Enhanced Systematic Idea Generation Methodologies using Generative AI
- A Case Study on Taekwondo Systems -
2025 年 17 巻 2 号 p. 13-25
詳細
2025 年 17 巻 2 号 p. 13-25
Intensified modernization within Taekwondo (TKD) systems is imperative due to recent advancements in technological innovation, particularly in artificial intelligence (AI)-driven training systems and smart wearable devices within the sports sector. Nevertheless, the evolution poses considerable challenges for TKD-focused social enterprises (SEs), which frequently operate under limited financial and human resources. This study proposes a strategic framework for generating novel design concepts for future TKD systems. Three ideation methodologies are integrated in the framework: the novelty potential/void matrix, a design catalogue informed by cooperative patent classification (CPC) codes, and generative artificial intelligence (Gen AI). Concepts with high novelty potential and minimal technological saturation (void) are identified through patent abstracts analysis before generating novel conceptual combinations with ChatGPT. Consequently, an innovative design concept consisting of a smart scoring system enhanced by an emotion-responsive platform is produced. Based on comparative evaluation data, the proposed ideation approach yields concepts with superior originality and no overlap with LLM-generated ideas, suggesting a more substantial potential for first-mover advantage than ideas generated solely through LLM models. Although the differences were not statistically significant (p>0.05), the trend still suggests that the proposed method may offer higher novelty compared to LLM-generated ideas. A detailed assessment of relevant CPC codes is also conducted to validate the novelty of the proposed concepts. The results confirm the absence of analogous patents, underscoring the uniqueness of the identified combinations for TKD SEs. Furthermore, semi-structured interviews with representatives from the Malaysian Taekwondo Federation and clubs are performed, procuring expert insights and feedback on the design catalogue. Additionally, the basketball example is presented as an exploratory application rather than a full second case study. Conclusively, this study presents a novel ideation methodology that can enhance the innovation capacity and competitive positioning of TKD SEs, aligning with broader objectives related to the Sustainable Development Goals (SDGs).
The global sports technology industry is experiencing rapid advances in new technologies, including wearable devices, artificial intelligence (AI), immersive augmented reality (AR)/virtual reality (VR) and sustainability-focused innovations. The market size is expected to reach 60 million USD by 2030, increasing at a compound annual growth rate (CAGR) of more than 20% [1]. The surge demonstrates the increasing demand for data-driven performance tools, digital systems and innovative equipment across a wide range of sports domains.
Originating in Korea, Taekwondo (TKD) is a globally recognized martial art that has evolved as an official Olympic sport. The sport is also popular among individuals across all ages, recording over 80 million practitioners in over 200 countries [2]. TKD was chosen as a representative case due to its rapid growth in patent filings, technological innovation, and strong social significance. Based on the number of patents, TKD demonstrates an increasing growth trend superior to other sports, including Judo and Karate, reflecting increasing technological innovation [3].
Furthermore, TKD contributes significantly to the athletes’ personal well-being through both physical fitness and mental activities, including reducing mental health disorders and enhancing self-defence ability. Post-COVID-19, the research in TKD has shifted towards smart, interactive, and artificial intelligence (AI)-driven systems to improve accuracy and enhance training performance. Behavior-based AI learning support systems [4] and intelligent recognition systems based on deep-vision learning for dynamic sports gestures [5] are among recent technological innovations in TKD. Consequently, TKD possesses a substantial potential for business expansion, integration of new technologies, commercialization and significant social contribution.
Nevertheless, numerous TKD organizations operate as non-profit organizations (NPOs) and face challenges with limited resources and funding, restraining innovation capabilities in developing new and advanced systems despite technological advancements and innovations. In the current digital transformation (DX) era, relying on one system is not feasible. Consequently, an integrated system for TKD social enterprises (SEs) based on the perspective of system of systems (SoS) [6, 7] is crucial. Leveraging interconnected subsystems also enables TKD SEs to achieve higher innovation potential, competitiveness, and long-term sustainability than isolated technological solutions.
This study proposed a novel approach to increase innovation in TKD SEs by creating novel design concepts for future TKD systems. Idea generation methodologies were employed, including the novelty potential/void matrix, a design catalogue based on cooperative patent classification (CPC) codes, and generative artificial intelligence (Gen AI). The proposed ideas were then benchmarked with large language model (LLM)-generated ideas evaluated by independent experts and CPC codes for verification.
Semi-structured interviews with the Malaysian TKD Federation and clubs were also conducted to obtain practical insights into innovation challenges and feedback on the proposed design catalogue. This study focused on a single case (TKD) for initial validation, which is a reasonable approach to confirm the effectiveness of the proposed design methodology. Additionally, the use of CPC codes and abstracted design knowledge enables potential application to other domains.
