Patent Search Using Triplet Networks Based Fine-Tuned SciBERT

TL;DR

This paper introduces a novel patent search method that fine-tunes SciBERT with a Triplet Network to generate fixed-size patent representations, enabling efficient similarity-based retrieval and outperforming baseline methods.

Contribution

It presents a new approach combining SciBERT and Triplet Networks for improved patent prior-art search accuracy and efficiency.

Findings

Outperforms baseline search methods

Efficient vector similarity computation

Effective patent representation with SciBERT

Abstract

In this paper, we propose a novel method for the prior-art search task. We fine-tune SciBERT transformer model using Triplet Network approach, allowing us to represent each patent with a fixed-size vector. This also enables us to conduct efficient vector similarity computations to rank patents in query time. In our experiments, we show that our proposed method outperforms baseline methods.