=Paper=
{{Paper
|id=Vol-3745/paper17
|storemode=property
|title=Identification of Core Technological Topics in the New Energy Vehicle Industry: The SAO-BERTopic Topic Modeling Approach Based on Patent Text Mining
|pdfUrl=https://ceur-ws.org/Vol-3745/paper17.pdf
|volume=Vol-3745
|authors=Jianxin Zhu,Yutong Chuang,Zhinan Wang,Yunke Li
|dblpUrl=https://dblp.org/rec/conf/eeke/ZhuCWL24
}}
==Identification of Core Technological Topics in the New Energy Vehicle Industry: The SAO-BERTopic Topic Modeling Approach Based on Patent Text Mining==
https://ceur-ws.org/Vol-3745/paper17.pdf
Jianxin Zhu1,2, Yutong Chuang1, Zhinan Wang1,2, ∗, Yunke Li1
1 Harbin Engineering University, School of Economics and Management 150001, China
2 Key Laboratory of Big Data and Business Intelligence Technology, Ministry of Industry and Information
Technology
Abstract
In the new energy vehicle industry, precise identification of core technologies is the key to
promoting innovation and maintaining market competitiveness. In this article, a comprehensive
approach combining information weight method and SAO-BERTopic topic model is proposed to
extract and analyze core technologies from large-scale patent data. Through in-depth analysis of
Incopat patent database, we use the information weight method to select high-quality core patents
from four dimensions: technological, strategic, law and market value. These selected patents form
the basis of the research data, which is then applied to the SAO-BERTopic model, which combines
the advanced semantic understanding capabilities of BERTopic with the fine-structured
characteristics of SAO analysis, greatly improving the efficiency and accuracy of identifying
complex technical topics. This innovation of this research is not only applicable to the analysis of
the technological development of the new energy vehicle industry, but also can provide valuable
reference for other high-tech industries such as biomedicine, renewable energy and information
technology. These fields also need to identify core technologies from a large number of patents.
SAO-BERTopic's structured analysis framework can help these industries to insight into technology
development trends, identify innovation opportunities, and provide data-driven decision support
for enterprises, research institutions and governments, thus playing an important role in
technology planning and market commercialization.
Keywords
technical identification, patent analysis, new energy vehicle, SAO-BERTopic
China has made remarkable progress in this
field[2,3]. However, in the face of the
In a globalized economic environment, strategic layout and potential containment of
the rapid development of the new energy traditional automobile powers such as the
automobile industry has become an United States, Europe, Japan and South Korea
important symbol of technological in terms of technology and industrial chain,
innovation and industrial transformation[1]. the sustainable growth and competitiveness
of China's new energy automobile industry
presents new challenges [4,5,6,7].
Joint Workshop of the 5th Extraction and Evaluation of Knowledge
Entities from Scientific Documents and the 4th AI + Informetrics
(EEKE-AII2024), April 23~24, 2024, Changchun, China and Online
zhjx@vip.163.com (Jianxin Zhu); chuangyutong@hrbeu.edu.cn
(Yutong Chuang); wzn6768@163.com (Zhinan Wang);
liyunke1209@163.com (Yunke Li)
© Copyright 2024 for this paper by its authors. Use permitted under
Creative Commons License Attribution 4.0 International (CC BY 4.0).
CEUR
ceur-ws.org
Workshop ISSN 1613-0073
Proceedings
116
In the research of technological progress patents from large-scale data in the patent
and innovation, patents and scientific papers analysis of the new energy vehicle industry
are indispensable resources. However, due to for the first time, and improve the efficiency
the large scale of these literatures, traditional and accuracy of technology identification.
manual analysis methods are difficult to cope
with, and expert analysis is subjective.
Therefore, based on the large amount of We adopt an innovative hybrid approach,
technical information contained in patents utilizing a phased scientific process to
[8], many scholars have devoted themselves identify the core technological topics within
to using data mining methods in recent years the new energy vehicle industry. This
[9]. Among them, the topic model is an process is divided into six detailed steps (as
approach that can automatically extract key shown in Figure 1), designed to ensure the
topics from documents, revealing technology efficient and systematic extraction, analysis,
trends and areas of innovation. Latent and determination of core technologies from
Dirichlet Allocation (LDA) is a word a broad range of patent data. After selecting
frequency-based models that may struggle to patents related to core technologies using the
capture semantic complexity [10,11,12,13]. information weight method from a large pool
Latent Semantic analysis (LSA) has of patents, the SAO-BERTopic model is
limitations in handling word sense diversity applied for an in-depth analysis of the
and polysemy, which can lead to information selected patents. This model identifies the
loss when analyzing specialized technical core technologies' SAO triplets and defines
documents [14,15]. Correlated Topic models these triplets as specific topics.