Although several idea-generation methods are available, only focusing on traditional approaches, such as brainstorming and mind mapping [8, 9], is insufficient and presents challenges to generating new and high-quality ideas. According to the novelty potential concept, combining different abstract concepts can lead to novel design concepts. Analyzing patent documents and extracting abstract concepts may also result in the identification of unexplored high-novelty combinations.
Furthermore, opportunities for innovation can be highlighted when combined with the void matrix to identify areas without existing patents. Based on the concept, Tachibana et al. generated novel ideas by combining sports and blockchain patents [10]. The ideas were evaluated with lexical and distributional word similarity, such as Word2vec. Syamimi et al. also produced new business ideas to create an advanced video platform integrating blockchain technologies for TKD SEs, utilising a similar concept [11].
2.2 Design CatalogueRoth produced the design catalogue, a powerful tool for generating design concepts. The catalogue also offers a systematic way for exploring design possibilities [12]. Furthermore, the design catalogue permits organizing collections of design knowledge, solutions, and components, providing a foundation for generating new concepts that can then be combined and adapted to address new design challenges [12].
In a design catalogue, each row represents a pair of issues and solutions. Multiple solutions for a single problem are also summarised, allowing comparisons and consideration of various solutions to identical issues. Elements, such as sketches, examples, and conditional expressions, are utilized to express the solutions.
A design catalogue is particularly useful in the early stages of design, aiming to explore a wide range of ideas and identify potential concepts. Nonetheless, the content of most design catalogues is limited to the mechanical engineering sector, and their versatility is finite. Maintaining and updating the manually developed content is also time-consuming and requires considerable effort. Furthermore, designers face challenges when finding solutions due to unclear classification criteria [13]. Accordingly, employing classification systems, including CPC and patent-based catalogues, enables cross-industry knowledge transfer.
2.3 Gen AIFoundation model-based Gen AI, including ChatGPT, has been widely applied to support design concept development and automate idea generation. For instance, ChatGPT is employed as a concept generation tool that aids in improving the number and novelty of the generated concepts [14]. The tool also enhances the TRIZ methodology, facilitating the ideation process in eco-design by generating guidelines based on the latest technical advancements from patents [15]. Moreover, ChatGPT can generate several ideas and adjust them based on the input provided.
Among the limitations of utilising ChatGPT is its ineffectiveness compared to traditional design methods [14]. The complex prompt engineering may also pose challenges to the design process in human-computer interactions [15]. Although ChatGPT supports human capabilities, it cannot substitute human design thinkers and human-generated insights [16]. Consequently, depending solely on ChatGPT to generate novel ideas may be inadequate. Viewing AI as a complementary tool within the design process in collaboration with human creativity and AI-generated ideas, is essential. Human oversight in AI-assisted design processes is also vital to ensure creativity, feasibility, relevance, and alignment with real-world applications.
2.4 Research ApproachAlthough numerous ideation tools are available, few studies have specifically addressed developing novel design concepts for TKD SEs. Consequently, this study suggested a comprehensive strategy that integrates multiple ideation methodologies, namely the novelty potential/void matrix, CPC-based design catalogue, and Gen AI, such as ChatGPT.
This study considered ChatGPT as an enhancement layer and not a standalone ideation tool, refining ideas generated from structured frameworks, including novelty potential, void matrix, and design catalogue. The integrative approach enabled synthesising diverse ideation techniques, facilitating innovative design concept generations tailored to the unique requirements and constraints of TKD SEs.
This study utilizes idea generation methodologies, novelty potential/void matrix, design catalogue based on CPC codes, and Gen AI to support novel design concepts creation in the early stages of product development. The approach is grounded in three complementary theories: combinational creativity, absorptive capacity and design thinking, which together create a robust theoretical foundation for knowledge-driven innovation.
Creativity in the human mind can be generated through three approaches, including exploratory creativity, transformational creativity and combinational creativity [17]. Among these, combinational creativity is considered the easiest approach to creativity by combining ideas that were previously unrelated or indirectly related. Such combinations can be derived through different elements such as concepts, ideas, words, sounds and images [17].
