Special Topics

123 Discipline Construction of Information Science and the Future of Information Services in the Artificial Intelligence Era Su Xinning DOI: 10.3772/j.issn.1000-0135.2025.02.001 Artificial intelligence (AI) has triggered a transformation in scientific research, and national security and development strategies have introduced new requirements for information science and services. This study provides a critical reflection on the evolution of information science in the era of AI. It emphasizes the need for an information service infrastructure to integrate seamlessly with AI. As AI enhances information capabilities, the focus of information services must be repositioned. And the development of a comprehensive information system is proposed to address this issue. Finally, this study demonstrates the disciplinary structure and construction of information science in the era of AI. It concludes that the conditions are favorable for establishing information science as a first-level discipline in the context of AI and presents an initiative for this development. 2025 Vol. 44 (2): 123-131 [Abstract] ( 12 ) HTML (72 KB)  PDF (661 KB)  ( 45 )

Intelligence Theories and Methods

132 AI for Science Revolution: The “Platform Research” Paradigm from the Perspective of Knowledge Services for Innovation Mao Jin, Zhou Fanqian, Wang Zhuohao DOI: 10.3772/j.issn.1000-0135.2025.02.002 Based on the scientific and technological intelligence knowledge service perspective, this paper outlines the connotation and framework of the “Platform Research” paradigm promoted by AI for Science (AI4S). According to Kuhn's paradigm theory, we discuss the inevitability of AI4S to promote innovation of the research paradigm; summarize the scientific research process using Bacon's inductive method as a framework; and elucidate the reciprocal and co-evolutionary relationship between knowledge service for innovation and the “Platform Research” paradigm as a theoretical guide. To support research and innovation activities, the main contents of the paradigm include scientific data management from the perspective of knowledge representation, universal knowledge base from the perspective of knowledge fusion, scientific hypothesis prediction from the perspective of knowledge inference, scientific experiment execution from the perspective of knowledge discovery, and industrial empowerment from the perspective of knowledge application. A framework of the paradigm from the perspective of knowledge service is proposed, which clarifies the core research content of innovative knowledge services in various key stages, and aims to become a growth point in the field of scientific and technical information research. This article provides a reference for China to seize the opportunity for research paradigm innovation. 2025 Vol. 44 (2): 132-142 [Abstract] ( 8 ) HTML (118 KB)  PDF (2186 KB)  ( 42 )

143 Measurement of Paper Innovation Quality by Integrating Novelty and Academic Influence Characteristics Li Jing, Qiu Xinpeng DOI: 10.3772/j.issn.1000-0135.2025.02.003 With an emphasis on the high-quality development of science and technology and breaking the “five principles” of science and technology evaluation, accurately measuring the innovation quality of academic papers will provide a methodological basis for improving the evaluation system of scientific and technological achievements. Based on the AHP (analytic hierarchy process) - entropy weight - TOPSIS (technique for order preference by similarity to an ideal solution) method, this study proposes a measurement framework for paper innovation quality considering the characteristics of paper novelty and academic influence. The novelty score of the papers is obtained by calculating the frequency of question words, method words, and problem-method combination words based on the principle of word frequency. The academic influence score is obtained by calculating the number and quality of citations. The empirical results show that the measurement method of paper innovation quality constructed in this study can, to some extent, address the limitations of existing paper evaluation methods, such as the use of a single evaluation index, low discrimination of results, and emphasis on the number of citations but neglecting the quality of citations, and can effectively identify papers with high innovation quality from journals with weak impact factors. 2025 Vol. 44 (2): 143-156 [Abstract] ( 9 ) HTML (221 KB)  PDF (1866 KB)  ( 36 )

157 Early Identification Strategies for Disruptive Technologies through the Lens of Technology Life Cycle Hou Yanhui, Chen Rong, Wang Jiakun DOI: 10.3772/j.issn.1000-0135.2025.02.004 This study proposes a method for the early identification of disruptive technologies that considers the technology life cycle stages and feature heterogeneity. It addresses the problem of ignoring technology evolution features in the current disruptive technology identification process. First, Sentence-BERT (sentence bidirectional encoder representation from transformers) is used to vectorize patent abstracts. Second, a filtering identification system is constructed: the first layer employs the local outlier factor with constraint integration (LOCI) anomaly detection algorithm to identify and classify outlier patents; the second layer uses an S-curve life cycle identification to filter patents in the maturity stage; the third layer measures the innovativeness of patents in the budding stage; and the fourth layer evaluates the disruptive nature of the patent text and technology reporting data in the growth stage to finalize the filtering process. Finally, the field of quantum information technology is used as a case study to illustrate this method. The study identifies three disruptive themes in the germination and growth stages, which align with the authoritative reports, verifying the feasibility and effectiveness of the proposed method. 2025 Vol. 44 (2): 157-170 [Abstract] ( 13 ) HTML (150 KB)  PDF (3689 KB)  ( 20 )

