1.Background

With the concurrent advancement of the economy, South Korea has seen an increasing interest in healthcare information among its citizens, propelled by advancements in medical technology and improvements in education. The desire for better health has escalated with the rapidly aging population. Both individuals and society are actively interested in enhancing health and fitness and prolonging lifespan. This growing interest extends to oral health, which plays a role in improving quality of life¹⁾. As consumer interest in oral health continues to rise, premium-grade toothbrush market competition within the oral care product industry intensifies. This has led to the introduction of various functional toothpaste products amid increasing competition²⁾. Interest in oral cleansers has been increasing, not only for the management of oral diseases and the oral environment but also for functions such as bad breath elimination and teeth whitening. The development of effective oral cleansers targeting periodontal disease and oral environmental management has been consistently researched. In South Korea, active research focuses on natural substances that induce changes in oral environments and examines their antibacterial effects on periodontal diseases and oral bacteria³⁾. Interest in dental treatments extends beyond oral health and includes an increased focus on aesthetic dental care. Owing to recent westernized dietary changes, there has been a growing demand and interest in orthodontics⁴⁾. Moreover, with an increasing proportion of adults seeking orthodontic treatment and an increase in income levels, there is a heightened desire for aesthetic procedures. Consequently, ongoing developments in orthodontic device innovations and clinical research are aligned with these trends. Transparent aligners and lingual orthodontic devices have been widely adopted as key tools in aesthetic orthodontic treatments. Companies that provide transparent aligners are developing various methods and auxiliary devices for tooth movement⁵⁾. In dental prosthodontics, numerous aesthetic materials and restorative products have been introduced, notably zirconia, which has garnered significant attention owing to its excellent strength, wear resistance, and high biocompatibility⁶⁾. As the demand for esthetic restorative materials increases, computer-aided design/computer-aided manufacturing (CAD/CAM) methods have been introduced. In dentistry, the utilization of three-dimensional (3D) printing technology is steadily increasing in various areas such as diverse tooth models, temporary teeth, transparent aligners, and implants, leading to continuous advancements in associated technologies^7,8).

As various studies have advanced with the development of oral health-related technologies, it is challenging to find comprehensive research that encompasses all oral health-related technologies. There are diverse methods for predicting promising technologies using technological trend analyses. Traditional methods such as the Delphi technique, analytic hierarchy process, scenario technique, expert panels, and trend extrapolation exist but are limited by the subjective opinions of relevant technology experts. Consequently, data-based methods utilizing paper and patent information have been predominantly employed⁹⁾. Patent data, for example, can be utilized as a metric for measuring trends and achievements in technological research and development. It refers to textual documentation containing specific technical and scientific information about an invention and detailing the elements intended for legal protection. Moreover, it is quantifiable and objective data on the latest technology, which is globally standardized according to specific formats, thus making it applicable for various purposes. Sawng et al.¹⁰⁾ and Jo et al.¹¹⁾ have previously conducted research utilizing patent information as analytical data for technological trends¹²⁾. In this study, we performed network analysis and topic modeling using text analysis programs to comprehensively analyze patent data. Network analysis involves breaking down words constituting sentences to extract meaningful concepts, and represents how words form relationships in a network format. Network analysis allows us to identify keywords and understand the connections between words, thereby enabling us to infer the context of sentences within a text¹³⁾. Kim et al.¹⁴⁾ also applied network analysis methods to analyze keywords related to dental hygiene. Topic Modeling is a technique used to estimate latent and meaningful topics within a collection of unstructured text data. Among these algorithms, Latent Dirichlet Allocation (LDA) is a method that calculates a specific number of topics by considering the probability distribution of terms related to the topics¹⁵⁾. Kim et al.¹⁶⁾ and Lee¹⁷⁾ conducted patent analyses by applying topic-modeling techniques to uncover hidden themes within a large volume of documents.

