ABSTRACT. High-Tech Industrial Development Zone (HTZ) plays an important role in China's regional innovation system, yet the determinants shaping the spatial distribution of top-tier HTZs remain insufficiently understood. This study examines how economic strength, innovation inputs, innovation outputs, higher-education resources, and industrial innovation capacity jointly influence provincial-level differences in HTZ development. Using cross-provincial data for 2024, we construct a set of theoretically grounded indicators—including GDP per capita, R&D intensity, patents per 10,000 people, the number of 985/211 universities, and the number of high-tech enterprises—to evaluate their relationships with HTZ presence and scale. Three models are implemented: logistic regression for the presence of top HTZs, Poisson regression for HTZ counts, and a negative binomial model as a robustness check. This research highlights R&D intensity has the strongest relationship to both the presence and scale of top HTZs, followed by GDP per capita. The study contributes to understanding regional drivers of high-tech development in China and provides a data-driven basis for evaluating innovation investment, optimizing HTZ planning, and informing future research on spatial innovation dynamics.

ABSTRACT. Electric vehicles have become a cornerstone of the global energy transition. This paper examines the ripple effects of the European Union’s tariffs on Chinese electric vehicles, focusing on their transmission through critical metal supply chains and their implications for global welfare. We develop a multi-country, multi-sector quantitative trade model, calibrated with global input–output data, and analyze three key metals—copper, nickel, and aluminum. Our results show that tariffs reshape not only EU–China trade in electric vehicles but also generate heterogeneous shocks across upstream and midstream metal sectors via price and demand channels. The copper supply chain is the most affected, experiencing sharp fluctuations in exports and prices; nickel shows high sensitivity in the mining stage; while aluminum demonstrates relative resilience due to its diversified downstream demand. Welfare effects are highly uneven: China, as the major producer and consumer, suffers substantial losses; EU manufacturing powers benefit from domestic substitution; and resource exporters such as Russia and Australia gain from their endowment advantages. Overall, we highlights how trade barriers in the electric vehicle sector reverberate through critical metal supply chains, underscoring the importance of supply-chain heterogeneity in shaping policy outcomes and offering new empirical evidence on the broader consequences of trade protection in the green transition.

ABSTRACT. How do technology sectors in emerging economies transition to green innovation? How do institutional factors stimulate the transformation of green technologies? This study adopts an integrated Sectoral/Technological Innovation System (STIS) framework, combining Multi-Criteria Decision-Making (MCDM) with qualitative insights from elite interviews. The analysis focuses on the Taiwanese green technology sector, drawing data from a panel of 15 experts and 12 high-level stakeholders from industry, government, and academia, each with over a decade of strategic experience. Our analysis identifies a specific 'Policy-Driven STIS Model,' characterized by the dominance of Cross-Sector Collaboration and Government Industrial Policy over purely market-based drivers. Notably, a divergence emerged between qualitative and quantitative data: while experts qualitatively emphasized the pandemic's impact, they prioritized structural factors in quantitative rankings. This implies that the pandemic served as a context-dependent accelerator for digitalization rather than a fundamental driver of structural change. These findings offer a replicable framework for policymakers in emerging economies, highlighting the need for hybrid innovation policies that align institutional support with technological capabilities.

ABSTRACT. In knowledge-intensive industries, the forward-looking patent layout in key or core technology fields has become an important indicator for measuring an enterprise's technological strength and innovation capabilities. However, existing external patent recommendation and matching methods are difficult to fully reveal the strategic needs of enterprises in different technology fields and the potential relationships between technology fields. Considering the outstanding ability of graph neural networks and attention mechanisms in depicting the intrinsic relationships within different technological fields in recent years, this paper proposes a patent recommendation method based on a relational graph attention network. Firstly, a time-integrated enterprise technology portrait model is constructed using patent IPC numbers. On this basis, a heterogeneous information graph containing three types of nodes: enterprises, patents, and IPCs is built. The relational graph attention network (RGAT) is introduced to learn the semantic and structural information between heterogeneous nodes, thereby achieving high-precision patent recommendations oriented to the technological needs of enterprises. In the empirical analysis section, the heavy gas turbine field is selected. The experimental results show that this method significantly outperforms traditional baseline models in key indicators such as Precision, Recall, and F1 score, and the ablation experiments further demonstrate that the dynamic technology portrait mechanism of long-term and short-term data fusion has a significant synergetic effect. The results of this research provide a new path reference for the R&D decision-making and technology management of enterprises in the heavy gas turbine field and also offer certain theoretical and methodological references for the intelligence of patent recommendation systems and the assessment of enterprise innovation capabilities.