171 The Influence of Scientific Collaboration Process Patterns on Performance in Funding Project Teams Yao Zhizhen, Rong Guoyang, Huang Xiaoming, Zhang Bin, Ma Feicheng DOI: 10.3772/j.issn.1000-0135.2025.02.005 The success of complex projects often depends on the ability to collaborate and solve problems. Scientific collaboration enables science teams to solve complex problems that individual researchers or disciplines cannot address, thereby demonstrating stronger potential for technological innovation. Previous studies have identified collaboration patterns from a static structural or network perspective without emphasizing how to dynamically coordinate to improve team performance. In this study, dynamic evolutionary modeling of the scientific collaboration process was conducted to intuitively analyze a team's dependency relationships and coordination mechanisms as well as the different improvement effects of the scientific collaboration process on project performance. A scientific collaboration process pattern mining approach based on graph and sequence mining was developed, and the optimal team collaboration process pattern was identified. The conclusions follow: 1) most scientific collaboration process patterns have a positive impact on project performance; 2) the most efficient collaboration process pattern begins with knowledge expansion and ends with knowledge enhancement; and 3) conversely, when a team adopts knowledge expansion as its overall development strategy or as a priority during project execution and relies on internal knowledge enhancement in the middle and later stages of development, it is not conducive to enhancing project performance. These findings have significant practical value and provide guidance for project applicants, leading to scientific cooperation. They can provide targeted suggestions and new implementation ideas for optimizing team formation, improving collaborative environments, promoting collaborative behaviors, and evaluating project performance. 2025 Vol. 44 (2): 171-184 [Abstract] ( 3 ) HTML (300 KB)  PDF (2272 KB)  ( 23 )

185 Research on Knowledge Co-creation Mechanism of Online Health Communities Based on Cognitive Assimilation Learning Theory Yi Ming, Xu Weizhuo, Zhou Yang, Li Han DOI: 10.3772/j.issn.1000-0135.2025.02.006 This study simplified the knowledge co-creation process in online health communities to propose disease treatment plans. Accordingly, this study used idea-type speech as the key object of group cognition analysis and the assimilation theory of cognitive learning to reveal the law of group cognition, and mapped the behavioral mechanism of knowledge co-creation in online health communities from inside and outside. The core of the assimilation theory of cognitive learning is the extraction of two typical cognitive modes: subordinate and derived cognition. Based on these distinct cognitive modes, different treatments of idea-type speech have been employed to construct a framework for analyzing collective cognition at macro and micro levels. At the macro level, collecting data about idea-type speeches from discussion posts was the analysis object. An algorithm was designed to mine the subordinate fulcrum, derived fulcrum, subordinate cognition, and derived cognition contained in the initial and idea-type speeches, and to extract the overall model of the patient group’s comprehensive use of the derived and subordinate cognition to produce various idea-type speeches for specific health issues. At the micro level, the analysis focused on group speeches in each time unit of discussion posts. Three indicators were designed to define the cognitive mode reflected in each time unit, and the distribution and transformation rules of dependent and derived cognition were explored in combination with life cycle patterns. An empirical analysis of 998 discussion posts on “Dancing with Cancer” revealed significant findings. At the macro level, for specific health issues concerning patients, the patient groups predominantly generated various idea-type speeches through the subordinate cognition-driven (15.93%), derived cognition-driven (49.30%), subordination cognition - derived cognition-synchronous driven (22.75%), and subordination cognition - derived cognition-iterative driven modes (12.02%). At the micro level, the life-cycle curve of 998 discussion posts can be divided into three patterns: gradual decline (28.56%), middle peak (26.35%), and tail rebound (45.09%). These patterns were dominated by group speeches from subordinate cognition and supplemented by group speeches from derived cognition with a ratio of approximately 8∶2. The distribution and transformation of the eight specific modes of subordinate and derived cognition were closely associated with the number of new idea-type speeches generated at each life cycle stage. 2025 Vol. 44 (2): 185-199 [Abstract] ( 8 ) HTML (167 KB)  PDF (5016 KB)  ( 15 )

Intelligence Technology and Application

200 Research on the Extraction and Application of Ancient Books' Restricted Domain Relation Based on Large Language Model Technology Liu Chang, Zhang Qi, Wang Dongbo, Shen Si, Wu Mengcheng, Liu Liu, Su Yushi DOI: 10.3772/j.issn.1000-0135.2025.02.007 Automatic extraction and structuring of fine-grained knowledge units in ancient books can provide a database for digital humanities research on ancient books, such as group biographies and historical maps. The extraction method based on the discriminative model is severely restricted by the semantic complexity of ancient Chinese and missing training samples, which limit the extraction and domain transfer effects. Related research is required to develop generative artificial intelligence technology. This study explores methods for restricted domain relation extraction in ancient texts based on large language models and the automatic generation of high-quality training corpora. After comparing the impact of different prompt templates on model extraction performance and proving the significance of fine-tuning methods in improving model performance, we utilize the ChatGPT4 application programming interface (API) service, combined with self-instruction, thought chains, and human feedback, to create a domain-specific relation extraction dataset for ancient texts. After data augmentation, F1 scores of 56.07% and 30.50% are achieved on two ancient text relation extraction datasets, exhibiting a significant improvement in transferability compared with models trained on the entire dataset. The study also explores the collaborative use of self-instruction and automatic evaluation models to synthesize training corpora, evaluation information, and trained models based on synthetic data, effectively alleviating the problem of insufficient training data. The findings indicate that using large language models to extract relational triplets and synthesize training data can significantly reduce the labor costs previously associated with domain-specific relation extraction and improve the efficiency of constructing knowledge graphs in the field of ancient texts. 2025 Vol. 44 (2): 200-219 [Abstract] ( 5 ) HTML (185 KB)  PDF (8085 KB)  ( 13 )

