This work analyzes the role of patents in the technology foresight process from both methodological and practical perspectives, contributing to the field of Data Driven Foresight. The central research question is: How can patents be utilized in the technology foresight process efficiently, and which use cases can be addressed accordingly? The main contributions are threefold. First, a comprehensive patent database implemented with a graph database and a full-text search engine, populated with patent data, and structured via an RDF ontology, is developed. Second, a systematic literature review of 26 practitioner-focused technology-foresight use cases relevant for patent analysis is conducted. Third, 23 patent-analysis methods are categorized into three requirement levels (low, medium, high), and a framework to streamline method selection and analysis workflows is introduced. Across the literature and practical examples, patent data are validated as a primary resource for foresight. Simpler techniques, such as descriptive statistics, bibliometrics, and basic text mining, are the most commonly used and typically sufficient for tasks such as surveying technology fields, tracing development trajectories, and spotting early indications of technological opportunity. Advanced approaches, including machine learning and neuronal network analyses, offer deeper insights but are currently less prevalent; their adoption is expected to grow as tools and expertise become more accessible. This study provides use case profiles that specify the most relevant patent data fields and analysis types for each use case, and documents complementary data sources and supporting literature. The work furthermore highlights areas for further research: certain patent data fields and analytical techniques remain rarely used despite their potential for use-case-specific insights. Moreover, the unique characteristics of patent data, such as unusual metadata conventions, heterogeneous standards across patent offices, and variable bibliographic quality, that require careful handling to avoid misinterpretation, are listed. Additionally, the integration of AI, particularly large language models, into patent-analysis workflows is evaluated. LLMs can, for instance, improve efficiency in literature mapping, semantic similarity detection, and automated categorization, thereby enhancing the output and timeliness of the foresight process. In summary, this study offers a practical toolkit for foresight practitioners and decision-makers: guidance for building patent databases, a method-selection framework tied to real-world use cases, and a curated overview of relevant literature and complementary data sources. The next steps for this work are automating literature categorization, expanding explored patent fields and methods, and improving accessibility via interactive online visualizations. Ultimately, this research clarifies the role of patents as a critical data source in technology foresight and paves the way for deeper integration of patent analysis and AI in future foresight practice.
Melanie Martini (Thu,) studied this question.