Structural Representation Learning and Disentanglement for Evidential Chinese Patent Approval Prediction

TL;DR

This work targets automatic Chinese patent approval prediction with a requirement for evidential, transparent decisions. It introduces DiSPat, a retrieval-based framework comprising BRR for base references, SPR for structurally encoding claim hierarchies, and DRL for disentangling representations into similarity and specificity to enable evidence-backed decisions. The approach is evaluated on three Chinese patent datasets (A47, C23, F24) showing state-of-the-art performance and improved evidentiality, with ablations confirming the value of structural prior knowledge and disentanglement. The results demonstrate practical impact for faster, explainable patent examination and decision backtracking in real-world settings.

Abstract

Automatic Chinese patent approval prediction is an emerging and valuable task in patent analysis. However, it involves a rigorous and transparent decision-making process that includes patent comparison and examination to assess its innovation and correctness. This resultant necessity of decision evidentiality, coupled with intricate patent comprehension presents significant challenges and obstacles for the patent analysis community. Consequently, few existing studies are addressing this task. This paper presents the pioneering effort on this task using a retrieval-based classification approach. We propose a novel framework called DiSPat, which focuses on structural representation learning and disentanglement to predict the approval of Chinese patents and offer decision-making evidence. DiSPat comprises three main components: base reference retrieval to retrieve the Top-k most similar patents as a reference base; structural patent representation to exploit the inherent claim hierarchy in patents for learning a structural patent representation; disentangled representation learning to learn disentangled patent representations that enable the establishment of an evidential decision-making process. To ensure a thorough evaluation, we have meticulously constructed three datasets of Chinese patents. Extensive experiments on these datasets unequivocally demonstrate our DiSPat surpasses state-of-the-art baselines on patent approval prediction, while also exhibiting enhanced evidentiality.

Figures (7)

• Figure 1: (a) Manual patent examination for patent approval prediction processes; (b) Automated patent examination based on the manual processes; (c) An analysis of the structure within and between patent claims.
• Figure 2: Overall Structure of DiSPat. The blue arrows indicate the transfer of the target patent between modules, and the green arrows indicate the same operation of the base reference patents.
• Figure 3: Performance when replacing or removing certain components from DiSPat on the three datasets. r.p. stands for replace and r.m. stands for remove.
• Figure 4: The results of comparison with different values of $k$, $n$, and $w$ on datasets A47, C23 and F24.
• Figure 5: Impact of different $\mathcal{L}_{con}$ compositions on accuracy in DRL module. The percentage indicates the improvement of the accuracy compared to $\mathcal{L}_{con}=0$.