220 A Paper Semantic Representation Method Incorporating Academic Network with Content Information Shi Bin, Wang Hao, Li Xiaomin, Zhou Shu DOI: 10.3772/j.issn.1000-0135.2025.02.008 With the increasing number of scientific research workers, the publication of scientific and technological papers published has increased rapidly, making the work of archiving, inputting, and analyzing documents increasingly burdensome. Most of the classification models focus on the content information of the paper, ignoring the relevant information. To solve this problem, this study proposes a paper representation model called PAITKG, which integrates content information and academic networks. Knowledge graph embedding technology is introduced to characterize multiple relationship patterns of literature; SciBERT, which is fine-tuned by Adapter, is used to extract content features and integrate the two. In the training process, this study improves the dynamic counter loss function to guide the model to pay more attention to error results. It applies this method to literature classification and analysis in the field of digital humanities. In the multilabel classification of scientific and technological literature, PAITKG showed significant improvement compared with the baselines, which greatly improved the classification accuracy. In addition, the representation of PAITKG has been more widely applied through the learning of upstream tasks. Without any additional training, the feature vectors extracted by the model can be applied to analysis tasks such as topic clustering and scholar recommendation. The experiments show that PAITKG can effectively integrate the associated literature information and improve the understanding of literature data by constructing and characterizing the academic networks of papers. Moreover, the representations learned by PAITKG have excellent generalization potential and can be applied to various literature analysis work. 2025 Vol. 44 (2): 220-233 [Abstract] ( 10 ) HTML (209 KB)  PDF (3257 KB)  ( 23 )

234 Research on Technology Theme Identification Based on Investment and Financing Events —Using Biotechnology as an Example Wang Yicheng, Jiang Xingyu, Qin Qing, Liu Yunong, Zheng Yanning DOI: 10.3772/j.issn.1000-0135.2025.02.009 Identifying technical themes in a field through investment and financing events and analyzing their evolutionary trends is of great significance for investment planning and technological innovation in the science, finance, and business sectors. First, feature words were extracted using the BERTopic static topic model, and theme clusters were generated using the HDBSCAN (hierarchical density-based spatial clustering of applications with noise) algorithm. The c-TF-IDF algorithm was then employed to extract thematic feature words from these clusters. Combining this with technical details from the websites of financing companies, technical themes were named based on biotechnology expertise. Visualization techniques were used to construct a map of the evolution paths of these technical themes, allowing for analysis of their evolutionary trends. Finally, by analyzing the corpus of investment and financing events in the biotechnology field, the popularity levels of different technical themes in the capital market were compared, expanding methods of study for identifying technological topics. 2025 Vol. 44 (2): 234-245 [Abstract] ( 7 ) HTML (122 KB)  PDF (2317 KB)  ( 19 )

Intelligence Discipline Development and Construction

246 Reflections on the Impact of Large Language Models on the Development of Information Science Li Yang, Sun Jianjun DOI: 10.3772/j.issn.1000-0135.2025.02.010 The application of large language models represented by ChatGPT has a continuous profound impact on human society; different knowledge domains are “compressed” and “mapped” into successive large language models. The rise of large language models has given the information world a new form. The typical characteristics of the information world are manifested in three aspects: the massive production of artificial intelligence generated content; the rising status of machines; and the emergence of large language models as the new quality productivity engines. Information science has always exhibited a high degree of sensitivity to new technologies in line with the information world. The new form of the information world exerts a profound impact on the research issues, target tasks, theoretical systems, research paradigms, and visibility of the discipline. As a result, two different paths have emerged, namely, tool perspective and object perspective. The former involves intelligent information analysis and processing empowered by large language models and the construction and application of information large language models for diversified scenarios, whereas the latter covers topics such as good governance of artificial intelligence generated content in the context of the convergence of security and development and information users and behaviors in the era of large language models. Therefore, it is necessary to explore the academic environment, data infrastructure construction, and education and talent training to further support the development of large language models and the construction of disciplinary discourse systems. 2025 Vol. 44 (2): 246-256 [Abstract] ( 15 ) HTML (107 KB)  PDF (1254 KB)  ( 33 )