Absorptive capacity theory defines a firm’s ability to utilize external knowledge through a sequential process. The process enables firms to identify the value of external knowledge, incorporate it through exploratory learning and apply it to generate new knowledge and value [18]. Finally, design thinking focuses on a human-centered approach, which emphasizes empathy and prioritizes user needs to create meaningful solutions [18]. Therefore, relying on a single idea generation method may be insufficient, as it limits the capabilities to produce novel outputs. Table 1 summarizes the theories’ relation to the study’s approach.
| Theory | Definition | Relation to approach |
|---|---|---|
| Combinational creativity | Generating novelty by associating ideas that were previously unrelated or indirectly related | Combining unrelated patents such as TKD and audience behavior to create new design ideas |
| Absorptive capacity | The organization’s ability to identify, acquire, assimilate and exploit external knowledge | Leveraging on external patent information such as CPC codes to recombine ideas |
| Design thinking | Human-centered problem framing to recognize most vital aspects of problems and creating solutions from potential solutions | Applying human evaluation and empathy to address TKD challenges in selecting possible solutions generated by GenAI |
Firstly, the design catalogue method, which utilizes patent data as information sources, was utilized to extract the purpose-means pairs known as design knowledge [19]. A retrieval approach was then suggested with the design catalogue to generate a design concept [19]. The design catalogue creation process consisted of patent data collection, design knowledge extraction from the patent data, and clustering design knowledge pieces into abstract purposes and means. In this study, P denotes purpose, M represents means, and F is the design problem.
3.1 Patent Data CollectionA keyword search in the field of “Title” OR “Abstract” OR “Claim” enabled patent data collection from patent databases involved in this study. Valuable information were extracted, such as titles, abstracts, application year, and CPC codes, from 2000 to 2019. In this study, “Audience Behavior (ab)” and “Taekwondo (tk)” were considered to be related to “digital technology” as keywords that are relatively close to design problems.
Furthermore, “Digital Transformation (dx)” and “Waste Management for Public Health (wm)” were selected to drive innovative design concepts, considering their growing importance driven by the recent emphasis on promoting sustainable development goals (SDGs). The keywords applied are as follows:
Following patent data collection, design knowledge pieces described in the patent from the title and summary of patent data were obtained with the feature OpenAI in GPT-4 [20]. The model can perform various tasks through few-shot learning, carefully leveraging constructed prompts with relevant examples and instructions rather than focusing on reinforcement learning. This study also constructed a prompt that included an instruction outlining the task of extracting P–M pairs, an example sentence illustrating the desired output format, and the title and summary of the target patent. Furthermore, the large language model (LLM) was instructed to include the original text segment (“Source”) from each design knowledge piece to mitigate hallucination, where the model might generate inaccurate or unsupported design knowledge. The step allowed human evaluators to assess the validity and traceability of the extracted information.
3.3 Clustering Design KnowledgeThe extracted P and M in design knowledge pieces in this study were clustered as follows:
This study applied Ward’s method, an agglomerative hierarchical clustering technique, in Steps 2 and 3. The classification was determined by the cosine distance between the distributed representations expressed in Equation (1).
 |
A design catalogue offers a designer various alternative Ms corresponding to the F by retrieving design knowledge with an abstract purpose. In this study, the following steps were performed according to design problem similarities:
In this study, representative concepts were evaluated with text mining co-occurrence network analysis based on the four extracted data types (p, pm, p-m, and m) to determine the most suitable type for assessing the novelty potential/void matrix for idea generation. Combinations with long distances between concepts were chosen during the new design concept combinations selection, indicating considerable novelty and zero void, which reflected no similar patents. Following combinations selection from the novelty potential/void matrix, a new design catalogue was generated according to CPC codes, yielding suitable means. The procedure was performed by referring to the original patent information collected, ensuring consistency and accuracy.
The CPC scheme is a hierarchical system that permits patent searches and analyses by categorizing inventions according to their technical domains, a more granular and detailed extension of the international patent classification (IPC) system [21]. Consequently, the CPC framework is particularly well-suited for identifying design cases and facilitating computational analysis, as patents are systematically classified by technical fields, including those defined by the IPC [19].
3.6 Generation and Verification of New Design ConceptIn this study, ChatGPT was provided with the new design catalogue to recommend means (solutions) based on technological trends (CPC codes). Based on the suggestion, an example of CPC combination for TKD SEs was logically selected. ChatGPT was then employed to refine the new idea combinations. To further illustrate the adaptability of the framework, an additional exploratory example was conducted in another sports domain, basketball.
The proposed method ideas were then compared and benchmarked with the ideas generated by LLM models, including ChatGPT, Gemini and DeepSeek. All ideas were independently evaluated by expert reviewers based on three variables:
A 5-point Likert scale (1 - very low; 5 - very high) was used to score each idea. In order to minimize bias, evaluations were conducted blindly with the order randomized. A total of eight evaluators from Osaka University participated, specializing in the mechanical and applied engineering department. The evaluators included one associate professor, three PhD students, one master’s student and three undergraduate students. The results were analyzed using the Kruskal-Wallis H Test, suitable for comparing more than two independent variables under non-normal distribution of data.
