Special Topics

637 Potential Disruptive Technology Identification Method Based on Graph Representation Learning Hot! Dou Yongxiang, Kai Qing, Wang Jiamin DOI: 10.3772/j.issn.1000-0135.2023.06.001 In addition to disrupting existing technology systems and value networks, disruptive technologies can also drive technological innovation. Most traditional disruptive technology identification methods based on bibliometrics use data from theses and patents to first construct keyword networks and sets and then artificially construct high-level data features for analysis; obtaining all the structural information is difficult, which leads to a decrease in the identification accuracy. This study introduces a semi-supervised disruptive technology identification method based on graph representation learning. First, using data from scientific and technical literature databases, a keyword-weighted network is constructed with the keyword co-occurrence frequency and journal influence. Second, the keyword network vector is obtained by learning anonymous wandering sequences using a backpropagation algorithm. Third, potentially disruptive technologies are identified by comparing the similarity between vector sequences, which can describe similarities in technology evolution, the technology to be identified, and the recognized disruptive technology. Finally, ten technologies were selected as experimental targets from recent strategic planning and prediction reports related to disruptive technologies at home and abroad. Of these technologies, this study could identify three potentially disruptive technologies and determine two pseudo-disruptive technologies to be non-disruptive technologies with pre-given five disruptive technologies. 2023 Vol. 42 (6): 637-648 [Abstract] ( 265 ) HTML (1 KB)  PDF (6244 KB)  ( 265 )

649 Integrating Science-Technology Knowledge Linkage to Predict Disruptive Patents Hot! Liang Zhentao, Mao Jin, Li Gang DOI: 10.3772/j.issn.1000-0135.2023.06.002 Identifying and predicting disruptive technologies is critical to the national need for strategic development. This study treats patent families as technology units and calculates their disruption index, technological features, and Science-Technology (S&T) association features based on two large-scale patent (PATSTAT) and bibliographic (MAG) datasets. An approach to predicting potential disruptive patents is proposed. The prediction is considered a supervised binary classification task, which predicts the patent disruptiveness in five years, given the features calculated in the year it was published. Our results show that: (1) disruptive patents are characterized by less prior knowledge, stronger teams, an underestimated commercial value, and a higher long-term impact; (2) S&T linkage is an important feature in predicting disruptive patents; and (3) the LightGBM model achieves the best results in terms of performance and efficiency. However, the prediction of disruptive patents remains difficult. Future studies should consider incorporating semantic features and multiple data sources to improve the performance. 2023 Vol. 42 (6): 649-662 [Abstract] ( 407 ) HTML (1 KB)  PDF (1283 KB)  ( 306 )

663 Patent Valuation Method Based on a Combination of Feature Stitching, Label Migration, and Deep Learning Hot! Zhao Xuefeng, Hu Jinjin, Wu Delin, Wu Weiwei, Sun Andong, Zhao Tao DOI: 10.3772/j.issn.1000-0135.2023.06.003 Patent value evaluation is of great practical significance in cracking down on abnormal applications and purifying the market environment. This study uses feature combination, label migration, and deep learning combination to construct a patent valuation method and explores actual performance based on patents in Guangdong Province from 2010 to 2020. Several sets of comparative models are introduced for experimental analysis. Our findings reveal the following conclusions. (1) Stitching together the information of bibliographic documents can construct more powerful patented research objects with technical characteristics, thereby overcoming the phenomenon that evaluation accuracy is not high owing to the insufficient reflection of the nature of patented technology. (2) We can quantify a more representative patent value from the patent law. While extending the research depth, the mismatch between traditional labels and the actual value of the patent is also resolved. (3) A patent value evaluation model with bidirectional encoder representations from transformers (BERT) and long short-term memory (LSTM) is established based on the construction principle of high-precision word vector, effectively solving the disadvantages of low evaluation accuracy caused by the lack of feature extraction ability of traditional models. This study has a strong application value and presents improvement strategies from the three aspects of research object effectiveness, label system, and model evaluation rate, providing a new tool for patent value evaluation. 2023 Vol. 42 (6): 663-680 [Abstract] ( 236 ) HTML (1 KB)  PDF (5236 KB)  ( 156 )

Intelligence Theories and Methods

681 Modeling and Simulation Analysis of Knowledge Exchange through Cross-Regional Scientific Research Cooperation Based on Cellular Automata Hot! Ye Guanghui, Wei Jinyu, Tan Qitao, Xia Lixin DOI: 10.3772/j.issn.1000-0135.2023.06.004 Knowledge exchange through cross-regional scientific research cooperation is significant in promoting innovation and value creation in scientific research. However, traditional analysis methods, such as social network, information econometric, and content analyses struggle to describe the complex process of knowledge exchange properly. Therefore, this study adopts the method of cellular automata modeling and simulation for analysis. First, we describe the individual characteristics of knowledge exchange through cross-regional scientific research cooperation. Then, based on the principle of cellular automata, we construct a knowledge exchange model and divide the knowledge exchange process into three stages according to the knowledge exchange characteristics. Finally, a three dimensional evaluation index system was constructed to evaluate the impact of different factors on the knowledge exchange effect. The results show that the hindering effect of geographical distance on the knowledge exchange through cross-regional scientific research cooperation gradually decreases with time, the knowledge exchange ability is key to improving the knowledge exchange efficiency, and knowledge innovation promotes the knowledge exchange constantly. Using modeling and simulation methods, this study explores the effects of different influencing factors on the knowledge exchange through cross-regional scientific research cooperation, discusses the process and characteristics of knowledge exchange, reveals the internal mechanism of knowledge exchange, and addresses the challenges to knowledge exchange in terms of space, time, and state. 2023 Vol. 42 (6): 681-689 [Abstract] ( 154 ) HTML (1 KB)  PDF (1548 KB)  ( 94 )

