Intelligence Theories and Methods

1139 Knowledge Diversity and Its Influence on Technical Team Innovation Performance Hot! Shi Jing, Sun Jianjun DOI: 10.3772/j.issn.1000-0135.2023.10.001 Team innovation is a distinctive feature of the “Big Tech Era.” Numerous studies have examined the innovation process of scientific teams, but technical teams have received insufficient attention. Particularly, the knowledge diversity of technical teams and its impact on innovation performance remains unclear. Considering the features of technical teams, we propose a new method to identify teams, then construct team knowledge networks, measure knowledge diversity, and explore its impact on team innovation performance. The impact is analyzed from three dimensions: innovation quantity, innovation quality, and innovation breadth. The results reveal that knowledge diversity can significantly improve the innovation performance in all three dimensions. Specifically, when aiming at innovation productivity, larger teams can better exert a positive effect of knowledge diversity. However, when focusing on innovation novelty and originality, knowledge diversity can improve innovation performance more obviously in a smaller team. This study can be considered with the relevant research on scientific teams to help understand the different characteristics and mechanisms in scientific and technical teams. It can also assist R&D departments and managers to allocate resources for different innovation goals. 2023 Vol. 42 (10): 1139-1150 [Abstract] ( 165 ) HTML (146 KB)  PDF (2905 KB)  ( 140 )

1151 Discipline Identification Methods for Knowledge Units: A Comparative Study towards Feature Mining Hot! Cao Yujie, Xiang Rongrong, Mao Jin, Wang Shiyun DOI: 10.3772/j.issn.1000-0135.2023.10.002 The features of knowledge units are the basis of the discipline identification of knowledge units. Mining the key features of knowledge units can help improve performance, so as to better serve the study of interdisciplinary research at the knowledge content level. In this study, with the help of 16 methods of discipline identification for knowledge units, we compared and analyzed the discriminative performance of these methods for knowledge units with different word frequencies and disciplinary coverages. Further, we evaluated the effect of the three features and feature combinations of disciplinary importance, disciplinary relevance, and disciplinary discriminability implied by the methods to mine the subset of features with the best effect. We also constructed a test dataset based on data from the cross-cutting field of “computational medicine.” The experimental analysis results showed that the combined use of the three features achieved better performance on all groups, while the performance advantage of disciplinary importance indicates that it is the most important among the three features; the discipline identification of high-frequency words needs to focus on disciplinary importance, while low-frequency words need to focus on disciplinary importance. For knowledge units with multidisciplinary coverage, it is necessary to consider disciplinary differentiation in addition to disciplinary importance. The findings of this study provide theoretical guidance and practical suggestions for the optimization of discipline identification methods for knowledge units. 2023 Vol. 42 (10): 1151-1165 [Abstract] ( 112 ) HTML (170 KB)  PDF (2788 KB)  ( 132 )

1166 Identification of Sleeping Beauties in Altmetrics Hot! Xiang Fei, Chen Huafang, Shen Tong, Cao Guang, Liu Yan DOI: 10.3772/j.issn.1000-0135.2023.10.003 In essence, Altmetrics (alternative metrics) and citations are both measures of the impact of academic achievements; therefore, the “sleeping beauties” (SBs) that appear in citations also appear in Altmetrics. Those based on Altmetrics (A-SBs), which are initially unknown, resulting in a waste of knowledge, are of great value. Realizing the early prediction of A-SBs can improve article utilization, enhance public wisdom, and reflect the public’s attention. Developing a method to identify SBs is the first step to realizing their early prediction. The two most important features of A-SBs are a long sleeping time and sudden increase in attention. According to quartiles and Bcp, an Altmetrics sleeping beauty (ASB) index was developed to identify A-SBs with the core of these features. This study included the articles with the top 1% of the attention as high-attention articles, which were used to test the effect of the ASB index. The effect of the ASB index was analyzed via the comparison of the accumulation curves of attention and index features of 10 articles with the highest, median, and lowest ASB values. The results indicated that the ASB index is effective in the recognition of A-SBs. 2023 Vol. 42 (10): 1166-1175 [Abstract] ( 112 ) HTML (155 KB)  PDF (1351 KB)  ( 119 )