A CPC code investigation was also performed to verify the proposed novel ideas. Furthermore, semi-structured interviews were conducted through online and face-to-face meetings with the representatives of the Malaysian TKD federation and clubs. Finally, a new system for future TKD SEs was developed. The outcomes of applying the proposed methodologies to generate novel design concepts are discussed in subsequent sections of this article.
This study selected the most suitable and practical approach for idea generation in TKD, utilizing the extracted data as concepts (p-type, pm-type, m-type, and p-m-type) to determine the novelty potential/void matrix concept. The objective was achieved by analyzing the contents utilizing text-mining of the co-occurrence network analysis. During this step, potential relationships of the particular node (concept or word) were identified. The four types are defined as follows:
Based on the results, p-type (2080) and pm-type (2775) had higher overall concept combinations than p-m-type (1596) and m-type (1830), potentially indicating the superior novelty of the generated ideas. A higher number of combinations also suggested a greater diversity of concept pairings, increasing the likelihood of uncovering unique or less conventional ideas. An elevated purpose alignment was recorded by p-type, as it generated ideas focusing on a clear purpose or goal as demonstrated in Figure 1.
Figure 1 The P-type in the text mining
For example, the cluster contained words directly applicable to TKD, such as improve, efficiency, and training, thereby supporting the selection of combinations with a strong purpose alignment and a relevant business case for performance enhancement in TKD training. Meanwhile, Figure 2 illustrates the pm-type cluster, where purpose (“what”) and means (“how”) were semantically linked through existing or new methods.
Figure 2 The Pm-type in the text mining
This supports the selection of ideas with a clear business purpose and the means to achieve it. The approach yielded words, including improve, accuracy, and real-time, which were considerably connected to words, such as feedback and video. Consequently, the idea included the purpose (to improve real-time accuracy) and means (employing bio-feedback game device or video analysis).
The p-m-type and m-type ideas might have possessed inadequate purpose alignment with business goals, rendering them less suitable for business creation. The p-m-type might also have resulted from disjointed ideas that were challenging to implement due to purpose and means separation. For example, a cluster presented words (enhance, user) that focus on purpose, while another cluster had words (processing, signal, circuit) focusing on means.
Methods were the primary focus of the m-type instead of a clear purpose or goals. An example includes a cluster with words, such as intelligent, training, and device, whereas another cluster documents system and method. Consequently, the idea combination was solely based on methods, which might be technologically innovative. Nevertheless, the idea combination presented insufficient straightforward application or purpose to the TKD business creation, potentially rendering the implementation of the ideas more challenging.
Based on the evaluations, this study applied the extracted data as concepts for “purpose” (p-type) as an example for TKD. In the p-type, 64 clusters and representative concepts were procured, of which 18 clusters were related to the TKD category identified in the analysis of 633 TKD patents. Nonetheless, the term “taekwondo training” was deemed too general, hence excluded during assessments.
This study chose idea combinations that included one or both TKD categories to enhance their relevance to TKD business creation before sorting them by novelty potential in descending order to gain some hints. The combinations were then evaluated from the top, which indicated the highest novelty potential level. Finally, combinations with a 0 void were selected. As a result, the combination of “secures the helmet to the wearer’s head (tk)” and “analyzes audience behavior (ab)”, which has novelty potential and void value: (0.279574, 0) was obtained.
Novelty potential represents the cosine distance between two concept vectors. A high cosine distance (close to 1) between concepts reflects a high novelty potential. A 0.279574 novelty potential value denotes a moderate level of novelty. Conversely, void facilitates identifying areas without existing patents, revealing potential innovation gaps. Consequently, a 0 void value denotes that existing patents do not cover the combination. The following section demonstrates the application of the design catalogue, structured around CPC codes, to identify appropriate mechanisms from the selected combination of conceptual elements.
4.2 Design Catalogue Based on Technological Trends (CPC Codes)Post-selecting concept combinations “secure the helmet” and “analyze audience behavior”, the means (solutions) were determined from the design catalogue, yielding new design concepts. Nevertheless, establishing and selecting a suitable combination of means was challenging due to the vast amount of patent information. As a solution, each design concept was classified with CPC codes in the design catalogue. The original patent information collected was also referred to ensure accuracy and consistency.
Most CPC codes (sub group classified) in the “secure the helmet to the wearer’s head (tk)” were related to A63B71/00 (games or sports accessories), followed by A63B71/10 (for the head), A61F5/00 (orthopaedic methods or devices for non-surgical treatment of bones or joint and B01D53/00 (separation of gases or vapours). Other CPC codes were related to A63B group, including A63B71/06 (scoring), A63B2225/50 (wireless data transmission) and A63B2220/53 (force of an impact, e.g. blow or punch) reflect the emerging integration of sensors and protective, safety-related technologies in TKD patents.