2.Objectives

Although previous studies have examined text analysis using patent data, there remains a shortage of research in the healthcare field. This study aimed to utilize text-mining techniques to examine the interrelationships among keywords in patents related to oral health and extract latent topics for visualization. Through this process, the study sought to derive significant keywords, analyze core and specific technologies within different topics, and comprehend technological trends. Additionally, it advocates the active utilization of text analysis techniques in dental hygiene and aims to utilize patent-derived analysis as fundamental data to promote advancements in oral health-related technologies.

This study progressed sequentially, as depicted in Fig. 1, and involved data collection, data preprocessing, keyword extraction, frequency analysis, network analysis and visualization, and topic modeling.

Fig. 1. Research methodology.

1.Patent data collection

As of July 31, 2023, a total of 11,710 patent documents related to oral health were collected from the domestic patent database service platform, KIPRIS (www.kipris.or.kr). The scope of data collection was not limited by time and included patents that were either publicly disclosed or registered, utilizing search queries constructed with keywords related to “dentistry,” “teeth,” and “oral health.” Keyword selection was based on synonymous words provided by the KIPRIS’s search term expansion function. After gathering the patent data, invention titles and abstracts were reviewed, excluding titles associated with pet-related contents. Additionally, among Korean invention titles containing “teeth,” “tooth,” or “dental” terms, those not semantically related were excluded. The final research dataset comprised 6,865 invention titles in Korean, collected from March 2003 to July 2023.

2.Data preprocessing and keyword extraction

We used Python version 3.10 (Python Software Foundation, Santa Clara, CA, USA) as an analytical tool capable of text analysis. For the Korean text, we employed the Hannanum morphological analyzer from the Korean Natural Language Processing in Python (KoNLPy) package. To extract meaningful keywords, we preprocessed the tokenized words. The words in Korean, such as “seuk-ru, im-peul-laen-teu, im-peul-lam-teu, beu-rae-kit, pik-seu-chyu, pik-seu-chyu-eo, pik-seu-chwo, yu-nit, yu-ni-teu, eo-bu-teu-meon-teu, and eo-byu-teu-meon-teu,” have been standardized to their equivalent words with the same meanings: screw, implant, bracket, fixture, unit, and abutment. This involved preprocessing by removing punctuation, numbers, and stop words, and extracting words consisting of more than one syllable for the analysis.

3.Keyword frequency analysis

Using the Pandas module and counter function in Python, we conducted a frequency analysis of keywords (37,155) extracted from patent invention titles. We extracted the top 1 to 50 keywords with the highest occurrence frequencies.

4.Network analysis and network visualization

We conducted a network analysis to understand the connections between the keywords. We compared the degree centrality indices of the top 50 keywords by frequency. This value is based on the degree of direct connections between nodes, indicating the centrality of the network. A higher degree of centrality implies that a node is highly connected directly to many other nodes, and is calculated based on the number of links directly connected to a specific node^18,19). The analysis was performed using a Python library network. From the total keywords, the 50 most frequently appearing keywords were used to derive a 2D adjacency matrix (50×50), indicating the co-occurrence relationships among these keywords. A portion of the top 10 results is presented in tabular form. For visual inspection of the degree of connectivity between nodes, we visualized the keyword network using Gephi 0.10.1. (Gephi Consortium, Paris, France) The nodes were positioned based on their degree centrality values. Nodes ranked from 1 to 10 were centrally placed and their color was set to red. The nodes ranked 11∼20 were orange, 21∼30 were yellow, 31∼40 were green, and 41∼50 were blue.