Intelligence Technology and Application

1176 Structured Abstract Generation for Scientific and Technological Papers by Integrating Moves and Text Features Hot! Xi Haixu, He Sheng, Huang Chunguo DOI: 10.3772/j.issn.1000-0135.2023.10.004 In the era of mobile Internet, mobile and fragmented reading have become the main means of public reading. One of the important ways to solve the problem of information overload is to provide key summary content for improving reading efficiency. The task of abstracting scientific and technological papers is more challenging than that of abstracting ordinary texts, such as news, because of their length, varying content, and domain knowledge. This paper proposes a structured summarization method for scientific papers. First, scientific papers are divided into different moves, and then, the texts of different moves are abstracted separately. The multiple features of the text are integrated into the iterative calculation process of the TextRank algorithm according to weight, and the MMR algorithm is introduced to redundantly process the pre-selected abstract set. Finally, the text is semantically analyzed using dependency syntax analysis, and the summary is further streamlined and combined into a structured summary. Experimental results show that this method is different from the benchmark model in terms of the relevance, diversity, and readability of different moves. Combined with manual evaluation, this method can significantly improve the diversity of the summary while simultaneously improving the relevance and readability of the summary to a certain extent. 2023 Vol. 42 (10): 1176-1186 [Abstract] ( 77 ) HTML (165 KB)  PDF (1196 KB)  ( 202 )

1187 Construction and Analysis of Semantic-Enhanced Full-Text Co-Occurrence Network Hot! Zhao Yiming, Yin Jiaying DOI: 10.3772/j.issn.1000-0135.2023.10.005 A co-occurrence network is an important method to investigate linguistic phenomena, while semantic features are important tacit knowledge in co-occurring words. Examining the semantic relationships and characteristics of such words can improve the research on co-occurrence networks from a semantic perspective and enrich the existing analysis methods with semantic knowledge. This study proposes a semantic-enhanced co-occurrence network construction and analysis method, which enriches the properties of co-occurrence network nodes and edges from the three dimensions of co-occurring, semantic, and network features. A semantic-enhanced co-occurrence network based on more than 140,000 news text data items is then constructed through experiments. Analysis of the semantic features of co-occurrence word reveals the value of semantic features in the field of computer linguistics and industry application. From the semantic perspective, this study expands the construction and analysis method of co-occurrence word network, describes the semantic characteristics of co-occurrence words, and verifies the asymmetric and transitive properties of semantic relations through experiments, which provides data verification for the classification and derivation of semantic relations. Through the semantic-enhanced co-occurrence network, semantic knowledge can be enriched in semantic disambiguation, word meaning understanding, and legal applications. 2023 Vol. 42 (10): 1187-1198 [Abstract] ( 103 ) HTML (114 KB)  PDF (3390 KB)  ( 251 )

1199 Research on Fine-Grained Technology Opportunity Analysis Based on Patent Text Mining Hot! Wu Keye, Sun Jianjun, Xie Ziyue DOI: 10.3772/j.issn.1000-0135.2023.10.006 In the new round of scientific and technological revolution and industrial transformation, the strategic position of technology opportunity analysis in R&D management and corporate decision-making is growing. However, the accuracy of technology opportunity analysis based on traditional link prediction indicators has reached a bottleneck, the stubborn expertise can hardly cope with the dynamics and complexity of technological innovation, and fine-grained technical opportunity identification and analysis are difficult to realize. As a result, this study proposes a fine-grained technical opportunity analysis framework based on patent text mining that combines patent text mining and the graph neural network link prediction method and divides technology opportunity analysis into three research subtasks: knowledge network construction and evolution analysis, element link prediction and technology opportunity assessment, and screening. An empirical study in the field of computer vision shows that the knowledge network built using multi-dimensional keyword features can fully present the knowledge panorama of cross-fields, and the combination of complex network indicators and time series can further reveal the context of technological development and provide direction for subsequent technological opportunity analysis guidance. The BERT model combined with the graph neural network method is suitable for the knowledge element link prediction task of each technology life cycle, and it shows higher accuracy and robustness than traditional prediction indicators. Following a comparison and evaluation with multi-source technical reports, it is confirmed that the nine technical opportunities based on this framework are in line with the current development of computer vision technology and have practical R&D value. 2023 Vol. 42 (10): 1199-1212 [Abstract] ( 142 ) HTML (135 KB)  PDF (5336 KB)  ( 242 )

Intelligence Users and Behavior

1213 Exploration and Analysis of the Information Empowerment of Information Behavior Research Hot! Liu Jinya, Wang Xinyue, Meng Gaohui, Li Yujia, Peng Hanqi, Song Xiaoxuan, Liu Chang, Wang Yanfei DOI: 10.3772/j.issn.1000-0135.2023.10.007 In the context of the requirements for Chinese-style modernization, the information science enterprise has entered a new stage of development. The introduction of the information empowerment concept injects new momentum into academic exploration in the field of information science and opens new space for research on information behavior research. Based on the philosophy of information empowerment, this study reviewed the achievement of information behavior research, aiming to explore the possibilities of the information empowerment concept empowering information behavior research. It is expected to achieve information empowerment by improving the efficiency of information works, such as clue discovery, analysis studies, and decision making, and enhance information users’ decision-making ability. Moreover, based on the philosophy of information empowerment, the practical value of information behavior research can be widely extended, and future research can open new possibilities. 2023 Vol. 42 (10): 1213-1223 [Abstract] ( 147 ) HTML (151 KB)  PDF (1005 KB)  ( 131 )