The top CPC codes for “analyzes audience behavior (ab),” was H04N21/00 (selective content distribution, e.g., interactive television or video on demand), followed by G06Q30/00 (commerce), A61B51/63 (by tracking eye movement, gaze and pupil change) and G06F17/00 (digital computing). The CPC codes demonstrated the importance of audience engagement, digital platforms and biometric feedback in audience behavior patents. The data are illustrated in Figures 3 and 4.
Figure 3 The top CPC codes in “secure the helmet to the wearer’s head”
Figure 4 The top CPC codes in “analyze audience behavior”
This study employed ChatGPT to suggest combinations based on the technological trends (CPC codes) classified in the design catalogue by employing the role–task–format (R–T–F) prompting approach. ChatGPT was assigned a specific role (‘expert in design concept creation using patents), provided with a task (to generate design concepts for TKD), and constrained to a specific output format (table). The structured prompting approach allowed consistent benchmarking and reproducibility across generated outputs.
An example of the prompt applied was “You are an expert in design concept creation using patents. This is a design catalogue to generate new ideas for taekwondo based on 2 concept combinations: (secure the helmet to the wearer’s head) and (analyze audience behavior). Based on the combination of CPC trends and means, please suggest a new design in table format (attach the design catalogue).”
Figure 5 presents the generation of 3,969 combinations of helmet and audience behavior, systematically categorized based on CPC codes, purpose (the intended function), means (the mechanism to achieve the function) and new design ideas. The integration of the helmet and audience behavior demonstrates how different CPC-based combinations can lead to innovative design concepts. In order to address the critical challenge of reducing scoring bias and improving accuracy in TKD trainings and competitions, the A63B71/06 (scoring), G06N20/00 (machine learning), and G06V20/41 (semantic clustering) CPC code combinations were chosen in this study. Table 2 summarises the findings.
Figure 5 An example of the CPC trend combinations generated by ChatGPT
| CPC codes | CPC description | Design knowledge identification (ID) | Means (M) |
|---|---|---|---|
| Secure the helmet to the wearer’s head | |||
| A63B71/06 | Indicating or scoring devices for games or players, or other sports activities | tk-621-7 | Employ the signal transmitted to the electronic device |
| Analyze audience behavior | |||
| G06N20/00 G06V20/41 |
|
ab-3-5
ab-3-6 |
|
Subsequently, ChatGPT was employed to refine the combinations, generating novel design concepts utilizing a similar prompt. “You are an expert in generating novel business ideas. Please generate new design ideas by combining all CPC codes and means for taekwondo in table format.
Means: use the signal transmitted to the electronic device.
Means: based on the audience video data for one or more audience members for each performance segment.
Three design concepts leveraging CPC codes were obtained: a smart scoring and emotion-driven system, an interactive training and evaluation platform, and a live event engagement and dynamic entertainment system.
Figure 6 illustrates new design concepts generated in basketball, such as a smart real-time scoring hoop and an interactive fan-driven scoreboard, specifically using the most promising CPC code combinations identified in TKD by prompting ChatGPT. The basketball example demonstrates exploratory transferability and is not a fully generalized case, whereby it did not involve new patent data collection and novelty potential/void matrix analysis.
Figure 6 An example of ChatGPT ideas for basketball
An example of the prompt applied was “You are an expert in generating novel business ideas. Based on the combination of CPC trends and means, please suggest new design concepts for basketball in table format.”
Means: use the signal transmitted to the electronic device.
Means: based on the audience video data for one or more audience members for each performance segment.
Means: for each performance segment to determine an emotion accuracy score.”
By extending the framework to basketball, it demonstrates that the methodology can generate meaningful concepts for other sports or sectors beyond TKD. The ideas were then verified by comparing them with only LLM-generated ideas evaluated by independent experts in the subsequent section.