ABSTRACT. Local Citation Recommendation aims to retrieve appropriate references from the candidate pool based on the context of specific citation position. Existing methods primarily rely on the semantic representation of citation contexts for candidate matching. However, they fail to effectively distinguish differences in citation intent across various writing scenarios. Consequently, the recommendation results often struggle to precisely align with users' actual needs. Meanwhile, existing methods mainly employ end-to-end architectures and lack the ability to explain recommendation results, which limits user trust and adoption rate. To address these issues, we propose Citation Intent-Aware Explainable Local Citation Recommendation method (IAE-LCR). The IAE-LCR identifies citation intent and embeds it throughout the entire recommendation procedure, achieving to align accurate recommendations and explainable outputs. Specifically, we first design a citation intent recognition module based on Pre-trained Language Models to extract citation intent features from the context. Second, we construct an intent-enhanced recall and re-ranking mechanism, where the intent information is integrated into both stages to achieve complementary optimization. Finally, we propose a templated explanation generation strategy based on intent labels and keyword evidence. This approach ensures standardization while effectively avoiding the hallucination risks associated with free-form generation. Experimental results on SciCite, ACL-200, and RefSeer benchmark datasets demonstrate the effectiveness of our proposed method. It achieves Macro-F1 of 86.95% on the intent recognition task. For the citation recommendation task on ACL-200, our performance significantly outperforms all baseline models. Furthermore, human evaluation indicates that the generated explanations receive high ratings in terms of relevance, faithfulness, and usefulness.

ABSTRACT. The dynamic topic model extends traditional topic models to analyze temporal data, enabling the detection of topic evolutions over time. However, most of these models usually require simultaneous access to all the data, making them unsuitable for processing data that continuously streams in. This paper introduces a new dynamic topic modeling method, Density-Adaptive Stream Clustering (DASteamTopic), which is specifically designed for stream data. DASteamTopic combines pre-trained language models with a new stream clustering method that uses micro-clusters to improve the memory efficiency and the adaptability of stream data. It also introduces a unique density-adaptive distance function to measure micro-cluster distances, realizing the automatic number identification of the micro-clusters and the detection of clusters with arbitrary shapes. This method has been verified on three datasets: DBLP, Tweet and New York Times News. Compared with the state-of-the-art dynamic topic models, DASteamTopic generates higher quality topics. Experiment results reveal an improvement in topic quality of up to 32%. Additionally, DASteamTopic experiences less topic drift, ensuring that topics remain more consistent over time.

ABSTRACT. To address three bottlenecks in technology opportunity identification—namely, the coarse granularity of analysis, limited capability for large-scale data processing, and the continued reliance on experts’ subjective judgments in the evaluation process—this study presents a fine-grained identification and evaluation framework built upon a Knowledge graph. Using patent documents as the data source, the framework first constructs a domain-specific Knowledge graph by extracting technology elements and their relationships. Building on this graph, a Relation-aware link prediction algorithm is employed to automatically identify potential combinations of technology elements and generate a corresponding set of candidate technology opportunities. The framework further integrates topological features of the Knowledge graph with technological semantic information to formulate an objective set of evaluation indices spanning the dimensions of novelty, feasibility, and value. The entropy weight method and the TOPSIS method are then used to support data-driven screening of technology opportunities. An empirical study in the quantum information domain demonstrates that the framework can identify and prioritize high-potential technology opportunities while revealing specific implementation paths, thereby improving interpretability and practical operability. Overall, the identification and evaluation framework offers methodological support and valuable references for innovation decision-making in frontier technological domains.

ABSTRACT. Understanding the evolutionary trajectories of emerging technologies is critical for strategic decision-making amid intensifying global scientific competition. Current methods, predominantly reliant on single structural features, exhibit analytical limitations that compromise prediction accuracy. To address this gap, we propose a novel multi-feature fusion approach integrating three patent-derived dimensions: citation network structures, classification co-occurrence networks, and semantic text content. We construct multidimensional feature vectors, apply temporal clustering to delineate technology clusters within sequential intervals, and compute inter-cluster similarities to trace evolutionary pathways. The method’s efficacy is validated using a dataset of 61,683 Chinese new energy vehicle patents. We identify 17 distinct technology themes for 2022, visualize their evolutionary relationships via similarity-based linkages across time windows, and detect thematic convergence/differentiation patterns. By synthesizing empirical evolution paths, thematic prominence trends, policy analysis, and expert validation, we make predictions about the key future trends in China's new energy vehicle: (1) battery performance breakthroughs, (2) waste battery recycling. Our findings demonstrate that fusing multidimensional patent features with time-series cluster analysis significantly enhances the detection of emerging technology evolution paths, offering a robust framework for innovation tracking and strategic planning.

ABSTRACT. Key generic technologies (KGTs) are crucial for industrial upgrading and competitiveness. Existing studies are mostly ex-post and single-source. This paper proposes a multi-source, multi-indicator framework for prospective KGT identification. BERTopic extracts technology topics, a multi-dimensional indicator system supports evaluation, and XGBoost explores evolution from “emerging” to “key generic.” Using the smart-grid domain, the study identifies multiple ETs and KGTs, three evolutionary paths, and five transition patterns, priority cultivation, reserve enhancement, maturity upgrading, baseline maintenance, and wait-and-see, providing actionable insights for technology foresight, policy, and industrial planning.

