Measuring latent combinational novelty of technology

Abstract

Novelty is considered an important driving force of scientific and technological innovation. How to measure novelty has drawn much attention in recent years. The comprehensive measurement of the novelty could help to identify novel patents as soon as possible and reduce the risk of delayed identification of important technologies. This research introduces a comprehensive measure that could identify novel technology using IPC in patents from a knowledge combinational perspective. The existing methods for measuring novelty commonly use the co-occurrence of knowledge pairwise combinations and identify novelty by assessing the new pairings that did not exist. Besides considering the number of direct co-occurrence of knowledge combinations to evaluate novelty, the proposed method integrates indirect link probability and hierarchical similarity in the IPC tree structure. The feasibility of the measure is demonstrated by applying it to the patent data in the field of Artificial Intelligence (AI). Compared with previous measures, the proposed measure could capture the latent distance between knowledge pairings and identify more novel combinations. The relationship between novelty and citations in the AI field shows that: High-novelty/high-conventional patents have a higher average number of citations and a higher probability of being “hit” patents, indicating that novel patents build on prior knowledge have a relatively higher future impact.