In traditional innovation practices, concept and IP generation are often iteratively integrated. Both processes demand an intricate understanding of advanced technical domain knowledge. Existing large language models (LLMs), while possessing massive pre-trained knowledge, often fall short in the innovative concept generation due to a lack of specialized knowledge necessary for the generation. To bridge this critical gap, we propose a novel knowledge finetuning (KFT) framework to endow LLM-based AI with the ability to autonomously mine, understand, and apply domain-specific knowledge and concepts for invention generation, i.e., concept and patent generation together. Our proposed PatentGPT integrates knowledge injection pre-training (KPT), domain-specific supervised finetuning (SFT), and reinforcement learning from human feedback (RLHF). Extensive evaluation shows that PatentGPT significantly outperforms the state-of-the-art models on patent-related benchmark tests. Our method not only provides new insights into data-driven innovation but also paves a new path to fine-tune LLMs for applications in the context of technology. We also discuss the managerial and policy implications of AI-generating inventions in the future.

Accurate interpretation of scientific funding policies is crucial for government funding agencies and research institutions to make informed decisions and allocate research funds effectively. However, current large language model (LLM)-based systems often generate responses without references, leading to a lack of interpretability needed for policy enforcement. This study introduces the Adaptive Two-stage Retrieval Augmented Fine-Tuning (AT-RAFT) method, a novel LLM-based approach specifically designed for science policy interpretation. AT-RAFT incorporates three complementary artifacts: a two-stage retrieval mechanism, adaptive hard-negative fine-tuning, and an interpretable response interface. It is trained directly on policy documents, allowing the model to provide reference answers based on retrieved text while also offering the original policy context to enhance interpretability. Our experiments demonstrate that AT-RAFT improves retrieval accuracy by 48% and generation performance by 44% compared to existing baseline systems, effectively supporting real-world decision-making tasks for stakeholders in research institutions and funding agencies. Our proposed method has been adopted by ScholarMate , the largest professional research social networking platform in China, and is now deployed on their platform, providing global users with access to advanced policy interpretation tools. Additionally, a demo version of the instantiated interface is available at https://github.com/renruntao/ResearchPolicy_RAG .

Research project evaluation upon completion is one of the important tasks for research management in government funding agencies and research institutions. Due to the increased number of funded projects, it is hard to find qualified reviewers in the same research disciplines. This paper proposes a machine learning and large language model integrated approach to provide decision support for research project evaluation. Machine learning algorithms are proposed to compute the weights of key performance indicators (KPIs) and scores of KPIs based on the evaluation results of completed projects, large language models are used to summarize research contributions or findings on project reports. Then domain experts are invited to consolidate the weights and scores for the KPIs and assess the novelty and impact of research contribution or findings. Experiments have been conducted in practical settings and the results have shown that the proposed method can greatly improve research management efficiency and provide more consistent evaluation results on funded research projects.

Abstract Academic collaboration plays an important role in undergraduate research. Current methods rely on offline social contacts for undergraduate students to collaborate with academic staff members in universities and research institutions. In big data era, it is difficult for undergraduate students to find suitable research project opportunities and supervisors to work with. This paper proposes a social recommendation system for undergraduate students to find research project opportunities and work with research project teams on an academic collaboration network. The proposed recommendation method integrates relevance, connectivity, and quality modules, where profiles of undergraduates are constructed with their self‐claimed information, research activities (e.g., studying and reading research publications and reading research projects), and social connections in the academic collaboration network. Suitable research projects are recommended based on the undergraduates' profiles. Experiments are conducted, and the results have shown that the proposed social recommendation system can facilitate undergraduates' selection of research projects.

Expert recommendation plays a vital role in the expansion of researchers’ academic communities and in the creation of potential collaboration opportunities. Current approaches for academic expert recommendation are mainly based on keywords-based research relevance and social network proximity between researchers. However, most proximity measures only focus on the individual level in the network. Therefore, we develop a new measure for the proximity at the institutional level that measures the link strength between researchers’ affiliated institutions. Moreover, a multilevel profile-based approach is proposed to identify the most suitable expert for research collaboration by integrating research relevance information, individual social network information and institutional connectivity information. The proposed approach has been implemented in ScholarMate, which is a research 2.0 innovation, promoting knowledge-sharing activities in the virtual scientific community. According to the results of the experiments conducted on the real-world dataset, institutional connectivity is proved to be an important factor for expert recommendation and the proposed hybrid method outperforms all the other benchmark algorithms significantly.