ABSTRACT. With the rapid development of artificial intelligence (AI) technologies, identifying potential application scenarios has become a critical task for bridging technological supply and industrial demand. However, traditional approaches to application scenario identification largely rely on shallow semantic similarity matching, which makes it difficult to uncover deep, cross-domain, and latent associations between technologies and application scenarios. To address this limitation, this study proposes an integrated methodological framework that combines large language models (LLMs) with heterogeneous graph neural networks (HGNNs) to enhance both the breadth and quality of application scenario identification in the AI domain. First, large language models are employed to achieve deep semantic understanding of patent and academic publication texts and to automatically extract interpretable technology themes and application scenario themes. Based on these results, a heterogeneous graph comprising multiple types of nodes and relationships is constructed. A heterogeneous graph transformer (HGT) is then applied to jointly aggregate structural and semantic information, enabling the learning of rich representations for technology and application scenario nodes. Finally, a link prediction approach is used to identify novel associations between technology themes and application scenario themes that have not been explicitly documented. Empirical results demonstrate that the proposed framework can comprehensively and effectively identify multiple categories of potential application scenarios in the AI field. By integrating deep semantic understanding with complex relationship modeling, this study provides a novel methodological framework for application scenario identification and offers valuable insights for strategic R&D planning and the practical deployment of artificial intelligence technologies.

ABSTRACT. Industry demand-oriented technology transfer is crucial for accelerating the transformation of university patents and enhancing innovation efficiency. Accurately matching enterprise technological demands with transferable university patents remains challenging due to semantic heterogeneity, structural complexity, and limited observable transfer records. This paper proposes a heterogeneous graph neural network–based solution for industry demand-driven university patent recommendation. First, large language models are employed to extract and represent enterprise technological demands and university patent knowledge from patent texts, capturing fine-grained semantic information. Then, a heterogeneous network is constructed to jointly model enterprises, technological demands, university patents, and technology fields, incorporating semantic, structural, and classification-based relationships. Based on this network, a heterogeneous graph neural network is applied to learn demand–patent representations through relation-aware neighbourhood aggregation, enabling effective identification of patents with high potential for technology transfer. Finally, a demand-oriented patent recommendation framework is developed to rank university patents for enterprise-specific technological needs. An empirical case study in the artificial intelligence domain demonstrates that the proposed approach outperforms multiple baseline methods, validating its effectiveness and robustness in supporting university–industry technology transfer decision-making.

ABSTRACT. Research topic selection is a pivotal decision in a scholar’s academic trajectory, yet identifying viable topics remains challenging due to information overload in academic literature. To address this problem, we propose a GNN–LLM synergy framework that integrates structural connectivity learning with semantic expert reasoning for personalized scholar-topic recommendation. Unlike approaches relying solely on graph topology or semantic generation, our framework decomposes the recommendation process into two stages: (1) constructing a scholar–paper–topic heterogeneous bibliometric network and learning embeddings via HetGNNs to capture long-term research trajectories; and (2) empowering large language models (LLMs) as domain experts through Chain-of-Thought (CoT) prompting for reasoning-based filtering and re-ranking. By constraining LLMs to reason over graph-derived candidate topics, the framework effectively suppresses hallucinations while alleviating sparsity in graph-based methods. Experiments on Scopus data from Renmin University of China, with cross-institutional validation on Peking University and Tsinghua University, demonstrate superior performance, achieving an F1-score of up to 85.91% and bridging structural topology with deep semantic insight in scientometric recommendation.

ABSTRACT. Science, Technology, and Innovation (STI) policies are central to national and international competitiveness, yet their complexity makes systematic mapping and continuous monitoring a persistent challenge. This study draws on one of the largest initiatives in the field, the OECD STIP Compass survey, which collects and organizes data on STI policy from OECD countries and has historically relied on extensive manual survey efforts to ensure global consistency. Large Language Models (LLMs) are redefining representation learning in NLP, enabling them to process and internalize knowledge from long unstructured documents. This paper presents a novel application of LLMs for structured information extraction and generation from STI policy documents, focusing on OECD data across six sample countries. We develop a data extraction pipeline based on long-context in-context learning to encode task-specific schemas that allow learning of survey taxonomy labels from public URLs referencing policy initiatives. The pipeline integrates validation steps using a secondary LLM for relevance and evidence scoring, and comparison with survey responses completed by human respondents. For evaluation, we apply multiple overlap measures, including overlap ratios, agreement scores between human-generated and LLM-generated policy indicators, and K-fold cross-validation for AI-generated labels. Our findings indicate that LLMs can achieve high overlap with human respondents for policy indicators (84-95%). Qualitative analysis reveals that the model tends to provide more detailed descriptions, complementing human-written content. Our approach points to the potential of an AI-assisted framework for STI policy monitoring, enhancing both efficiency and quality in international policy intelligence.

ABSTRACT. This research investigates whether large language models (LLMs) can provide a content-based complement to existing journal evaluation systems in biological sciences. Using UK REF 2021 submissions in UoA 5 (Biological Sciences), we construct a dataset of journal articles with departmental REF quality profiles, citation data, and cleaned English abstracts. An LLM is prompted as a “virtual REF reviewer” to perform comparative quality judgements on pairs of titles and abstracts, producing both binary preferences and 1*–4* REF-style ratings. From these outputs, we derive LLM-based quality indicators (LLM_REF and BT_score), which are benchmarked against REF-derived departmental quality indicators (Dept REF mean) and citation-based impact indicators (MNLCS and SJR). Analyses examine citation patterns across star levels, overlap in highly ranked journals, novelty preferences, and rank correlations among indicators. The project aims to assess how far LLM-based content judgements can support, rather than replace, expert-led journal evaluation.