(CTM) high computational complexity limits Step One: Patent Selection — The first
its rapid application on large document step is to use Incopat patent database to
collections [11,16] . download patents, and use the information
In view of the limitations of existing topic weight method to establish indicators in four
models in the identification of core dimensions of patent technical, strategic,
technologies in the new energy vehicle legal and market value; Step Two: Textual
industry, we propose an innovative topic Vector Representation — This step is text
model— SAO-BERTopic. The model preprocessing and then vectorization via
combines SAO analysis and BERTopic, an BERT(Bidirectional encoder representation
advanced natural language processing from the transformer) model; Step Three:
technology, to improve the accuracy and Dimensionality Reduction and Clustering —
efficiency of technology recognition. In this step, UMAP(Uniform manifold
The main innovations of this article are: Approximation and projection) algorithm is
1. By combining SAO analysis and used to reduce the dimensionality of the text
BERTopic topic model, we construct SAO- vector, and then HDBSCAN(hierarchical
BERTopic method, which effectively solves densi-based applied spatial clustering with
the problem of identifying deep-level noise) algorithm is used to cluster the topic;
technology trends and key innovation points Step Four: SAO Triplet Extraction — Using
in the new energy vehicle industry. the C-TF-IDF(class-based word frequency-
2. By using the combination of the inverse document frequency) algorithm, SAO
information weight method and the SAO- triples are extracted from the clusters and the
BERTopic model, we realize the selection of topic representation is refined using the
highly innovative and market-valued core MMR(Maximum marginal correlation)
117
algorithm to balance correlation and integration with electric vehicles. Topic 3 is
diversity; Step Five: Result Validation — The "Electric Vehicle Battery Pack Configuration
Calinski-Harabasz index and Davies-Bouldin and Structural Design", covering various
index were used to compare the LDA, aspects of the physical configuration,
BERTopic and SAO-LDA topic models to structural design of battery packs, and how
evaluate the clustering effectiveness. In these designs impact battery pack
addition, dimensionality reduction performance. Topic 4 is "Cathode Materials
visualization via UMAP provides an intuitive and Composition for Secondary Batteries",
comparison of cluster distributions; Step Six: focusing on the composition and design of
Strategic Recommendations — Based on the cathode materials for batteries, especially
results, recommendations are made from the secondary batteries, which directly impact
corporate, industry and governance battery performance aspects like capacity,
perspectives. energy density, charge/ discharge speed, and
Massive patents in the new
energy automobile industry
Core technology patent
text
Generating Topic
Clusters using
lifespan.
HDBSCAN
Information Patent technical value POS tagging
Patent strategic value Tokenization
weight method preprocessing
Patent legal value Stop words filtering
Patent market value Lemmatization
Dimensionality
Text Vectorization with
Core technology patent Reduction of Text
BERT
Vectors using UMAP
Step1
Patent Screening
Step2
Vector Representation of
Text
Step3
Cluster Representation of
Documents
This research successfully applied the
information weight method to screen high-
Extracting SAO Triples
LDA BERTopic
Technology
quality core patents from a large-scale patent
dataset, and employed the SAO-BERTopic
from Topic Clusters SAO- Identification
using C-TF-IDF SAO-LDA BERTopic Suggestions to
Calinski harabasz index
model to identify core technological topics
Index comparison
method Davies bouldin index
Enterprise
Visual comparison Dimensionality reduction
method visualization Industry
Adjusting Topic
from the filtered data. Our empirical analysis
Comparison of Government
Representation using
Clustering Results
MMR
Step5 Step6
results highlight the superior clustering
Step4
Topic Verification Put forward suggestions
Topic Results
Figure 1: Overall research process performance of SAO-BERTopic in identifying
technological topics in the new energy
vehicle domain compared to traditional
This article uses the Incopat patent models. The approach presented in this
database to collect data pertinent to the new research, through its efficient processing of
energy vehicle industry, the topics are finally technical terms and complex concepts and its
refined into five core topics. precise screening of patent data,
Topic 0 is "Energy Management and demonstrates the ability to identify core
Power Transmission Technology in Hybrid technologies and drive innovation across a
Electric Vehicles". Topic 1 is "Electric Vehicle wide range of fields. Therefore, this research
Charging Systems and Energy Management", not only has a direct contribution to the
where the SAO triples contained in topic 1 technological progress of the new energy
focus on describing the functionalities of vehicle industry, but also provides a new
receiving, managing, and providing energy in perspective and tool for technology
charging panels and systems. Topic 2 is identification and innovation management in
"Battery System Integration and Energy various fields, which is helpful to promote
Efficiency Management", where the SAO the technological development and
triples collectively depict various aspects of innovation strategy formulation in a wider
energy storage system design, management, range of fields.
