Patent Claim Generation by Fine-Tuning OpenAI GPT-2

TL;DR

This paper explores fine-tuning GPT-2 to automatically generate patent claims, addressing a unique language structure and providing experimental insights and tools for future research in automated patent claim creation.

Contribution

First to generate patent claims with GPT-2, introducing new sampling methods and providing experimental analysis and a tool for further exploration.

Findings

GPT-2 can generate coherent patent claims

New sampling approach improves text diversity

Fine-tuning process insights for patent language

Abstract

In this work, we focus on fine-tuning an OpenAI GPT-2 pre-trained model for generating patent claims. GPT-2 has demonstrated impressive efficacy of pre-trained language models on various tasks, particularly coherent text generation. Patent claim language itself has rarely been explored in the past and poses a unique challenge. We are motivated to generate coherent patent claims automatically so that augmented inventing might be viable someday. In our implementation, we identified a unique language structure in patent claims and leveraged its implicit human annotations. We investigated the fine-tuning process by probing the first 100 steps and observing the generated text at each step. Based on both conditional and unconditional random sampling, we analyze the overall quality of generated patent claims. Our contributions include: (1) being the first to generate patent claims by machines and being the first to apply GPT-2 to patent claim generation, (2) providing various experiment results for qualitative analysis and future research, (3) proposing a new sampling approach for text generation, and (4) building an e-mail bot for future researchers to explore the fine-tuned GPT-2 model further.