690 The Interaction and Relevance between Chinese Government Information Disclosure Policy and Related Research from the Perspective of Academic Citation Hot! Zhai Yujia, Xu Jia, Liang Yixiao DOI: 10.3772/j.issn.1000-0135.2023.06.005 From the perspective of academic citations, this paper takes government information disclosure policies and related social science research as the main research object and analyzes the interactive relationships between them from three dimensions: topic, citation, and association. The results show that, regarding the topic dimension, Chinese scholars’ focus on the research topics and research contents of government information disclosure is relatively scattered and migrates with the time nodes of different proclaimed policies. Regarding the citation dimension, researchers mainly directly cite policies or their contents as background or theoretical basis in the introduction and review/method/analysis sections. The citations to “Regulations of the People’s Republic of China on the Disclosure of Government Information” are mainly concentrated in the first three chapters, which are embodied in the citations to the subject, scope, method, and procedure of government information disclosure. Regarding the association dimension, this study constructs a co-occurrence network of government information disclosure policies by citing coupling relations. The results show that “Regulations of the People’s Republic of China on the Disclosure of Government Information,” “Archives Law of the People’s Republic of China,” “Law of the People’s Republic of China on the Preservation of State Secrets,” and “Constitution of the People’s Republic of China” have the most co-occurrences. In addition, the policy contents that have received more attention primarily focus on two aspects: the discussion as to whether there are problems with a specific clause in different policies and the theoretical analysis of the relationships between clauses in different policies, which are mainly manifested as mutually supportive or contradictory. 2023 Vol. 42 (6): 690-701 [Abstract] ( 233 ) HTML (1 KB)  PDF (3642 KB)  ( 167 )

702 Quantitative Evaluation Model for Humanities and Social Sciences Journals Based on Association Tautology Hot! Han Songtao, Li Jie, Jin Jiali, Huang Chen DOI: 10.3772/j.issn.1000-0135.2023.06.006 Grounded in the studies on “association tautology”, this article aims to develop an academic evaluation model by evaluating published papers from scholars of universities with first-class disciplines. Through a quantitative analysis of 500,000 papers from more than 5,000 journals in Social Sciences Citation Index (SSCI) and Arts & Humanities Citation Index (A&HCI) over the past five years, a data-based evaluation model for humanities and social sciences journals is proposed—an evaluation model for first-class disciplines. This evaluation model can effectively avoid subjective uncertainty and is empirically proved available. Using scholars’ submission behaviors as a starting point, this sustainable dynamic evaluation model has unparalleled advantages over current citation analysis models, and can be improved in application. 2023 Vol. 42 (6): 702-710 [Abstract] ( 127 ) HTML (1 KB)  PDF (777 KB)  ( 149 )

Intelligence Technology and Application

711 Identifying “Sleeping Beauties” in Cell Biology and Exploring Their Classical Applications through an Improved BP Network and Function Fitting Hot! Hu Zewen, Jin Xinyue, Cui Jingjing DOI: 10.3772/j.issn.1000-0135.2023.06.007 Identifying “sleeping beauties” from a large number of studies and recommending them to the scientific community can enable the full use of their scientific and technological value, thus driving the development of science and technology. In this study, we designed and implemented an improved back propagation (BP) neural network model by merging the K-value algorithm, quadratic function fitting method, least squares method, and an iterative algorithm. We then used these methods to identify “sleeping beauties” from 401,130 papers in the field of cell biology, from 1990 to 2010, and explored the classical applications of the identified papers. The results show that: (1) the BP neural network can improve the degree of automation in identifying “sleeping beauties.” However, it is necessary to identify some “sleeping beauties” in advance in a training set to train the recognition model. The improved bivariate quadratic function fitting method and Gini coefficient, based on the least squares method, an iterative algorithm, and a slicing algorithm, demonstrate optimal speed in identifying “sleeping beauties”. (2) The recognition effect of the bivariate quadratic function fitting method is not affected by the length of the citation period. However, the recognition effect of the Gini coefficient is influenced by the length of the citation period. This is illustrated by the fact that the number of identified “sleeping beauties” from papers within a shorter citation period (i.e., published between 2001 and 2010) is 15 times as much as that from papers within a longer citation period (i.e., published between 1990 and 2000). (3) In the same field, there is a difference in the number of “sleeping beauties” identified using different methods. As an illustration, among 257,562 papers, the K-value algorithm, BP neural network model, and quadratic function fitting method with optimal recognition effect can identify between 30 and 223 “sleeping beauties”, an identification percentage that is less than 0.09%. The Gini coefficient with a poorer recognition effect is influenced by the length of the citation period, and identifies a maximum of 1066 “sleeping beauties”; i.e., the percentage increases to 0.41%. (4) The annual distribution of the number of identified “sleeping beauties” is maintained at a stable percentage between 0.02% and 0.17%. (5) The mechanism of identifying “sleeping beauties” can be widely applied in the comparison and analysis of bibliometric features among different types of literatures, as well as the recognition and recommendation of research hotspots. Moreover, the value of the contents of such excellent papers identified as “sleeping beauties” could then be realized. 2023 Vol. 42 (6): 711-728 [Abstract] ( 170 ) HTML (1 KB)  PDF (4773 KB)  ( 193 )

