Exploration of Large Language Model Usage in Computer Technology Patent Texts

Using Large language models (LLMs) to augment human authors in writing professional texts leaves traceable evidence. By examining computer technology patents filed in the United States, this study investigates latent patterns that can be attributed to the widespread adoption of LLMs by finding word tokens that became more prevalent after the release of ChatGPT. After manually reviewing and excluding tokens attributed to inventions' technological elements, we found one token that has been used more frequently after the release of ChatGPT across all parts of patent documents. This research extends the study of LLM's widespread impact on academic manuscripts into examining technical documents where LLMs as an augmentation tool are actively explored. The results of this study show that LLM's impact is observable and quantifiable on the scale of a specific corpus, which has a significant implication for both researchers who study existing documents and practitioners who author new documents.