*5.1. A Transdisciplinary Approach*

Our proposed transdisciplinary approach integrates the following disciplines and/or research fields: transnational innovation ecosystems, universities' societal engagement, international relations (EU and China), Helix models of innovation, institutional theory, social network theory, social network analysis, professional social matching, ML, and KR and R. Our long-term goal is to design an AI-based system that can predict and match potential university and industry collaborators in transnational contexts, particularly matching European industrial firms with potential Chinese partner firms through their common connections to EU–China university cooperation, as demonstrated in the first two layers of boxes in Figure 5. The users of the system will be European SMEs. The users are expected to input their own professional network, including their collaborations with European university actors, on a voluntary basis, to the system to keep the database growing. Meanwhile the system will reward the users with suggested optimal industrial partners from China as well as the information concerning U–I co-innovation networks in between. Behind the user interface is the ML-based matching system. Along the lines in the second layer box, the dark grey rectangular boxes indicate examples of open data sources, used for ML.

**Figure 5.** Illustration of an artificial intelligence (AI)-based system for building transnational university and industry (U–I) co-innovation networks.

The core technologies are ML and KR and R (demonstrated in the third layer boxes in Figure 5). To train the computer to make autonomous prediction, training data from various sources, such as those examples in the grey boxes, will be gathered. Here the data are about existing links between TIC and TUC.

The integration of social sciences and AI is shown in the third- and fourth-layer boxes in Figure 5. The insights of social sciences theories and studies, as mentioned early in the paper, will be used for guiding empirical research on TIC and TUC synergy building. The possible methods are case studies,

Living Lab and survey. The findings (data) of the research will be the sources of validation and test data, which is important for improving and optimizing the ML algorithm. On the other hand, the models developed by the AI system will help enhance understandings of TIC and TUC interactions in social science studies.

While Figure 5 illustrates an overall view of our approach, we try to give a bit more details of the AI-based methods (or the technology core in the third layer box). Our modelling and prediction of TIC and TUC co-innovation networks are on the organizational level, though our data mining will process the data of individual organizational members. Our proposed approach is composed of three phases: (1) Data Preparation, (2) Data Analysis, and (3) TIC and TUC matching. The latter is the step wherein, based on the reasoning of semantic information, the hidden cooperation links are inferred. This section describes each phase in order to provide understanding on the interrelated techniques and parts of the approach, which is depicted in Figure 6.

**Figure 6.** Different phases of the approach for inferencing transnational industry cooperation (TIC) and transnational university cooperation (TUC) matching.

Data Preparation is the first step in understanding the data and transforming it into interpretable information. In fact, data preparation is divided into three subtasks: collection, homogenization and population of data. First, the data are collected from both open source data and different partners who are willing to cooperate within the network on a voluntary basis. To achieve this, future partners will have access to a platform wherein they can fill out profile information. This information includes aspects such as name and type of organization, interests, availability of human resources and so on. Additionally, existing databases (e.g., the CORDIS EU project databases, Web of Science) and other web resources, such as provided website links and social media, are also added for further data mining. ML-based techniques will be used to mine data from web links, aiming to finding patterns that can be later used to match information to other partners in the network. Second, the data are homogenized in the platform format, which in turn is compatible with the semantic model to be populated. In this context, as the semantic model is built within ontologies, the selected format is the Ontology Web Language (OWL), which can be queried by other parts of the platform within SPARQL Protocol and RDF Query Language (SPARQL) queries. On top of the OWL statement, the model includes a set of semantic rules that derive from inferences at the third stage of the proposed approach.

Once the data are prepared and thus included in the semantic model, the data are analyzed in the Data Analysis phase. There are two types of analysis in this phase. First, as stated previously, web sources are mined in order to find common patterns. Found patterns permit the creation of new statements that will extend the ontology. Furthermore, the semantic model consistency is validated within a semantic reasoner. In this approach, the Pellet reasoner is used for such a validation.

Finally, once the model is populated and validated, the third stage regards finding the potential cooperation between parties within the TIC and TUC matching phase. This is achieved within the inference of ontological descriptions enriched within semantic rules. More precisely, Semantic Web Rule Language (SWRL) rules are used, as they are compatible with Resource Description Framework (RDF)-based models and hence OWL ontologies. The finding of implicit relations between ontological resources within similar interaction of technologies and languages have been demonstrated on previous research work by the authors [125,126].

