[Paper Review] A Survey on Patent Analysis: From NLP to Multimodal AI

Homaira Huda Shomee, Zhu Wang+2 more|arXiv (Cornell University)|Apr 2, 2024

Intellectual Property and Patents6 citations

TL;DR

A comprehensive survey of AI-based methods for patent analysis (2017–2023), covering text and image data across classification, retrieval, quality analysis, and generation, with a new multimodal taxonomy and future directions.

ABSTRACT

Recent advances in Pretrained Language Models (PLMs) and Large Language Models (LLMs) have demonstrated transformative capabilities across diverse domains. The field of patent analysis and innovation is not an exception, where natural language processing (NLP) techniques presents opportunities to streamline and enhance important tasks -- such as patent classification and patent retrieval -- in the patent cycle. This not only accelerates the efficiency of patent researchers and applicants, but also opens new avenues for technological innovation and discovery. Our survey provides a comprehensive summary of recent NLP-based methods -- including multimodal ones -- in patent analysis. We also introduce a novel taxonomy for categorization based on tasks in the patent life cycle, as well as the specifics of the methods. This interdisciplinary survey aims to serve as a comprehensive resource for researchers and practitioners who work at the intersection of NLP, Multimodal AI, and patent analysis, as well as patent offices to build efficient patent systems.

Motivation & Objective

• Provide a consolidated overview of AI tools for patent analysis across four main tasks (classification, retrieval, quality analysis, generation).
• Introduce a novel taxonomy for categorizing methods by patent life cycle tasks and AI modalities.
• Summarize key datasets, models, and evaluation metrics used in patent AI research from 2017–2023.
• Highlight challenges, gaps, and future directions including multimodal learning and generative AI for patents.

Proposed method

• Survey and synthesis of 40+ papers from 26 venues (2017–2023).
• Taxonomy development linking tasks (classification, retrieval, quality analysis, generation) with AI methods (traditional ML, neural networks, LLMs, multimodal).
• Organization by patent life-cycle tasks with sections on datasets, methods, and evaluation metrics.
• Discussion of multimodal and generative AI trends for patents.
• Recommendations for future research directions and potential datasets.

Experimental results

Research questions

• RQ1 What AI methods have been applied to patent analysis tasks (classification, retrieval, quality analysis, generation) from 2017 to 2023?
• RQ2 How do text-only and multimodal (text + image) approaches compare in patent analysis tasks?
• RQ3 What datasets, evaluation metrics, and trends are prevalent in AI-based patent analysis research?
• RQ4 What future directions (multimodal learning, generative AI, knowledge graphs) show promise for patent analysis?

Key findings

• AI tools are increasingly used for patent classification, retrieval, quality analysis, and generation.
• There is a shift from traditional ML to deep learning and large language models, with domain-specific variants like SciBERT.
• Multimodal approaches integrating text, images, and metadata are emerging as a research direction.
• A novel taxonomy ties patent life-cycle tasks to AI methods and data modalities.
• Future directions include multimodal datasets, generative patent AI, knowledge graphs, and domain-specific evaluation for generated content.