Zheng Xinle, Wu Qing, Luo Xingyu, Lv Kun
Business School, Ningbo University, Ningbo, China.
Merchants' Guild Economics and Cultural Intelligent Computing Laboratory, Ningbo University, Ningbo, China.
PLoS One. 2025 Jun 5;20(6):e0324933. doi: 10.1371/journal.pone.0324933. eCollection 2025.
As the wave of technological innovation propels national development into the future, technological advancement has emerged as a crucial pillar for enhancing international competitiveness. Unraveling the evolutionary trajectory of technologies and their associated themes provides a solid theoretical foundation for strategic decision-making in fostering future industrial technological upgrades, thereby aiding in seizing the initiative in technological innovation. This study, adopts a multi-source data perspective, employing the life cycle theory to delineate temporal windows. We use the BERTopic model to extract technological themes and construct a co-occurrence network of theme keywords. Three network centrality indices are computed to filter key theme terms, and the Word2Vec model is leveraged to calculate cosine similarities. Ultimately, we map out the evolutionary pathway of technological themes using Sankey diagrams. Taking the field of artificial intelligence as an example, the study found that the proposed method could effectively identify 48 technical theme keywords and analyze the technological evolution paths of these keywords in areas such as scenario applications, network services, human-computer interaction, intelligent detection, and natural language processing. Furthermore, all evaluation metrics of the model outperformed those of comparable topic models. The rationality of the empirical results was validated through examination against national policies and market application scenarios.
随着技术创新浪潮推动国家发展迈向未来，技术进步已成为提升国际竞争力的关键支柱。揭示技术的演化轨迹及其相关主题，为推动未来产业技术升级的战略决策提供了坚实的理论基础，从而有助于在技术创新中占据主动。本研究采用多源数据视角，运用生命周期理论来划定时间窗口。我们使用BERTopic模型提取技术主题并构建主题关键词共现网络。计算三个网络中心性指标以筛选关键主题词，并利用Word2Vec模型计算余弦相似度。最终，我们使用桑基图描绘技术主题的演化路径。以人工智能领域为例，研究发现所提出的方法能够有效识别48个技术主题关键词，并分析这些关键词在场景应用、网络服务、人机交互、智能检测和自然语言处理等领域的技术演化路径。此外，该模型的所有评估指标均优于可比主题模型。通过对照国家政策和市场应用场景进行检验，验证了实证结果的合理性。