With the increasing pressure on researchers to produce scientifically rigorous and relevant research, researchers need to find suitable publication outlets with the highest value and visibility for their manuscripts. Traditional approaches for discovering publication outlets mainly focus on manually matching research relevance in terms of keywords as well as comparing journal qualities, but other research‐relevant information such as social connections, publication rewards, and productivity of authors are largely ignored. To assist in identifying effective publication outlets and to support effective journal recommendations for manuscripts, a three‐dimensional profile‐boosted research analytics framework (RAF) that holistically considers relevance, connectivity, and productivity is proposed. To demonstrate the usability of the proposed framework, a prototype system was implemented using the ScholarMate research social network platform. Evaluation results show that the proposed RAF‐based approach outperforms traditional recommendation techniques that can be applied to journal recommendations in terms of quality and performance. This research is the first attempt to provide an integrated framework for effective recommendation in the context of scientific item recommendation.

Research project selection is an important task for government and private research funding agencies. When a large number of research proposals are received, it is common to group them according to their similarities in research disciplines. The grouped proposals are then assigned to the appropriate experts for peer review. Current methods for grouping proposals are based on manual matching of similar research discipline areas and/or keywords. However, the exact research discipline areas of the proposals cannot often be accurately designated by the applicants due to their subjective views and possible misinterpretations. Therefore, rich information in the proposals' full text can be used effectively. Text-mining methods have been proposed to solve the problem by automatically classifying text documents, mainly in English. However, these methods have limitations when dealing with non-English language texts, e.g., Chinese research proposals. This paper presents a novel ontology-based text-mining approach to cluster research proposals based on their similarities in research areas. The method is efficient and effective for clustering research proposals with both English and Chinese texts. The method also includes an optimization model that considers applicants' characteristics for balancing proposals by geographical regions. The proposed method is tested and validated based on the selection process at the National Natural Science Foundation of China. The results can also be used to improve the efficiency and effectiveness of research project selection processes in other government and private research funding agencies.

Both similarity-based and popularity-based document ranking functions have been successfully applied to information retrieval (IR) in general. However, the dimension of semantic granularity also should be considered for effective retrieval. In this article, we propose a semantic granularity-based IR model that takes into account the three dimensions, namely similarity, popularity, and semantic granularity, to improve domain-specific search. In particular, a concept-based computational model is developed to estimate the semantic granularity of documents with reference to a domain ontology. Semantic granularity refers to the levels of semantic detail carried by an information item. The results of our benchmark experiments confirm that the proposed semantic granularity based IR model performs significantly better than the similarity-based baseline in both a bio-medical and an agricultural domain. In addition, a series of user-oriented studies reveal that the proposed document ranking functions resemble the implicit ranking functions exercised by humans. The perceived relevance of the documents delivered by the granularity-based IR system is significantly higher than that produced by a popular search engine for a number of domain-specific search tasks. To the best of our knowledge, this is the first study regarding the application of semantic granularity to enhance domain-specific IR.

In the first decade of the 21st century, China’s Research Community (CRC) is struggling to achieve better performance by increasing growth in knowledge quantity (e.g., publications), but has failed to generate sound growth in knowledge quality (e.g., citations). An innovative E-government project, Internet-based Science Information System (ISIS), was applied nationwide in 2003 with a variety of embedded incentives. The system has been well received and supports the National Natural Science Foundation of China (NSFC) to implement managerial control to cope with pressing demands relating to China’s research productivity. This paper explores the impact of Information Systems (IS) from the perspective of agency theory based on CRC empirical results. Since the nationwide application of ISIS in 2003, CRC outcomes have markedly improved. The discussion and directions for future research examine implications of IS for E-government implementation and business environment building in developing countries.