729 Visualizing the Temporal and Spatial Emotional Trajectory of Poets Based on Knowledge Reconstruction: Focusing on Xin Qiji Hot! Zhang Qiang, Gao Jinsong, Long Jiaqing, Yang Xiaoyan, Xia Hongyu, Jiang Zhihui DOI: 10.3772/j.issn.1000-0135.2023.06.008 The extensive application of digital technology to humanities research provides new research methods and paradigms for addressing traditional humanities problems. While traditional humanities scholars often focus on “time” over “space” when researching poets, digital humanities technology facilitates exploring the changes in the temporal and spatial emotional trajectories of poets from a macro perspective, thereby clarifying the overall characteristics of poets. Accordingly, this study reconstructs the relevant data of Xin Qiji (1140-1207) along multiple dimensions. First, a knowledge map of Xin is constructed in a top-down manner to complete static knowledge associations. Second, combined with GIS technology, this study demonstrate the trends in his emotional trajectory over time and space to understand the complete the dynamics of knowledge display. Finally, from the time, space, emotion, and other dimensional of Xin, this study display the knowledge discovery. This verification shows that the method can effectively describe and visualize poets in full, explore hidden knowledge about them, and provide new practical means for humanistic research. 2023 Vol. 42 (6): 729-739 [Abstract] ( 284 ) HTML (1 KB)  PDF (3461 KB)  ( 115 )

740 Ancient Chinese Word Segmentation Based on Graph Convolutional Neural Network Hot! Tang Xuemei, Su Qi, Wang Jun, Yang Hao DOI: 10.3772/j.issn.1000-0135.2023.06.009 The syntax of ancient Chinese is characterized by the omission and inversion of word order, and morphology is characterized by the word-class shift and the abundance of pronouns and nouns. These features increase the difficulty of ancient Chinese word segmentation (CWS) and lead to the serious out-of-vocabulary (OOV) problem. Recently, deep learning methods have been widely used on ancient CWS tasks and achieved significant success. However, these works paid more attention to improving the performance of CWS and ignored the OOV issue, a major challenge in CWS. Therefore, we propose an ancient CWS framework that combines the pre-trained language model and the graph convolutional neural network, integrating external knowledge into the neural network model to relieve the OOV problem. The experimental results on three ancient Chinese CWS datasets (Zuo Zhuan, Stratagems of the Warring States, and The Scholars) demonstrate that our model improves the word segmentation performance of the three datasets. Further analysis illustrates that our model can effectively integrate lexicon and N-gram information. In particular, N-gram helps to alleviate the OOV problem. 2023 Vol. 42 (6): 740-750 [Abstract] ( 261 ) HTML (1 KB)  PDF (1230 KB)  ( 105 )

751 Simulation Research on the Incentive Strategy in Scientific Research Data Sharing and Expected Revenue Analysis between Researchers and Sharing Intermediaries Hot! Gao Xiaoning, Gao Mingzhu DOI: 10.3772/j.issn.1000-0135.2023.06.010 In the process of scientific research data sharing, intermediaries tend to lack the willingness to truly convey information about data quality and requirements, leading to hidden dangers such as low efficiency of scientific research data sharing and researchers taking most of the risks. Based on the principal-agent theory, this study investigates the relationship of interest game between data sharing intermediaries and researchers, introduces the revenue sharing coefficient to restrain the non-standard behaviors of data sharing intermediaries, then constructs an incentive mechanism of both parties to share risks, and solves the optimal incentive strategy. The conclusions of this study provide management inspirations and decision support for scientific research personnel. Research indicates that the expected returns of researchers are concave to the incentive coefficient, and those of shared intermediaries are concave functions of their efforts. This reveals the existence of an optimal revenue sharing coefficient that enables data sharing intermediaries to transmit the real information of sharing process according to the wishes of the researchers. Moreover, the optimal incentive strategy also changes dynamically with the change of shared intermediary cost input, risk aversion degree, random factor, and other parameters. 2023 Vol. 42 (6): 751-760 [Abstract] ( 176 ) HTML (1 KB)  PDF (2084 KB)  ( 82 )