5.Topic modeling

To extract meaningful topics from the patent data, we performed topic modeling based on the connection rules between keywords. We used the “gensim.models.LdaMulticore” module from the Gensim library, version 4.3.0 in Python, as our research tool. To determine the optimal number of topics, we conducted topic optimization experiments based on topic coherence values. We varied the number of topics from 1 to 20, measured the coherence value for each topic, and selected the number of topics that exhibited the highest coherence value as the appropriate number of topics for our study. The repetition count for topic sampling was set to 500 times²⁰⁾. We used the LDA method to categorize data collected from the library into topics. For the parameter settings, we designated the number of topics (num_topic) as eight. The chunk size, which represents the number of documents processed in a single training session, was set to 2,000. The total number of training passes (passes) were set to 20, and the per-document iteration count (iterations) was set to 1,500²¹⁾. The results of the topic modeling were visualized using the pyLDAvis library, which demonstrated the trained model. The keywords associated with topics were generated by setting the lambda (λ) value to 0.6²²⁾. Lambda values range from 0 to 1 and act as hyperparameters that determine the diversity of word selection²³⁾. Furthermore, the lambda value represents the weight that indicates the degree of relevance between each topic and keyword. When setting the weight to 1, it generates a ranking of familiar terms based on the words that frequently appear in each topic. However, setting the weight to 0 prioritizes the selection of words that show significant differences among topics^24,25). The research findings of topic modeling were interpreted by two researchers (dental hygienists) under the guidance of a topic-modeling expert. The criteria for topic selection were centered on five main keywords. During the topic-labeling process, topics were structured to include at least one of the five keywords.

1.Keyword frequency analysis and network analysis

Table 1 presents the results based on the top 50 words with a high frequency of occurrence in invention titles, along with their respective centrality values. According to the analysis of occurrence frequency and centrality, “Method” exhibited the highest centrality value. Following this, “Apparatus” and “Tooth” showed high frequencies. Words with a centrality of 0.9 or higher, excluding words with a centrality of 1, were “Apparatus”, “Treatment”, “System”, “Inclusion”, “Possibility”, and “Oral”. Examining words ranked 11th to 15th based on centrality, we observed “Implant,” “Orthodontics,” “3D,” “Processing,” and “Usage.” Among the top 50 keywords, those with lower centrality included “Extracts,” “Zirconia,” and “Active ingredient,” which ranked 48th, 49th, and 50th, respectively (Table 1).

Frequency, Degree Centrality of Top 50 Keywords

Word	Frequency			Degree Centrality
	Rank	Frequency		Rank	Degree Centrality
Method	1	2,347		1	1
Apparatus	2	1,677		5	0.979
Tooth	3	1,502		2	1
Oral	4	1,491		10	0.918
Manufacture	5	1,001		3	1
Composition	6	949		17	0.775
Implant	7	939		11	0.897
Orthodontic	8	736		12	0.897
System	9	626		7	0.938
Use	10	577		4	1
Inclusion	11	564		8	0.938
Toothbrush	12	326		43	0.51
Dentition	13	254		34	0.632
Instrument	14	245		20	0.734
For treating	15	238		27	0.693
Prevention	16	223		45	0.489
Treatment	17	202		6	0.979
Extracts	18	194		48	0.387
Recording	19	193		24	0.714
Prosthetic	20	189		28	0.673
Picture	21	188		25	0.714
Containing	22	182		44	0.51
Bracket	23	172		47	0.408
Guide	24	171		18	0.775
Disease	25	169		39	0.591
Procedure	26	166		35	0.632
Processing	27	161		14	0.836
Imaging	28	147		38	0.612
Management	29	145		21	0.734
Possibility	30	140		9	0.938
Production	31	131		19	0.755
Data	32	130		22	0.734
Artificial	33	130		23	0.734
3dimension	34	128		13	0.877
Medium	35	128		29	0.673
Scanner	36	123		40	0.571
Image	37	121		36	0.632
Provide	38	120		16	0.795
Active ingredient	39	118		50	0.367
Generate	40	117		26	0.714
3D	41	114		30	0.673
Scan	42	113		31	0.653
Computer	43	113		32	0.653
Abutment	44	113		41	0.551
Toothpaste	45	112		46	0.448
Usage	46	105		15	0.836
Zirconia	47	98		49	0.387
Digital	48	95		37	0.632
Purpose	49	94		42	0.53
Medical	50	94		33	0.653

2.Visualization of network analysis results

After reviewing Fig. 2, which displays the results of the network visualization and examines the 2D adjacency matrix (50×50) showing the co-occurrence relationships, it is evident that thicker links between nodes represent a higher frequency of co-occurrence. For instance, when “Method” is linked to “Manufacturing (866),” “Tooth (593),” and “Apparatus (680),” the links appear thicker. “Oral” is found to be connected to “Composition (451).” “Implant” shows multiple connections with “method (260),” “Apparatus (137),” and “Tooth (260).” “Orthodontics” exhibits close associations with “Method (228),” “Apparatus (266),” “Tooth (228),” and “Dentition (220)” (Fig. 2).