ABSTRACT. Limitations and future work sections are a core component of scientific papers, explicitly inviting authors to reflect on constraints encountered during the research process and to articulate directions for further inquiry informed by this experience. Despite their central role in scholarly communication, these sections have received limited systematic attention as an object of analysis in their own right. This research-in-progress paper presents an exploratory, methodological study of limitations and future research statements in recent Digital Libraries research, with the aim of establishing an initial analytical framework for studying such reflective sections. We focus on papers published between 2020 and 2025 that are classified under the cs.DL (Digital Libraries) subcategory in arXiv. As a proof-of-concept, we draw on a random subsample of 50 papers, from which 20 contain a dedicated section or subsection explicitly addressing limitations and/or future work. Using arXiv metadata and full-text PDFs, we extract and analyze these sections. To support the analysis, we explore the use of large language models (LLMs) to assist with the extraction, summarization and classification of limitations and future work statements. Based on an initial inspection of the extracted statements, we propose a preliminary typology consisting of ten broad categories that can be applied consistently to both limitations and future work. Applying this typology to the proof-of-concept subsample indicates that data- and method-related considerations are most frequently foregrounded, while aspects such as usability, computational constraints, and scope limitations are less commonly articulated. These observations are indicative rather than definitive and primarily serve to demonstrate the analytical potential of the proposed framework. By framing limitations and future work sections as a systematic object of analysis, this paper contributes to meta-research on scholarly communication and provides a preliminary analytical framework that can be extended to larger corpora, other research fields, and longitudinal studies of reflective scientific discourse.

ABSTRACT. This study examines the impact of Large Language Models (LLMs) on the research practices and productivity of young researchers. Through an analysis of PhD students at French universities who defended their theses between 2019 and 2025, we investigate how the emergence of LLMs has been integrated into research activities and the changes it has brought to academic practices. Using a matching strategy that compares students sharing the same supervisor before and after ChatGPT's release, combined with a placebo design, we provide evidence on the dimensions and effects of LLM adoption. Our quantitative analysis of nearly 6.000 STEM theses is complemented by a survey of 265 doctoral graduates, shedding light on the modalities and concerns related to LLM implementation among young scholars.

ABSTRACT. Citations constitute the core mechanism of scientific knowledge dissemination. However, existing research predominantly focuses on rhetorical functions or authorial intent, overlooking the substantive knowledge contributions of cited works. More critically, prior studies often analyze isolated contribution statements rather than authentic citation contexts, resulting in findings divorced from real-world scenarios. Consequently, accurately identifying the knowledge contributions of citations has become a pivotal challenge in understanding scientific knowledge production and evolution. To address these challenges, we propose the Scientific Knowledge Contribution (SKC) classification framework, which for the first time categorizes citations from an objective contribution perspective into five classes: Method & Technology, Resource & Tool, Theory, Empirical Finding, and Background. Through two-stage sampling, we construct a high-quality annotated dataset and employ six representative models for systematic evaluation, establishing performance benchmarks for this task. This study achieves a methodological shift in citation analysis from subjective intent to objective contribution. The theoretical framework reveals knowledge dissemination characteristics in the NLP domain while uncovering performance limitations of existing models. This research provides essential infrastructure for fine-grained knowledge contribution identification, establishing foundations for large-scale scientific knowledge flow analysis and disciplinary evolution studies.

ABSTRACT. Building on the context of big science facilities, this study provides a more micro method to identify the scientific mobility procedure, named short-term scientific mobility hereafter, and associates with ability to yield novel and disruptive knowledge. We classify external users of big science facilities into two types (travelers and locals) by measuring the number of facilities the focal author used in one-year, represented by the number of publications supported by the corresponding facilities. We also include these authors’ other publications without supports from big science facilities. Scientific performance is measured by novelty score and the ten-year disruption index. Results show that locals associated with more disruption while travelers perform better in novel knowledge production. Diverse variables related to travel validate the robustness of our preliminary results. This study contributes to understanding the performance evaluation and science policy in the context of big science facilities and enriches the research in scientific mobility. The discoveries also could be a reference for those short periods scientific activity related to mobility but without visible information to quantify.

ABSTRACT. This study draws on the guideline module of the Chinese Medical Journal Full-text Database to systematically analyze the characteristics of academic references cited in Chinese clinical guidelines during the COVID-19 pandemic and to explore how they reflect the societal impact of research. Based on an analysis of 49 guidelines containing 1,617 references, the study reveals the following: (1) Journal articles constituted the predominant type of references (82.7%), indicating the strong reliance of guidelines on scholarly publications as evidence; (2) the temporal distribution exhibited pandemic-driven, stage-specific features, with a marked citation peak in 2020, reflecting the amplified demand for evidence during a public health emergency; (3) the country distribution was dominated by journals from the United States and the United Kingdom (58.9% combined), while papers published in Chinese-language journals (17.7%) played a distinctive role in providing contextualized evidence; (4) the journal distribution showed that leading international medical journals and core domestic journals jointly formed the evidence base; and (5) case analysis demonstrated that the evidence base comprised both domestic epidemiological and policy studies and international high-quality clinical and pathological studies, jointly supporting the rapid response and evidence-based decision-making of the guidelines during the pandemic. The findings suggest that evidence adoption in Chinese clinical guidelines followed a dual-track model of rapid domestic response combined with authoritative international support. This approach enhanced both the scientific rigor and contextual adaptability of the guidelines, while also providing a China-specific empirical pathway for evaluating the societal impact of research.