and application, including modular
composition, energy control and
management, system power supply, and
118
[9] N. Su, Z. Tan, Review and vision for the
future of the research on technology
[1] Y. Li, The impact of economic systems opportunity analysis methods, Inf. Stud.
and financial systems on new energy Theory Appl. 43 (2020) 179–186.
vehicle industry, Adv. Econ. Manag. [10] D. M. Blei, A. Y. Ng, M. I. Jordan, Latent
Polit. Sci. 33 (2023) 162-170. dirichlet allocation, J. Mach. Learn. Res.
doi:10.54254/2754-1169/33/20231622. 3 (2003) 993-1022.
[2] Z. Liu, H. Hao, X. Cheng, F. Zhao, [11] T. L. Griffiths, M. Steyvers, Finding
Critical issues of energy efficient and scientific topics, Proc. Natl. Acad. Sci. U.
new energy vehicles development in S. A. 101 (2004) 5228-5235.
china, Energy Policy 115 (2018) 92-97. doi:10.1073/PNAS.0307752101.
doi:10.1016/J.ENPOL.2018.01.006. [12] J. Lafferty, D. Blei, Topic models, in: A.
[3] P. Yu, J. Zhang, D. Yang, X. Lin, T. Xu, N. Srivastava and M. Sahami (Eds.) Text
The evolution of China's new energy mining, Chapman and Hall/CRC, New
vehicle industry from the perspective of York, NY, 2009, pp. 71-93.
a technology-market-policy framework, doi:10.1201/9781420059458.CH4.
Sustainability 11 (2019) 1711. [13] M. Hoffman, F. Bach, D. Blei, Online
doi:10.3390/SU11061711. learning for latent dirichlet allocation, in:
[4] M. Kendall, Fuel cell development for 24th Annual Conference on Neural
new energy vehicles (NEVs) and clean Information Processing Systems 2010,
air in china, Prog. Nat. Sci. Mater. Int. 28 NeurIPS, New York, NY, 2010, pp. 856–
(2018) 113-120. 864.
doi:10.1016/J.PNSC.2018.03.001. [14] S. Deerwester, S. T. Dumais, G. W.
[5] T. Yang, C. Xing, X. Li, Evaluation and Furnas, T. K. Landauer, R. Harshman,
analysis of new-energy vehicle industry Indexing by latent semantic analysis, J.
policies in the context of technical Am. Soc. Inf. Sci. 41 (1990) 391-407.
innovation in china, J. Clean. Prod. 281 doi:10.1002/(SICI)1097-
(2021) 125126. 4571(199009)41:6<391::AID-
doi:10.1016/J.JCLEPRO.2020.125126. ASI1>3.0.CO;2-9.
[6] X. L. Xu, H. H. Chen, Exploring the [15] T. K. Landauer, P. W. Foltz, D. Laham,
innovation efficiency of new energy An introduction to latent semantic
vehicle enterprises in china, Clean analysis, Discourse Process 25 (1998)
Technol. Environ. Policy 22 (2020) 1671- 259-284. doi:10.1080/01638539809545028.
1685. doi:10.1007/S10098-020-01908-W. [16] J. Lafferty, D. Blei, Correlated topic
[7] S. Chen, Y. Feng, C. Lin, Z. Liao, X. Mei, models, in: Proceedings of the 18th
Research on the technology innovation International Conference on Neural
efficiency of China's listed new energy Information Processing Systems, MIT
vehicle enterprises, Math. Probl. Eng. Press, Cambridge, MA, 2005, pp. 147–
2021 (2021) 6613602. 154. doi:10.5555/2976248.2976267.
doi:10.1155/2021/6613602.
[8] C. Lee, A review of data analytics in
technological forecasting, Technol.
Forecast. Soc. Change 166 (2021) 120646.
doi:10.1016/J.TECHFORE.2021.120646.
119