning rules expect to be further optimized. Conclusions This research scheme improves the efficiency accuracy of automatic extraction of large language model through knowledge distillation operation. At the same time, this study can support the mining of cutting-edge innovative and hot technologies from patent texts, and support higher quality intelligent patent analysis.
Hits
YES
Hits
Downloads
Comment