Fig. 2. Network connectivity of the top 50 keywords (The nodes’ colors were represented based on betweenness centrality values: the top 10 are in red, rankings 11∼20 in orange, 21∼30 in yellow, 31∼40 in green, and 41∼50 in blue). As the frequency of co-occurrence increases, the thicker links between nodes in the network appear.

3.Topic modeling

To determine the optimal number of topics, topic coherence values were computed for varying topic counts ranging from 1 to 20. The coherence value for the eight topics was highest at 0.807 (rounded to the fourth decimal place) (Fig. 3).

Fig. 3. Coherence scores by number of topics (the dashed lines indicate the highest coherence scores for the number of topics).

Table 2 displays the top ten core keywords for each of the eight topics, representing the labeling results for each topic. Each topic was analyzed based on primary keywords to understand the respective subjects. Topic 4 is identified with the highest weight, characterized as “Implants.” Meanwhile, the themes for the lowest-weighted topics, Topic 1 and Topic 3, are “Instrument and Materials for Oral Health” and “Toothbrush and Oral Health Care,” respectively.

Result of the Topic Modeling

Numbers of topic	Topic label (Keywords 1∼10)
1	Instrument and Materials for Oral Health (Instrument, Ingredient, Care, Hygiene, Improvement, Washing, Sintering, Structure, Impression Material, Tissue)
2	Orthodontics (Method, Apparatus, Orthodontic, Dentition, Recording, Picture, Data, Processing, Tooth, Image)
3	Toothbrush and Oral Health Care (Toothbrush , Fixing, Able, Removing, Electric, Unit chair, Gum, Driver, Toothbrush hair)
4	Implant (Manufacture, Implant, Method, Use, Guide, Abutment, Prosthetic, Tooth, 3D, Fixture)
5	Oral Composition for Prevention and Treatment (Composition, Oral, Inclusion, For treating, Prevention, Disease, Containing, Extracts, Active ingredient, Toothpaste)
6	Dental Treatment Aid Apparatus (Apparatus, Measurement, Mask, Auxiliary, Root canal, Dental, Occlusion, Material, Medical, Cleaning)
7	Apparatus or Method Based on Artificial Intelligence (Tooth, Orthodontic, Artificial, Model, Apparatus, Whitening, Purpose, Derived, Transparent, Attachment)
8	Oral Care System or Service (Management, System, Prove, Service, Method, Oral, Zirconia, Predict, Graft material, Health)

The lambda (λ) value, determining the diversity of word selection, was set to 0.6.

Fig. 4 presents a visualization of topic modeling using pyLDAvis after setting the number of topics to eight. It displays the Intertopic Distance Map (IDM) and the 30 core keywords. In the IDM, topics 6 and 7 intersect within one quadrant, while in another quadrant, topics 2, 4, and 8 are situated closely, overlapping with each other. In contrast, topics 1, 3, and 5 appeared relatively distant from the other groups and did not overlap with any other topics. The weightage of each topic was: Topics 1 (5.5%), 2 (19.6%), 3 (5.5%), 4 (22.0%), 5 (17.4%), 6 (8.9%), 7 (10.5%), and 8 (10.5%). The topic with the highest weightage is represented by the larger circle indicating “Implant,” while the lowest-weighted topics are indicated by smaller circles denoting “Instrument and Materials for Oral Health” and “Toothbrush and Oral Care.” The core keywords forming the entire set of topics are “Manufacturing,” “Composition,” “Implant,” “Oral,” and “Method,” with the most frequent keyword being “Methods” (Fig. 4).

Fig. 4. Intertopic distance map.

