Knowledge bases and patent examination efficiency in Artificial Intelligence: the roles of knowledge diversity, technology cycle, and firm heterogeneity

Patent examination speed is critical for firms in rapidly evolving AI technology. Drawing on knowledge recombination perspectives, this study examines how knowledge diversity and technology cycles influence patent examination duration and how firm size moderates these relationships. Using patents in the natural language processing (NLP) sector, we employed Cox proportional hazards models with time-dependent coefficients. The results show that knowledge diversity follows an inverted U-shaped relationship with patent examination duration, with moderate levels of knowledge diversity associated with longer examination durations and higher levels associated with shorter durations. A longer technology cycle is associated with a shorter patent examination duration. Furthermore, firm size strengthens the beneficial effects of knowledge diversity and mature technological knowledge on the efficiency of examinations. This study contributes to the innovation economics literature by highlighting the nonlinear effects of knowledge characteristics and the role of firm heterogeneity in explaining variations in patent examination efficiency.