4.3 Comparison of Novelty Potential / Void Matrix and Design Catalogue Ideas with LLM ModelsThe methods proposed in this study were compared and benchmarked by employing the novelty potential/void matrix and design catalogue against LLM modelsgenerated ideas such as ChatGPT, Gemini and DeepSeek. The LLM models’ ideas were generated by prompting “You are an expert in design concept creation, please suggest new design concepts for martial arts such as taekwondo and potential use for other sports or sectors in table format.” The summarized comparison of ideas is demonstrated in Table 3.
| ChatGPT | Gemini | DeepSeek | Novelty Potential/Void Matrix Idea |
|---|---|---|---|
| Digital Systems | |||
|
None |
|
|
| Apparel/Sportswear | |||
|
|
|
None |
| Facilities | |||
|
|
|
None |
| Training equipment | |||
|
|
None | None |
According to the results, ChatGPT produced the highest number of ideas (10) with the most diversity such as an AI-powered sparring assistant and biometric belt system compared to Gemini and DeepSeek (4). However, LLM-generated ideas tended to converge on mainstream concepts such as AR dojangs and smart uniforms, indicating lower domain novelty. Additionally, the ideas were not prioritized or ranked according to structured criteria, including novelty, and were significantly dependent on the quality of the prompt.
In contrast, the proposed method yielded ideas with no overlap, such as emotion-driven scoring systems, suggesting its strength in niche innovation. A more systematic evaluation in a structured, quantifiable manner is possible. The attribute facilitates prioritizing ideas based on clear criteria, which is beneficial when involving multiple concepts. Furthermore, our proposed approach plays a critical role in overcoming the limitations of prompt engineering by providing a structured, CPC-based knowledge that reduces prompt dependency to generate distinctive, domain-specific innovations.
Nevertheless, limitations, such as time-consuming, dependence on existing knowledge, including patents, and a necessity to learn the design catalogue method, remain. The advantages and weaknesses are summarized in Table 4.
| LLM ideas | Novelty Potential/Void matrix and Design Catalogue ideas | |
|---|---|---|
| Advantage |
|
|
| Weakness |
|
|
The novelty, feasibility and market potential between the novelty potential/void matrix ideas and LLM ideas were scored by independent experts. Table 5 presents the descriptive statistics of expert evaluations. Overall, the ideas produced a mean novelty score of 3.27 (SD = 0.97), indicating moderate novelty compared to current market solutions. Feasibility was rated the highest (M=3.66, SD= 1.162), suggesting that most ideas were perceived as technically achievable. Market potential gained a mean score of 3.46 (SD=1.121).
| N | Mean | Std. Deviation | |
|---|---|---|---|
| Novelty | 152 | 3.27 | 0.970 |
| Feasibility | 152 | 3.66 | 1.162 |
| Market potential | 152 | 3.46 | 1.121 |
| Group classification | 19 | 2.58 | 0.902 |
The Kruskal-Wallis H test in Table 6 indicates that the novelty potential/void idea scored the highest in novelty (M=15.50) than those generated by LLM models, ChatGPT (M=8.60), Gemini (M=10.38), and DeepSeek (11.75).
| Group | Mean rank | H (df=3) | p-value | |
|---|---|---|---|---|
| Novelty | NP/Void | 15.50 | 2.279 | 0.517 |
| ChatGPT | 8.60 | |||
| Gemini | 10.38 | |||
| DeepSeek | 11.75 | |||
| Feasibility | NP/Void | 6.50 | 1.437 | 0.697 |
| ChatGPT | 11.20 | |||
| Gemini | 9.75 | |||
| DeepSeek | 8.13 | |||
| Market potential | NP/Void | 8.50 | 0.351 | 0.950 |
| ChatGPT | 10.20 | |||
| Gemini | 10.88 | |||
| DeepSeek | 9.0 |
For feasibility, ChatGPT scored the highest (M=11.20), followed by Gemini (M=9.75), DeepSeek (M=8.13) and novelty potential/void (M=6.50). In terms of market potential, Gemini scored highest (M=10.88), followed by ChatGPT (M=10.20), DeepSeek (M=9.0) and novelty potential/void (8.50). Although the differences were not statistically significant (p>0.05), the trend still suggests that the proposed method may offer higher novelty compared to LLM-generated ideas. This implies potential value in domains requiring original, domain-specific ideation. Table 7 summarizes the top-rated ideas (score ≥4 or 5) from at least half of the evaluators across the four groups.
| Top novelty | Top feasibility | Top market potential | |
|---|---|---|---|
| NP/Void (n=1) | 1 | 0 | 1 |
| ChatGPT (n=10) | 3 | 8 | 6 |
| Gemini (n=4) | 2 | 3 | 4 |
| DeepSeek (n=4) | 3 | 1 | 4 |
Out of 19 ideas, 9 were rated as highly novel, including the kineti-form (8 evaluators), AR dojang (7 evaluators) and smart scoring platform (6 evaluators). A total of 12 ideas were rated as highly feasible such as the cultural fusion uniforms and eco-friendly gear line, each rated by 8 evaluators. Finally, 15 ideas were considered to have high market potential, including AR training pods and gamified progress tracker app, with 7 evaluators recognizing their potential. The novelty of the proposed ideas is further discussed through an analysis of CPC codes in the subsequent section of this article.