1.Interpretation

This study was conducted to understand technological trends through keyword network analysis and topic modeling based on domestic patents related to oral health. When considering the occurrence frequency, words, such as “Method,” “Apparatus,” “Tooth,” “Oral,” and “Manufacture,” showed higher proportions. This could be attributed to the collection of data that primarily focused on patents related to oral health, hence reflecting these prominent terms. According to previous research²⁶⁾ analyzing patents using network analysis methods, there are cases where words, such as “Method,” “Apparatus,” and “System,” were not considered as stop words and were derived as key keywords. Considering that specific technical domains might influence the interpretation of the results, these words were not excluded and were analyzed together. Additionally, in studies related to medical procedures and patents, it is noted that medical procedures involving human life are excluded from patent rights to prevent restrictions on such procedures. Instead, according to patent office practices, the functional and systemic operating methods of medical devices themselves are recognized as patentable, not as medical procedures or treatments²⁷⁾. This can be interpreted as one of the reasons why words related to “method” and “apparatus” accounted for a significant proportion in the frequency analysis results of this study.

To identify keywords with relatively higher importance in relation to their frequency of occurrence, we analyzed their degree of centrality. There are cases where keywords are perceived as important, despite having a low frequency of occurrence but a relatively high degree of centrality. For instance, the keyword “Use,” ranking 10th in frequency, had the highest degree centrality of 1, while “Treatment,” ranking 17th, showed the next highest value of 0.979. Therefore, relying solely on the frequency of occurrence to derive key terms should be avoided. Furthermore, when comparing the rank differences between frequency of occurrence and degree centrality, “3D” ranked 34th in frequency with 128 occurrences but secured a relatively higher position at 13th with a degree centrality of 0.877. The importance of “3D” is also evident in the network visualization. It is observed to be connected with nodes such as “Data,” “Scanner,” and “Scan.” In the collected patent data, titles like “Method for 3D scan data processing for dental prosthesis manufacture” and “Apparatus and method for restoring 3D oral scan data using computerized tomography images” were identified. Moreover, the relationships between keywords were confirmed within the sentences. Digital dental technology based on 3D printing, one of the core technologies of the Fourth Industrial Revolution in dentistry, is advancing. The introduction of 3D scanners into dentistry has significantly affected CAD/CAM technology in prosthetic manufacturing systems, and digital impression techniques using intraoral or extraoral scanners are commonly used in clinical settings²⁸⁾.

We conducted topic modeling and keyword analysis to extract potential themes from the text. Determining the number of topics before executing the model is crucial in this regard²⁹⁾. There is no statistical solution for determining the appropriate number of topics, as it depends heavily on the interpretability, validity, and research questions that influence the topics generated through topic modeling. The determination of the number of topics is subjective and is decided by researchers based on what they believe would yield the most meaningful results through topic modeling.

2.Key results and comparison

To minimize subjectivity, topic coherence values were used as in a previous study³⁰⁾. In this study, a function measuring topic coherence values for each number of topics from 2 to 20 was implemented in Python, following the methodology of a previous study to determine the appropriate number of topics³⁰⁾. In topic modeling analysis, topics with low relevance to the research subject are sometimes excluded based on the researchers’ judgment. In related studies, it has been observed that researchers reviewed the words included in the derived topics to eliminate those with low relevance to the research topic from the analysis. This exclusion of low-relevance topics by the research team has been acknowledged and considered in the topic inference method^25,31).

Topic 1 comprises words like “Instrument” and “Ingredient,” followed by “Care,” “Hygiene,” and “Improvement.” Interpreting this as a comprehensive concept encompassing oral health, it was named “Instrument and Materials for Oral Health.”

Topic 2 mainly consists of words such as “Orthodontic” and “Dentition,” and this topic was named “Orthodontics.” The analysis also confirmed the association of words like “Recording,” “Picture,” and “Data” with “Orthodontic.” Orthodontics involves clinical assessments based on quantitative analyses of the human skeletal structure and dental alignment. In clinical practice, there are ongoing developments in orthodontic analysis systems in South Korea, such as WebCeph software (AssembleCircle Corp., Seongnam, Korea), using imaging data, indicating a significant attempt to employ artificial intelligence (AI) technology and digital image recognition and processing in the field of dental alignment³²⁾.