1224 Research Interest Space Mining and Double-Task Research Recommendation Based on Heterogeneous Network Embedding Hot! Cui Hongfei, Feng Zihan, Zhang Jingyu DOI: 10.3772/j.issn.1000-0135.2023.10.008 The Internet contains rich literature databases that are crucial resources for researchers to understand various advances in their fields. Efficient information mining on a massive scale of research outcomes within specific fields can provide researchers globally with clearer directions amidst knowledge flow. In this regard, based on more than 130,000 scientific papers spanning 11 years (2010-2021) collected from a well-known biomedical literature database, PubMed, this study investigated the historical behavioral information of researchers. A heterogeneous information network containing authors, papers, and keywords was built, in which the nodes were subsequently embedded into a “heterogeneous research interest space” using the metapath2vec heterogeneous network embedding algorithm. Simultaneously, both research collaborator and keyword recommendation tasks were implemented based on the similarity metric between vectors in the space. Compared with existing studies, the proposed method can achieve double-task recommendation, obtaining not only meaningful results in the traditional research collaborator recommendation task but also significantly surpassing existing performance in heterogeneous keyword recommendation between researchers and keywords. Furthermore, an in-depth mining of author and keyword vectors in the space was conducted, proving that the heterogeneous research interest space can indeed provide inductive direction in understanding the semantic meaning of researchers’ interests in various research fields. The characteristics of the research interest space will provide new perspectives for researchers in terms of mining and comprehensive understanding of research interests. 2023 Vol. 42 (10): 1224-1237 [Abstract] ( 82 ) HTML (151 KB)  PDF (3581 KB)  ( 109 )

Intelligence Reviews and Comments

1238 Review of Patent Mining Methods in Technology Opportunity Discovery Hot! Wei Tingting, Feng Danyu, Song Shiling, Zhang Jiantao DOI: 10.3772/j.issn.1000-0135.2023.10.009 In the context of the ongoing technological revolution, research on technology opportunity discovery has received widespread attention. Technology opportunity discovery aims to discover new technology trends and infer possible technology forms and development points in the field; this is critical for technological innovation and industrial development. This paper presents a systematic review of the current situation of patent mining methods in technology opportunity discovery, summarizes the representative research of five underlying common analysis methods, clarifies the relationship between the analysis methods and research content, and provides a reference basis for the technology selection of subsequent research and practices in this field. It is shown that the methodological tools used for technology opportunity discovery have not yet been targeted for innovations in method application, despite their following of deep learning developments. Finally, ideas for improvement are proposed from three perspectives: the data, method application, and evaluation system. 2023 Vol. 42 (10): 1238-1250 [Abstract] ( 126 ) HTML (162 KB)  PDF (2140 KB)  ( 133 )

1251 Review on Patent Technology Complementarity: Concepts, Measures, and Applications Hot! Zhang Jinzhu, Shi Jialu, Zhang Chengzhi DOI: 10.3772/j.issn.1000-0135.2023.10.010 As an important reference for technological innovation in various organizations, patent technology complementarity has recently received extensive attention from scholars worldwide. This paper first presents a review of the evolution of the technology complementarity concept and generalizes its measurement methods from four perspectives: industry/trade classification, patent classification, patent citation relationship, and patent content feature association. Then, various applications of patent technology complementarity are summarized. This review focuses on the concept of patent technology complementarity and indicates that relevant studies primarily use patent classification or a patent citation network to form technology complementarity measurement indices and methods, which are mostly applied in the innovation of the performance factor determination, enterprise of merger and acquisition (M&A) decision making, and potential partner discovery. In the future, we intend to continue to refine the technology complementarity concept, make comprehensive use of multimodal heterogeneous data such as patent text, chart, and market information, design fine-grained quantitative measurement indices, improve the accuracy of the patent technology complementarity measurement by introducing recently developed methods such as deep learning, and broaden the application effect and scope of patent technology complementarity. 2023 Vol. 42 (10): 1251-1264 [Abstract] ( 124 ) HTML (179 KB)  PDF (1141 KB)  ( 220 )