ABSTRACT. Scientific knowledge serves as the theoretical foundation for technological development and innovation, and the scientific knowledge network, with scholarly literature as its knowledge carrier, plays a crucial role in full-cycle technology forecasting. While current technology forecasting research has made significant algorithmic progress in handling large-scale data using black-box AI models, such as deep learning, it provides limited support for domain experts’ judgment and lacks explainable insights and theoretical foundations for the complex interactions among scientific and technological entities. Building upon existing technology forecasting theoretical models, this study proposes a new theoretical model of influencing factors for technology forecasting, thoroughly analyzing the mechanism by which semantic content and citation structure features in scientific knowledge networks enhance forecasting effectiveness. To validate the theoretical model, we designed a human-AI collaborative forecasting framework based on matrix decomposition and trend extrapolation. Using computer science literature data, we constructed a multiplex graph and employed frequency, semantic and citation integrated feature to predict the frequency of keywords representing sci-tech research topics. Feature ablation experiments and temporal performance analysis across different network scales demonstrate that semantic and structure features in scientific knowledge networks significantly improve forecasting performance. Moreover, the results reveal a widespread teleconnection phenomenon of multi-dimensional fused features across the network. This study offers new theoretical foundations and methodological guidance for technology forecasting in theory, practice, and policy making.

ABSTRACT. Understanding how the amount of experimental resources used in scientific experiments relates to research innovativeness is an important issue in the scientometric analysis of large-scale research infrastructures. While existing studies have extensively examined the aggregate productivity and citation-based performance of major facilities, relatively limited attention has been paid to whether variations in experimental resource use at the paper level are systematically associated with differences in research innovativeness. This study addresses this issue by examining the relationship between experimental procedural complexity and paper-level innovativeness in the context of synchrotron radiation research.

Using publication data from the European Synchrotron Radiation Facility (ESRF), experimental procedural complexity is operationalized as the number of beamline stations employed per paper. This measure reflects the amount of experimental resources required to conduct a study and serves as a parsimonious proxy for the procedural complexity of experimental execution, insofar as experiments involving multiple beamlines typically entail more diverse measurement techniques and experimental configurations. Research innovativeness is measured using the Relative Independence Index (IND), a citation-network-based indicator capturing the degree to which a publication positions itself independently from its referenced literature. Unlike citation volume or diffusion-based metrics, IND reflects a structural aspect of innovativeness related to deviations from established citation pathways.

Methodologically, the study adopts a generalized propensity score matching (GPSM) framework to estimate the dose–response relationship between experimental procedural complexity and research innovativeness. Treating beamline usage as a continuous treatment variable and adjusting for a set of publication-level covariates, GPSM allows for a flexible assessment of how expected innovativeness and its marginal changes vary across different levels of experimental procedural complexity.

The empirical results indicate that publications based on experiments involving higher procedural complexity tend to exhibit greater innovative relative independence. The estimated dose–response function shows an overall increasing pattern, suggesting a positive association between the number of beamline stations used and research innovativeness. At the same time, marginal effect estimates reveal diminishing returns, with the positive association being more pronounced at lower levels of beamline usage and gradually weakening as additional beamline stations are incorporated. These findings point to a non-linear relationship between experimental procedural complexity and innovativeness rather than a simple linear scaling effect.

By demonstrating how scientometric methods can be applied to quantify the innovative implications of experimental procedural complexity, this study provides empirical evidence relevant to the assessment of experimental designs in large-scale research infrastructures. The results offer a methodological basis for informing discussions on experimental resource utilization at synchrotron radiation facilities, while remaining agnostic about the underlying organizational or managerial mechanism.

ABSTRACT. Patent citations embody the flow of technical knowledge and reflect the value and influence of patented technologies. This study constructs a three-layer network of keywords, IPCs, and inventors based on three perspectives: textual semantics, technical structure, and social cooperation. Combining co-occurrence and citation networks, a method for measuring and extracting potential citation relationships is designed. Co-occurrence, citation, and potential citation relationships are extracted for each layer, resulting in a multi-layer heterogeneous network of patent technologies. The R-GCN model is introduced to achieve patent representation learning based on multi-layer heterogeneous relationships. An attention mechanism is used to select and fuse important relationships, identifying key factors in patent citations to improve patent citation prediction capabilities.

ABSTRACT. To address the issue of unidimensional and incomplete technological main paths due to time lags in patent citations, this study proposes an integrated framework for extending main paths for technological development and identifies multi-category technology innovation opportunities. The multi-dimensional technological main paths are initially extracted combining the community detection and SPC algorithms. Then the technical-efficacy matrix based on SAO semantic analysis is mapped to acquire hot technical efficacies of each main path. Finally, we apply FOS to retrieve the latest scientific papers from similar technical domains with the hot efficacies, and then use technological novelty indicators and similarity indicators to screen and identify the technology innovation categories, thereby extending the original main paths. In summary, five main paths and three types of innovation opportunities in the field of ASSLIBs are identified.