One of the benefits of this approach is that the described stages are automatically executed with only the need for the profiles of university and industry actors. The platform engines manage the tasks regarding the corresponding data and present the results, i.e., the potential cooperation, to the collaborative partnerships. This is provided to the platform owner as a report to be later shared with the platform users.

#### *5.2. Potential to Answer the Research Questions Using the Transdisciplinary Approach*

Once our proposed approach has been developed, it will better answer the three research questions raised at the outset of the paper. To make clear how our approach advances state-of-the-art research, we contrast the potential contributions of our proposed approach to the research questions and the limitations of existing literature (Table 1).


**Table 1.** Answers to research questions by existing literature and our proposed approach.

#### **6. Conclusions**

This paper has demonstrated how social sciences and AI could be integrated to develop a transdisciplinary approach to TIC and TUC synergy building, thus contributing to the knowledge pool in which studies on TIC and TUC have been separately reported in spite of a growing awareness on necessary of synergy building between TIC and TUC. Specifically, the originality of our research lies in four aspects.

First, while the research on TIC and TUC synergy building is almost an uncharted field, our research maps a landscape of the research area with identification of specific research gaps through extensive analysis of relevant literature. Our efforts are around discussions on how existing research may offer useful insights and have limits in answering three questions: Why should TIC and TUC be looked at as synergetic entities? How can the synergy building be theoretically elucidated? How can the synergy building be methodologically realized? The diagnose of advantages and weaknesses of state-of-the-art research in the field not only serves as clear point of departure of our research on TIC and TUC synergy building, but also helps guide more scholars to plunge into the field.

Second, we propose a transdisciplinary approach to TIC and TUC synergy building by integrating insights of social sciences, such as Helix Model of innovation, institutional theory and social network theory, and AI, such as ML and KR and R. While the insights from social sciences have the potential to answer the question of why TIC and TUC should be regarded as synergetic entities, AI technologies can help answer the question of how such synergies can be realized. Current methods using AI (i.e., ML) in social science research tend towards two extremes: they are used either for verifying assumptions about human intelligence or for independent prediction, thus tending to replace human intelligence. We try to integrate between social-science-based theoretical modelling and data-based computational modelling, particularly regarding the understanding of TIC–TUC interactions. Empirical findings guided by social sciences theories will help improve ML algorithm by providing validation and test data. The models learnt by machine will be useful input for advancing understanding of TIC and TUC synergy building in social sciences studies.

Third, our proposed approach will specifically identify/predict hidden/missing connections between actors in TIC and TUC co-innovation networks by analyzing unstructured data from various sources, such as public databases (e.g., research projects, co-publications, patents), website text (on cooperation activities) and auto-generated survey data. In doing so, on one hand we try to advance current social network analyses, which mainly map out anticipated networks by processing classified data [107]. On the other hand, we will open new horizons for studies on professional social matching [34] from both cross-sectoral and transnational perspectives.

Finally, our approach of AI-based matching system will help realize the potential role of university for institutional change and trust building, which are important to the sustainable dimension of innovation ecosystem development.

Nevertheless, our paper is primarily on the conceptual level discussions. Despite promising potential of our approach, it must be first tested and verified with sample data. This will be our next research task. It should also be noted that matching TIC and TUC, e.g., in the EU and China context, is just the first step to building EU–China transnational co-innovation networks. To achieve full synergy of the networks, there are also other important issues that need to be deeply explored, such as research and innovation policies, entrepreneurship, knowledge management, intellectual property rights, and inter-cultural communications. This will indeed open a new area of multidisciplinary research. Studying the TIC and TUC synergy building in the EU and China context also propels an urgent demand for closer communication between two separate research communities, namely international researchers and Chinese researchers both conducting research on the topic [127].

**Author Contributions:** Overall research design: Y.C.; Designing and describing AI-based approach: Y.C., B.R.F. and J.L.M.L.; Writing original draft: Y.C.; Extensive review and editing: B.R.F. and J.L.M.L.

#### **Funding:** The APC was funded by MDPI.

**Acknowledgments:** We are grateful to the four peer reviewers' valuable comments, which has greatly helped us to improve the paper. We also appreciate our colleagues Martti Juhola, Nina Helander, and Thomas Olsson, with whom we have been working together to prepare a research project proposal around the ideas addressed in the paper. Their insights in our discussions on planning the research proposal provide us inspirations for writing the paper.

**Conflicts of Interest:** The authors declare no conflict of Interest.