4.5 Verification with CPC CodesThe novelty of the ideas generated using the methods proposed in this study was verified through the investigation of CPC codes trend combination of A63B71/06 (scoring), G06N20/00 (machine learning), and G06V20/41(semantic clustering). Patent databases covering application years from 2000 to 2019 were examined. The findings are illustrated in Figure 7.
Figure 7 Individual CPC trends based on application year from 2000 to 2019 (A63B71/06, G06N20/00, G06V20/41)
Based on the results, an upward trend for all CPC codes was observed, particularly accelerating from 2014 onwards. The G06N20/00 demonstrated the highest number of patents and growth, reflecting the rising interest and advancements in AI technologies, such as machine learning, within the last decade. The observations aligned with global sport innovation trends, which made AI a crucial component of technological advancements.
The CPC code G06V20/41 demonstrated steady growth, specifically after 2012, highlighting the vast applications and commercialization of image-processing technology across various fields with improved sensors, cameras, and computational capabilities. Similarly, the TKD-related CPC code, A63B71/06, revealed a steady but moderate increment compared to the other two CPCs. The results suggested a consistent interest in and incremental innovation within the sports equipment industry, potentially reflecting niche technological advancements in scoring systems.
Following the combination of the abovementioned CPC codes, no existing patent was identified. The information suggested that the proposed new combined ideas, which integrated AI, advanced image recognition, and scoring, might represent a novel idea. The findings from the semi-structured interviews conducted with representatives of the Malaysian TKD Federation and clubs are highlighted in the subsequent section.
This study conducted semi-structured interviews with representatives of the Malaysian TKD Federation and clubs to obtain insights and understand the challenges and importance of innovation. Feedback on the design catalogue as an innovation tool for idea generation was also gathered during the interviews. The respondents were provided with a list of questions they could answer freely. This study considered the responses as the private opinions of the individuals, not the opinions of the entire federation.
4.5.1 Importance of innovation for TKDBased on the responses from the semi-structured interviews conducted in this study, innovation was considered notably vital for TKD. Novel ideas, such as adopting technology, integrating AI, leveraging digital advancements, sustainable facilities and training environments are crucial in increasing innovation and ensuring the growth and competitiveness of the federations. For instance, the electronic scoring system, protective scoring system (PSS), has improved transparency in sparring matches. Nevertheless, PSS implementation in training sessions presents challenges due to the notable costs associated with the technology. Video assistant referee (VAR) technology has recently been introduced to assist referees in enhancing decision-making accuracy, particularly in blindside scoring. The implementation of VR is another example which enabled global remote sparring in TKD competitions.
4.5.2 Design catalogue as an innovative tool to generate ideasThe interviewees’ familiarity and attitudes towards employing a design catalogue for generating ideas are summarized as follows:
The principal challenges and barriers identified by the interviewees that may hinder the adoption of a design catalogue are discussed. The following are the findings:
The interviewed representatives recommended the following to enhance the usability, accessibility, and effectiveness of the design catalogue within the TKD community:
This section discusses the results of the proposed methodologies including independent expert scoring and new design concept for TKD SEs. Despite the statistical results yielding no significant differences (p>0.05), the observed trend suggested that our proposed method may offer higher novelty compared to LLM-generated ideas. The outcome may result in potential biases based on the evaluators’ background, whereby they were specialists in mechanical and applied engineering rather than TKD market experts.
This may have led to the underestimation of ideas that were difficult to visualize for practical implementation. Consequently, zero-void combinations may have been perceived as emergent ideas that remain difficult for evaluators to envision in terms of practical implementation. Therefore, it is essential to consider how the evaluators’ backgrounds may have influenced their assessments of novelty.
Based on the results, this study proposed a new design concept for TKD SEs, a smart scoring system combined with an interactive emotion-driven platform, leveraging advanced technologies including machine learning, semantic classification, and sensor technologies. Figure 8 illustrates the framework for the smart scoring system and emotion-driven platform for TKD SEs.
Figure 8 The smart scoring system and emotion-driven platform framework for TKD SEs
The system developed in this study utilizes CPC codes, such as A63B71/06 (scoring), G06N20/00 (machine learning), and G06V20/41 (semantic clustering). While existing scoring systems primarily focus on athletes’ physical or technical performance, integration with emotional capabilities remains limited. Incorporating emotional aspects can provide a new engagement layer, where reactions of the audience or athletes (such as recognizing excitement or tension) can be employed to enhance feedback during training or competition.