ABSTRACT. Against the backdrop of a new round of scientific and technological revolution and industrial transformation, we examine in this study from a demand-side perspective how industrial innovation performance shapes the structure of technology convergence. Using four-digit CPC co-occurrence in biopharmaceutical patents, we construct a technology convergence network, incorporate network endogenous structure, structural positions, industrial innovation performance, and factors related to knowledge proximity, resource dependence, and path dependence, and apply an exponential random graph model (ERGM) to identify multiple mechanisms from the perspective of micro-level tie formation. The results show that the technology convergence network is globally sparse but locally clustered, and nodes with high centrality act as hubs in network expansion. After controlling for structural factors, industrial innovation performance, collaboration intensity, and policy support significantly promote the formation of technological ties, whereas certain knowledge diversity indicators and excessively strong market demand inhibit the creation of new cross-domain ties; disruptive technologies participate more frequently in cross-domain connections. This study provides empirical evidence for understanding the evolution of technology convergence structures under demand-side drivers and for optimizing industrial innovation policies.

ABSTRACT. To address the limitations of existing quantitative methods that rely on single indicators, such as traditional co-classification and co-citation, which struggle to identify synergistic relationships between technologies at the semantic level, this study constructs a systematic conceptual model of patent complementarity and proposes an intelligent identification method integrating Large Language Models. The research aims to achieve precise and efficient semantic identification of patent complementarity, providing a reference for technology layout and strategic decision-making. First, through literature review and conceptual analysis, a conceptual system for patent complementarity is established across dimensions including connotation, extension, core features, constituent elements, and multi-dimensional classifications (technical, legal, market, temporal, and contextual). This clarifies the essential characteristics of value dependence and synergistic enhancement and categorizes nine types of complementarity across the problem, solution, and effect layers within the technical dimension. Second, a method for "Problem-Solution-Effect" (P-S-E) extraction and complementarity identification is designed. This conceptual model is utilized to guide the design of LLM prompts, enhancing semantic understanding and logical reasoning capabilities. Finally, using 7,536 patents in the field of humanoid robotics as experimental samples, 23 categories of technical problems, 27 categories of technical solutions, and 40 categories of technical effect themes were extracted. A total of 5,919 (78.54%) complementary patents were identified, and high-value complementary patents were recommended by integrating complementarity value and endogenous value. Through expert evaluation and multi-model cross-validation, the method demonstrated stable performance in tasks involving P-S-E extraction (P1), complementarity identification (P2), complementarity type identification (P3), and complementarity rationale generation (P4). The overall accuracy reached 90% or higher, with the DeepSeek model performing best in hallucination control. The results indicate that the conceptual model and associated methods effectively break through the limitations of traditional quantitative research on patent complementarity, significantly enhancing the capability to identify complex technical complementarity relationships. This study still has limitations, such as coarse granularity in technical element representation, insufficient adaptability to corporate strategies, and a single data modality. Future research will further develop a hierarchical technical element system and integrate corporate strategic contexts and multi-modal data to improve the interpretability of identification results and decision-support capabilities.

ABSTRACT. This paper proposes a multi-dimensional graph convolutional patent classification model based on cross-modal fusion and technological evolution. It aims to improve classification efficiency while reducing manual effort in intellectual property management. Method: In this paper, graphs are constructed from both spatial and temporal dimensions, and two Graph Convolutional Networks (GCNs) are employed to extract hierarchical structural features and dynamic historical technological features. The structured features and the unstructured representations of patent texts are treated as heterogeneous bimodal data. To bridge these modalities, a novel cross-modal attention mechanism is introduced to uncover latent correlations in the semantic space, thereby achieving semantic alignment and effective fusion between hierarchical structural features and textual patent features. Results: Experiments on Chinese and English patent datasets demonstrate that the proposed model outperforms all baseline models across every evaluation metric.

