Sunday 02 February 2025
The quest for a more efficient and accurate patent drafting process has led researchers to explore the potential of artificial intelligence (AI) in this domain. In a recent study, scientists have developed a novel approach to revising patent claims using large language models (LLMs). The results are promising, showing that AI can improve the quality of patent claims while reducing the time and effort required for drafting.
Patent claims are critical components of patent applications, as they define the scope of an invention and ensure its protection. However, drafting high-quality patent claims is a labor-intensive process that requires expertise in both patent law and technical fields. AI models, on the other hand, can analyze vast amounts of data, identify patterns, and generate text based on those patterns.
The researchers employed several LLMs, including CoEdIT-XL, SaulLM-7B, Mixtral-8×7B, and GPT-4, to revise patent claims. They fine-tuned these models using a dataset of 1,500 patent claim texts and evaluated their performance using both human evaluation and automated metrics.
The results show that AI can significantly improve the quality of patent claims. For instance, CoEdIT-XL achieved an average score of 81.7 in human evaluation, surpassing the copy baseline. The model demonstrated strength in completeness of essential features, conceptual clarity, consistency in terminology, and technical correctness of feature linkages.
The researchers also found that fine-tuning LLMs using a patent-specific dataset can improve their performance. SaulLM-7B, which was trained on a large corpus of legal texts, including patent texts from the United States Patent and Trademark Office (USPTO), achieved an average score of 82.6 in human evaluation.
Automated metrics, such as SARI, BLEU, and ROUGE-L, also showed promising results. G-Eval, a metric developed specifically for evaluating the quality of generated text, demonstrated that AI can accurately identify essential features, clarify complex concepts, and link technical terms.
The study highlights the potential benefits of using AI in patent drafting, including reduced time and effort required for drafting, improved consistency and accuracy, and enhanced collaboration between attorneys and inventors. However, the researchers also acknowledge the limitations of their approach, noting that AI models may struggle with nuances of patent law and the need for human oversight.
As the patent landscape continues to evolve, the potential applications of AI in patent drafting will likely expand.
Cite this article: “Artificial Intelligence Enhances Patent Claim Drafting with Improved Quality and Efficiency”, The Science Archive, 2025.
Artificial Intelligence, Patent Drafting, Large Language Models, Patent Claims, Patent Applications, Natural Language Processing, Intellectual Property, Machine Learning, Patent Law, Patent Texts
