Firm-Specific Patent Recommendation Using Topology-Based Link Prediction and SAO Embedding Method
提出一种专利推荐方法，平衡公司技术基础与外部知识，通过拓扑链接预测和SAO-Doc2vec模型提升推荐专利的相似性、新颖性和多样性，并在3D打印等领域验证有效性。
In the context of rapidly developing exploratory innovation, relying solely on in-house R&D without external knowledge is impractical, while abandoning existing technological advantages to explore new fields is equally unrealistic. Firms should actively seek external knowledge transfer to drive their own innovation. To address this, this study proposes a patent recommendation method that balances a firm's technical foundation with external knowledge. First, the topology-based link prediction method assesses the firm's technical foundation and identifies candidate patents using external knowledge. Then, the SAO-Doc2vec model analyzes the similarity between candidate and target patents for recommendation. Experimental comparisons with three benchmarks demonstrate that the proposed method outperforms in terms of patent similarity, novelty, and diversity. The method is further applied in two domains, including 3D printing firm Stratasys, the method uncovers undeveloped areas like G05B, which involves control systems. The study also explores the distribution and competition of integrating G05B with 3D printing across countries and firms. Interestingly, an extended case with Lingdong Technology demonstrates the method's effectiveness for emerging firms with underdeveloped patent portfolios. This study provides actionable insights for firms seeking reference patents and supports strategic R&D decisions, helping firms explore innovation opportunities.