ABSTRACT. Against the backdrop of globalization and technological transformation, technological convergence continues to drive innovation and profoundly reshape the global industrial landscape. However, current research on technological convergence has primarily focused on binary convergence, with insufficient exploration of multi-technology convergence. Moreover, existing studies on predicting technological convergence generally rely on existing network edges and overlook the informational gains from introducing new edges. Therefore, this study uses pharmaceutical patent data as a case study to construct a higher-order network and focuses on ternary relationships—a typical pattern of multi-technology interaction—to predict potential ternary technology convergence trends from the perspective of network edge dynamics. Specifically, we first construct a multimodal heterogeneous network based on patent information, linking "inventors–CPC–technical terms." Using a meta-path approach, we uncover latent connections between edges within the CPC co-occurrence network layer. Based on this, we form an expanded CPC co-occurrence network that includes all existing edges as well as newly identified links. Additionally, we develop a multidimensional technical similarity metric based on CPC classification codes, inventor information, and technical terms. Finally, we design three comparative experiments to systematically analyze, through machine learning models, the impact of network edge dynamics (i.e., the introduction of new edges) and multidimensional technical similarity metrics on technology convergence outcomes. Experiment 1 serves as the baseline control, predicting potential future ternary technology convergence using only the initial network state and the CPC-based technical similarity metric. Experiment 2 recalculates the CPC-based technical similarity metric within the expanded network and predicts ternary technology convergence, aiming to validate the improvement in prediction accuracy brought by the introduction of new edges. Experiment 3 builds upon Experiment 2 by further incorporating inventor similarity and technical term similarity, exploring the optimization of prediction performance through a multidimensional similarity metric system. The results indicate that introducing new edges into the CPC co-occurrence network used to characterize technology convergence significantly enhances the model's predictive performance and generalization capability. Furthermore, incorporating inventor similarity and terminology similarity metrics further optimizes the prediction outcomes. This study not only provides a novel methodological framework for predicting technology convergence but also offers important scientific evidence for deepening the understanding of the technology convergence process and optimizing technology convergence management practices.

ABSTRACT. Technology convergence is a critical driver of innovation in complex ecosystems like Digital Health. However, existing measurement approaches often rely on macro-level indicators (e.g., citations, IPC codes), failing to reveal the granular micro-mechanisms of how technologies combine at structural versus functional levels. To address this gap, this study proposes a novel analytical framework utilizing Large Language Models (LLMs) and Subject-Action-Object (SAO) semantic analysis. By employing the DeepSeek V3 model for precise knowledge extraction and Bertopic for topic modeling, we construct a “Dual-Perspective Network”—distinguishing between Structural Composition and Functional Interaction networks based on semantic action categorization. Applying this framework to 26,727 digital health patent families (2011-2024), we introduce a “Normalized Coreness Shift Index” to quantify evolutionary dynamics. Results reveal a striking divergence: the functional network evolves towards high complexity with “Super Topics,” while the structural network maintains a stable core. Specifically, “Sensor Module Design” exhibits a continuous core-evolution trend, whereas “Telemedicine Services” shows an infrastructure-sinking effect. These findings provide a high-resolution view of convergence, offering strategic insights for R&D decision-making.

ABSTRACT. To scientifically and effectively evaluate the transfer and transformation potential of patents, this study integrates relevant patent evaluation guidelines, suggestions from field experts, and data availability to enrich the existing patent transformation evaluation system. It innovatively incorporates semantic-dimensional indicators into the evaluation of patent technical value, designs 20 evaluation indicators from four dimensions (technical value, legal value, economic value, and applicant entity), and constructs a comprehensive evaluation system. Aiming at the imbalance of patent transformation samples, the Synthetic Minority Over-sampling Technique (SMOTE) is adopted to balance the categories. Two gradient boosting decision tree models, XGBoost and CatBoost, are used to perform classification tasks, and the SHAP method is applied to explain the importance of classification features. The empirical data consist of 139,273 authorized invention patents of Chinese institutions in Sichuan Province, retrieved from the Incopat Global Patent Search Database. The results show that transformed patents have slightly higher technical novelty and lower technical dependence; their average examination duration and average patent life are both longer than those of non-transformed patents; they also have a greater number of simple patent families, extended patent families, and more countries covered by the patent family, while having fewer patents under the same IPC main classification code. After oversampling, the CatBoost model demonstrates better performance, with an F1-score of 92.0%.

ABSTRACT. Against the backdrop of cross disciplinary technology integration becoming the core driving force of innovation, university patents are an important carrier reflecting the trend of technology integration. The existing patent research in universities has limitations, such as a single research perspective and insufficient exploration of inter domain correlations. To address these gaps, this study uses patents from Wuhan University of Technology from 2007 to 2025 as research samples, covering areas such as ships, materials, and others. It constructs a multidimensional intelligent analysis framework by integrating text mining, topic clustering, and social network analysis methods. The results indicate that patents exhibit overall cross disciplinary but locally clustered characteristics, with materials being the main core area; The turning point in the evolution of patent themes occurred in 2018, with research focus shifting towards green energy and data-driven approaches; Strategic emerging industry patents have high innovation potential. This study constructs a multi method integrated patent analysis framework, providing data support and decision-making references for universities to optimize research layout and promote technology transfer.

ABSTRACT. Against the backdrop of intensifying global technological competition and the deepening of the digital economy, early-stage tech startups serve as crucial engines of innovation. However, existing innovation potential evaluation systems face two core limitations: reliance on long-term operational data that startups lack and overgeneralized industry-agnostic indicators that fail to capture sector-specific innovation logic. To address these critical gaps, this study constructs a four-dimensional evaluation framework—integrating innovation accumulation, innovation team, innovation capital, and innovation ecology—grounded in the and innovation system theory. Tailored indicators are developed for four strategic emerging industries (semiconductors and integrated circuits, artificial intelligence, biopharmaceuticals, high-end medical devices) to enhance industry relevance. Using 20 AI startups as empirical samples, multi-source validation and case studies confirm the framework’s validity and strong discriminatory power. A key methodological contribution lies in prioritizing accessible "soft indicators" (e.g., core team background, patent quality, policy alignment) over traditional financial metrics, enabling effective capture of startups’ implicit technical value and team potential. This framework provides actionable tools for investors, policymakers, and enterprises to optimize innovation resource allocation, offering a targeted solution to the long-standing challenge of evaluating early-stage tech startups with limited data and heterogeneous industry characteristics.