Multiperson decision making (MPDM) problems with different formats of preference information are one of the emerging research areas in decision analysis. Existing approaches for dealing with different preference formats tend to be unwieldy. This paper proposes a new method to solve the problem, in which the preference information on alternatives provided by experts can be represented in four different formats, namely: 1) utility values; 2) preference orderings; 3) multiplicative preference relations; and 4) fuzzy preference relations. An optimization model is constructed to integrate the four formats of preference and to assess ranking values of alternatives. The model is shown to be theoretically sound and complete via a series of theorems, and then a corresponding algorithm is developed. A numerical example is given to illustrate the procedure. The proposed approach is more efficient and simpler than existing approaches because it does not need to unify different formats of preferences or to aggregate individual preferences into a collective one. Therefore, it overcomes a major shortcoming of existing approaches that lose or distort the original preference information in the process of unifying the formats

Group decision-making methods have been developed extensively, but their adaptation for use in organizations has been problematic. According to Arrow's (1963) Impossibility theorem, one conceivable reason is that a group decision outcome could never satisfy all decision makers' individual preferences. In order to accommodate individual decision maker's preferences in a group decision-making task, this correspondence presents a fuzzy multiperson multicriteria decision making (MMCDM) model and a structured group decision-making process. The fuzzy MMCDM model includes fuzzy individual preference generation and group preference aggregation. The structured decision-making process keeps the group interaction on track, so that the fuzzy MMCDM model can be effectively applied to the group decision-making task. Based on the proposed model and the decision-making process, a fuzzy group support system (GSS) has been developed and applied to a group assessment task. An empirical study was conducted and the experiment results showed that use of the fuzzy GSS enhanced individual understanding, consensus, and satisfaction of the group decision outcome.

Based on the additive multi-attribute value model for multiple attribute decision making (MADM) problems, the paper investigates how the set of attribute weights (or weight-set thereafter) is determined according to the preference orders of alternatives given by decision makers. The weight-set is a bounded convex polyhedron and can be written as a convex combination of the extreme points. We give the sufficient and necessary conditions for the weight-set to be not empty and present the structures of the weight-set for satisfying the preference orders of alternatives. A method is also proposed to determine the weight-set. The structure of the weight-set is used to determine the interval of weights for every attribute in the decision analysis and to judge whether there exists a positive weight in the weight-set. The research results are applied to several MADM problems such as the geometric additive multi-attribute value model and the MADM problem with cone structure.

Decision models play an important role in decision-making, and supporting model-building is one of the most important functions of model management in decision support systems. As concepts at different abstract levels have to be used in the process of model-building, representing these concepts in a coherent way has been recognized as a key research topic. In this paper, a model-building framework is proposed which integrates object-oriented analysis with action logic as the representation tool. This model-building framework can provide representations for concepts at different abstract levels and can describe the process of abstracting decision models from decision situations or problems represented in lower abstract level concepts.

The current trend in EDI implementation is moving away from traditional networks, Value Added Networks (VANs), to the Internet, because the Internet can provide lower network service charges, larger size of user groups, and more available network services. Internet EDI is now playing an important role in electronic business and/or electronic commerce applications. One of the most important components of Internet EDI is the EDI software. The paper proposes a component based approach to the development of EDI software for Internet EDI. The EDI software is developed to support the multiple document standards for government funded R&D project management.

This paper presents dynamic output feedback controller design for fuzzy dynamic systems. Three kinds of controller design methods are proposed based on a smooth Lyapunov function or a piecewise smooth Lyapunov function. The controller design involves solving a set of linear matrix inequalities (LMI's) and the control laws are numerically tractable via LMI techniques. The global stability of the closed-loop fuzzy control system is also established.

An adaptive fuzzy system implemented within the framework of neural network is proposed. The integration of the fuzzy system into a neural network enables the new fuzzy system to have learning and adaptive capabilities. The proposed fuzzy neural network can locate its rules and optimize its membership functions by competitive learning, Kalman filter algorithm and extended Kalman filter algorithms. A key feature of the new architecture is that a high dimensional fuzzy system can be implemented with fewer number of rules than the Takagi-Sugeno fuzzy systems. A number of simulations are presented to demonstrate the performance of the proposed system including modeling nonlinear function, operator's control of chemical plant, stock prices and bioreactor (multioutput dynamical system).