Topic 3 is focused on words like “Tooth,” “Removing,” and “Electric,” thus designated as the theme “Toothbrush and Oral Health Care.” Choi et al.³³⁾ analyzed 512 patents from 2005 to 2014, to examine the patent trends related to toothbrushes. They suggested that various shapes, materials, functions, and socially oriented aspects of toothbrushes have been consistently researched and patented. Moreover, they anticipated the continued production of electric toothbrushes.

Topic 4 is inferred as “Implant” based on words such as “Manufacture,” “Implant,” “Method,” “Use,” and “Guide.” Unlike in the past, dental implants are now considered and proposed as the foremost treatment when teeth are lost, demonstrating a high long-term success rate and reliability. Studies continue to focus on the materials and surface forms of implants for osseointegration, aiming to address the drawbacks or complications associated with them³⁴⁾. In the field of surgery, diagnostic models that replicate the structure or form of surgical sites are being developed to aid in surgical decision-making. Templates and surgical guides that utilize 3D printing technology are being developed to enhance the accuracy and safety of surgeries⁸⁾. Recently, the majority of the domestic implant companies in Korea have expanded the distribution of advanced software and 3D printers. This has enabled practitioners to directly develop surgical guides for implants. The use of implant surgical guides has become widespread, facilitating their diverse production using various materials and methods. Several researchers are actively studying this trend^35,36).

Topic 5 was inferred based on the words “Composition,” “Oral,” “For treating,” “Prevention,” and “Extracts,” and was named “Oral Composition for Prevention and Treatment.” According to the study of Hwang et al.³⁷⁾, when analyzing patents related to toothpaste on KIPRIS, composition accounted for 35% of the patents, representing the most developed field. Various extracts have been used in patents related to preventing and treating periodontal diseases. Moreover, studies are being conducted on the effects of extracts, such as their antibacterial effects, on oral health³⁷⁾.

Topic 6 appears to be associated with “Dental Treatment Aid Apparatus,” based on the words “Apparatus,” “Measurement,” and “Root Canal.” Through research on devices for occlusal force measurement³⁸⁾, wireless handpieces³⁹⁾ for root canal treatment, and devices for root canal irrigation⁴⁰⁾, it is evident that devices and instruments related to measurement and root canals are under investigation. Devices for measuring root canal length, vertical height, and bone density were identified in the collected patent inventories.

Topic 7 highlights the development of devices for artificial tooth processing and orthodontic systems utilizing AI, with prominent keywords including “Tooth,” “Orthodontic,” “Artificial,” “Model,” and “Apparatus.” Therefore, it was named “Apparatus or Method Based on Artificial Intelligence.” The future of AI foresees its extensive adoption in various healthcare sectors, particularly in the diagnostic imaging field, and it is anticipated to be widely applied in dentistry. Research suggests that AI will enhance diagnostic efficiency by automating repetitive tasks, thereby reducing costs and alleviating the burden on healthcare systems in an aging society. This efficiency is expected to extend to dental inventory management, patient scheduling, automatic generation of electronic medical records, and surgical feedback, eventually evolving into an integrated dental operation and management system that utilizes AI⁴¹⁾. In a study by Choi et al.⁴²⁾, it was highlighted that clinical trials of AI-assisted software aiding physicians’ diagnoses in South Korea increased significantly from six cases in 2018 to 17 cases in 2019. Moreover, the range of conditions targeted by AI technology has expanded beyond prostate and breast cancer to include diverse conditions, such as lung diseases, vertebral compression fractures, and dental conditions⁴²⁾. The potential for the advancement of patented technology related to “Apparatus or Method Based on Artificial Intelligence” is considerably high and requires continuous attention.