ABSTRACT. Intellectual property securitization (IPS) has emerged as a novel financial instrument for mobilizing intangible assets and supporting innovation driven growth. Despite growing policy attention, systematic firm level evidence on the innovation incentives of IPS remains limited. This study exploits the introduction of local intellectual property backed securities as a quasi-natural experiment to examine the impact of IPS on firm innovation, using panel data on SMEs listed on China’s A share markets from 2015 to 2023. A difference in differences approach is employed to identify the causal effect of IPS on innovation outcomes. The results indicate that IPS significantly enhances firms’ innovation activity. Mechanism analyses further show that internal pressure weakens the innovation enhancing effect of IPS, whereas stronger external supervision substantially amplifies it. These findings suggest that although IPS can alleviate financing constraints by transforming future intellectual property income into tradable securities, its effectiveness depends critically on firms’ internal governance conditions and the strength of external monitoring. Additional heterogeneity analyses reveal that the innovation promoting effect of IPS is more pronounced for firms with fewer collateralizable tangible assets and stronger securitization capacity. Overall, this study provides new evidence on how intellectual property based financial instruments influence firm innovation behavior. It contributes to the literature on technology finance and innovation policy and offers policy relevant implications for improving intellectual property centered financial support systems.

ABSTRACT. Today, rapid technological change and the shortening of technology life cycles are strengthening the demand for systems that enable firms and nations to assess technology maturity in a timely and consistent manner. Technology Readiness Level (TRL) is used as an indicator to evaluate technology maturity in a staged manner; however, it relies heavily on expert judgment, which limits its reproducibility and scalability. In addition, patent-based indicators and bibliometric indicators also have difficulty adequately explaining the stage-to-stage transition processes of technology maturity and their interpretation. Therefore, to address these limitations, this study proposes a Retrieval-Augmented Generation (RAG)-based multi-agent framework that emulates expert panel–based TRL evaluation structures. The proposed framework consists of three stages. (1) Diverse technology-related documents, including academic papers, patents, corporate reports, and industry news, are integrated to construct role-specific RAG-based knowledge databases that reflect the entire technology life cycle. (2) Multiple expert agents representing the perspectives of basic research, technology development, system implementation, and the external environment independently evaluate TRL using role-specialized RAG structures based on the same technology information. Their evaluation results are then integrated through a consensus-based procedure to converge on a final TRL stage. (3) Based on the derived TRL results, the development period required for transitions between specific TRL stages is estimated and quantified as time delays in inter-firm technology gaps with respect to a common target TRL. By integrating automated TRL assessment with time-based technology gap analysis into a unified framework, this study presents a data-driven decision-support tool for technology strategy formulation, R&D investment prioritization, and inter-firm technology competitiveness comparison.

ABSTRACT. Technological leap is commonly regarded as a form of discontinuous technological advancement associated with positive developmental expectations. However, studies on its impact on subsequent domain development largely rely on qualitative discussions, with limited quantitative and empirical evidence in specific technological contexts. To address this gap, this study examines additive manufacturing technologies and develops a counterfactual analytical approach integrating propensity score matching and the synthetic control method (PSM-SCM). By estimating outcomes under a non-leapfrogging scenario, the study evaluates the impact of technological leapfrogging while accounting for endogenous technological evolution. The results show that: (1) technological leapfrogging has an overall positive effect on subsequent domain development. Without leapfrogging, the impact would decrease by an average of 88.33%, 60.89%, and 71.41% in breadth, depth, and persistence, respectively; (2) the effects exhibit clear type-based differences, categorized into eight types (T1–T8). T1 (72.94%) shows declines across all dimensions without leapfrogging, while T8 (8.24%) performs relatively better under the non-leapfrogging scenario. The remaining types (T2–T7, 18.82%) display differentiated patterns across dimensions. This study constructs a multidimensional measurement framework and proposes a PSM-SCM-based counterfactual approach, providing a methodological reference for evaluating the effects of technological evolution on domain development.

ABSTRACT. Identifying emerging technologies is inherently challenging due to their weak signals, small scale, fragmented distribution, and cross-domain characteristics. In patent-based analyses, conventional BERTopic models face limitations in domain-specific semantic representation, outlier detection, and automated topic labeling. This study proposes an optimized BERTopic modeling framework for emerging technology identification. The framework integrates patent-domain embeddings, a main–sub hierarchical modeling strategy for reclustering outlier documents, multi-source keyword generation with semantic ranking, and large language model–based topic labeling. Empirical validation using digital health patents from 2015 to 2024 demonstrates that the proposed framework significantly improves semantic coherence, topic coverage, and sensitivity to marginal and cross-domain emerging topics compared with standard BERTopic configurations. The results indicate that the optimized framework enhances the applicability of BERTopic in patent-based emerging technology analysis and provides a reusable methodological reference for technology evolution analysis and forward-looking assessment.