The developed approach also leverages real-time monitoring, feedback, and decision-making, ultimately connecting multiple stakeholders, including athletes, coaches, audiences, and sponsors. For example, the TKD PSS can be combined with emotion recognition technology by collaborating with sports technology and AI companies, such as Affectiva, MoodMe, and Kick.ai. Affectiva’s emotion AI technology, utilizing machine learning and computer vision, including camera sensors can be employed to analyze real-time TKD athletes’ facial expressions, gestures, and body posture.
Another technology developed by MoodMe’s solutions, such as Emotion Fan Cam, enables real-time capture and analysis of audience emotions, reactions, and engagement levels. The audience’s emotional data streams, along with scoring results displayed during live competitions, can significantly enhance fan engagement and provide important insights for event sponsors and organizers. Furthermore, the integration of blockchain technologies—such as Scalar DLT—with scoring system software like iPS Host has the potential to enhance fairness and minimize the risk of disputes in TKD judging. Scalar DLT employs a decoupled consensus mechanism to establish a tamper-resistant distributed ledger [3].
The platform proposed in this study also adopts a SoS approach, which is essential in integrating independent yet interconnected subsystems. Each component in the system operates independently while maintaining managerial autonomy—meaning that one module can be updated without disrupting the entire system [22]. For instance, new machine learning algorithms can be implemented into the scoring system without disrupting the emotion analysis component. This integrated approach ensures the platform is flexible and adaptable to any technological advancements or changing user requirements, thereby meeting the evolving needs of TKD SEs.
From the viewpoint of a management system [23], the platform developed in this study can serve as monitoring, decision-support and feedback, data integrity and transparency, and stakeholder engagement systems. Several valuable propositions for the new system are as follows:
The global market demand for the sports technology industry is experiencing significant growth. In the past five years, an 11.32% annual rate and over 2,300 new companies have emerged [24]. An example of key players in sports technology that have revolutionized fan engagement is IBM Corporation. Their AI-powered platforms such as Watson enhance the sports viewing experience by integrating predictive analytics and real-time information [1]. Consequently, converging smart scoring systems and emotion-driven platforms in TKD aligns with the growing demand for data analytics and personalized training in sports. The methods proposed in this study and the SoS approach can support TKD SEs maintain innovation and competitiveness.
This chapter summarizes the key contributions of the study, acknowledges its limitations, and outlines directions for future research.
First, this study proposed a new strategy to generate novel design concepts for future systems. A case study on TKD utilizing idea generation methodologies, such as the novelty potential/void matrix, design catalogue employing CPC codes, and Gen AI (ChatGPT), were implemented. A highly novel design concept and logical approach was developed, which incorporates a smart scoring system with an interactive, emotion-driven platform that integrates machine learning, semantic classification, blockchain, and sensor technologies. It provides real-time and precise scoring that reflects training performance and analyzes the emotional aspects of athletes and audiences.
Our approach yielded ideas with no overlap compared to LLM-based approaches, indicating its strength in niche innovation. Nevertheless, the CPC codes assessments verified the absence of similar patents, indicating potentially novel combinations and an increased potential for a first-mover advantage for TKD SEs. Furthermore, the responses from the semi-structured interviews with representatives of the Malaysian TKD Federation and clubs served as an initial practitioner-level validation offering insights and feedback on the design catalogue as an innovation tool for idea generation.
Second, this study had several limitations, including limited generalizability due to only performing a single case study focusing on TKD. Another disadvantage of this study is that the interviews was restricted to a small sample obtained from a single country’s perspective, Malaysia. The quality of the prompt employed with ChatGPT to generate ideas can also affect reproducibility. Additionally, this study utilized patent data, which may lead to potential biases from representation and geographical factors, as not all inventions were patented or disclosed. CPC codes were also relied on heavily, which are subject to potential classification inconsistencies due to human judgment. This may have resulted in uneven granularity and coverage.
Third, accordingly, future research can consider exploring cross-sector applications by expanding the case study beyond TKD to include different sports, sectors and technological domains. This would further strengthen the generalizability and robustness of the proposed methodologies. Future international interviews should be conducted with TKD practitioners from Korea, Europe, and North America to validate the design catalogue’s applicability across cultural and organizational contexts.
Additionally, comparing the design catalogue with formal TKD education systems, specifically in Korean universities recognized as global benchmarks for structured TKD training, could provide valuable insights and potential for broader adoption in educational settings. Furthermore, multilingual adaptation through other languages such as Korean, Malay and Japanese would encourage wider adoption and inclusivity across global TKD communities. Finally, the design catalogue can be refined as an innovation tool for TKD SEs based on the TKD representatives’ recommendations.