Topic 8 was named “Oral Care System or Service,” derived from major keywords such as “Management,” “System,” “Provide,” “Service,” and “Method,” concluding the analysis. With the growing interest in oral health, governments have been continuously implementing oral health promotion programs targeting vulnerable populations, such as students, seniors, and people with disabilities, by establishing oral health clinics and regional clinics for dental care^43,44). With the evolving trends of the times, the approach to providing oral health services has transformed. The widespread adoption of smartphones has led to increased interest in applications aimed at improving the quality of life of people with disabilities. For instance, there was a case where an Android-based “15 Minutes of Daily Oral Exercise” app was developed to enhance oral motor skills for individuals with cerebral palsy⁴⁵⁾. Moreover, studies analyzing oral health-related apps have highlighted their significance as crucial educational tools for acquiring oral health knowledge and proper maintenance methods, as seen in related research⁴⁶⁾. This development indicates the advancement of technology related to oral care services associated with the theme of Topic 8.

3.Suggestion

The technologies associated with oral health have advanced in various aspects. However, while there have been studies on specific technological trends in dentistry and dental hygiene, there is a lack of comprehensive research analyzing patented technologies. Therefore, a significant contribution of this study lies in deriving the potential topics related to oral technology over the past 20 years and segmenting them into particular topics. This not only aids in setting future directions for detailed research and formulating patent strategies but also provides valuable data. Despite the methodological limitations, the value of this study lies in deriving results based on a vast amount of patent data in the dental hygiene field using Python and proposing a method to extract keywords representing topics. We hope that the methodology and visualization tools used in this study will find applications in various research fields within dental hygiene. Moreover, the results from network analysis and topic modeling can be utilized to guide technological development in the field.

4.Limitations

In this study, text mining techniques were employed for network analysis, whereas in topic modeling, efforts were made to maintain objectivity by undergoing refinement processes and utilizing stop-word lists. While objectivity regarding the number of topics was maintained using topic coherence values, the process of inferring and naming topics unavoidably involved qualitative interpretation by the researchers, potentially incorporating subjective judgments based on their choices. Exploring various methods to maintain objectivity in topic modeling is essential, and further research is required to select topic labels that semantically represent topics effectively. Additionally, applying time-series analysis to topic modeling to analyze patent trends across different years would aid in predicting competitiveness and promising technologies for various technical changes over time.

5.Conclusion

This study conducted text analysis on 11,710 patent data related to oral health and dentistry obtained from KIPRIS as of July 31st, spanning 2023 years, using a program developed in Python. Of these, 6,865 keywords highly relevant to the field were selected. The analysis involved determining the relationships among keywords in patent titles by analyzing centrality based on the top 50 words by frequency and applying topic modeling using LDA to infer topics.

The analysis of the connectivity between adjacent words revealed that the centrality index was highest for “Method” and lowest for “Active ingredient.” In network visualization, “Method” exhibited thicker links when connected to “Manufacture,” “Tooth,” and “Apparatus,” while “Oral” was closely linked to “Composition.”

Regarding topic modeling derived from keywords, the topic weights indicated “Implant” as the highest, followed by “Orthodontics,” “Oral composition for prevention and treatment,” “Apparatus or Method Based on Artificial Intelligence,” “Oral Care System or Service,” and “Dental Treatment Aid Apparatus.” Conversely, topics related to “Instrument and Materials for Oral Health” and “Toothbrush and Oral Health Care” had the lowest weightage. Therefore, in clinical and research contexts, studies primarily focused on “Implant” and “Orthodontics” regarding patent technology trends.

This study aimed to comprehensively understand and categorize technological trends related to oral health and utilize the frequency and relationships of the analyzed keywords to grasp technological advancements. It is anticipated that this research will serve as fundamental material for technological developments associated with patents in dentistry and dental hygiene.

None.

No potential conflict of interest relevant to this article was reported.

Not applicable.

Conceptualization: Hee-Kyeong Bak. Data acquisition: Hee-Kyeong Bak. Formal analysis: Hee-Kyeong Bak. Supervision: Han-Na Kim. Writing-original draft: Hee-Kyeong Bak and Han-Na Kim. Writing-review & editing: Hee-Kyeong Bak, Yong-Hwan Kim, and Han-Na Kim.

None.

Raw data is provided at the request of the corresponding author for reasonable reason.