1,720,961 research outputs found
LearnAIng: Generative Artificial Intelligence to boost teaching and training in technical field
Full text linkGenerative Artificial Intelligence is revolutionizing the field of education, offering innovative tools to support students in the learning process and teachers in their activities. This paper presents an AI-powered software developed to support the teaching of systematic innovation courses, following TRIZ methodology, through the integration of Large Language Models (LLMs), Retrieval Augmented Generation (RAG) techniques and the access to patent source. The article describes the methodology implemented and the results of an experimental phase conducted on a large sample of students. The analysis shows how the use of AI in technical Problem-Solving enhances a more structured approach but at the same time is more effective in stimulating creativity and lateral thinking, reducing psychological inertia and boosting Technology Transfer. The findings highlight significant improvements over traditional didactics methods, both in learning effectiveness and instructional support
The Use of AI to Classify Sustainability Patents About AM Metal Powders According to EU Environmental Objectives
No full textAwareness of the issue of environmental sustainability of industrial products and processes is becoming increasingly topical, especially in EU countries, and with it the need to assess and certify the environmental impact of production activities. Among the most popular practices are methods based on life-cycle assessment (LCA), both current and prospective, which require the research of a vast amount of data, accurate, and up-to-date. Whereas in the past the scientific community relied on data reported in the scientific literature and in European projects, there is now increasing recognition of the importance of including patent information, which provides a more detailed view of the scalability of processes, use and end of life of a product. One of the main challenges in making the best use of patents as a source of technical data is managing the complexity of the database and language structure of the documents themselves. A crucial step in facilitating access to this data is through the accurate identification and classification of patents dealing with sustainability to circumscribe the search to a more delimited and accurate scope. Although the criteria for classification are well defined by European regulations, the techniques for actually operating this classification are still highly inefficient. In this paper, we compared traditional patent search methods by keywords and by using Cooperative Patent Classification, in terms of precision and recall, then proposing as an alternative the use of AI tools, such as next-generation NER using a bidirectional transformer encoder and RAG, to overcome the limitations that exist with traditional approaches. An exemplary case is conducted on a pool of patents on metal powders for AM
Discovery Omnia: Dynamic RAG for Enhanced Patent Analysis and Systematic Innovation
No full textAccess to structured and reliable knowledge sources is a crucial step in any innovation process. While general-purpose Large Language Models (LLMs) excel in language generation, they often struggle with factual consistency and precision in specialized technical domains. This research investigates how integrating LLMs with algorithmic routing and structured databases, particularly in the field of patent analysis, can mitigate issues related to hallucinations and improve targeted knowledge retrieval. This information can then be leveraged to carry out specific strategic tasks, such as extracting Patent Intelligence data to derive business insights and support decision-making, conduct a prior art search to secure a patent FTO, enhance Technology Transfer in-domain or cross-domain, but also boosting problem-solving activities and systematic innovation with TRIZ.
This research focuses on the development of a hybrid framework where natural language queries are dynamically reformulated through a routing mechanism, optimizing knowledge retrieval. This method improves the accuracy and relevance of extracted information by systematically addressing query categorization and structured search methodologies. To be able to answer such diverse goals, it is not enough to use a conventional RAG structure, but some hints will be shown on how to design a “Dynamic RAG” nowadays, which means that it is populated each time according to a complex system of translating, analyzing, expanding and routing the user question, select the appropriate source and query to retrieve the information needed and construct the final answer by AI
Investigating the Impacts of Misspellings in Patent Search by Combining Natural Language Tools and Rule-Based Approaches
Full text linkAmong all sources of technical information, patent information is one of the richest and most comprehensive. Knowing how to search in this mass of documents is becoming increasingly crucial. However, many users have limited knowledge of patents and search strategies, so they must use intuitive, often approximate approaches that can lead to highly inaccurate searches and be timeconsuming. To address this problem, there are tools that help expand queries to increase recall so as not to miss good documents, however, it remains an open problem dealing with misspellings based strategies. Typically, the problem of the presence of misspellings in patent text is underestimated even by experts in the field, and there is no specific functionality to handle it in the tools available, both free and paid. The goal of the article is to raise awareness about the difficulties in making a proper patent strategy that also takes into account the possible presence of misspellings. It is important to know where we expect to find them and how much these may affect the final result. In particular, it is chosen to divide misspellings into categories, distinguishing between misspellings associated with a generic keyword or multiword from misspellings in acronyms, chemical formulas, names of applicants, inventors, or names of specific formulas or theorems. At least one example case is given for each category, showing when and how it may affect the result. Finally, an integrated approach combining word and contextual embedding models based on deep learning with a rule‐based algorithm based on wild cards and truncation operators is suggested for correcting the query, automatically suggesting the most consistent misspellings, thus achieving a more accurate and reliable result
AI Based Pointer to Geometric Effects
No full textThe aim of this work is to rediscover one of TRIZ’s best-known tools, the Pointer to geometric phenomena and effects. It is part of a large group of pointers that also includes that to physical effects, to chemical effects and to specific technologies. They were introduced by Altshuller as direct tools for solving a problem. Easy to use but difficult to construct, these tools largely remained on paper, with the exception of the pointer to physical effects, which instead has had numerous software implementations, including recent ones. Studies on the geometric one, on the other hand, are stuck in 1989, in the never-translated book by Vikentiev and Yefremov. The authors’ work started from there, translating it from Russian, recovering what had already been done and developing it at its weakest points. In the absence of a rigorous theoretical basis, they worked on a definition of geometry and the nature of the relationship between the geometry of an element of the system and its functionality. The concept of multilevel shape and topology was introduced, and material was added to the list of pointers. For the construction of the libraries that enable the pointer to function, a systematic working methodology was developed that benefited from the latest text mining technologies, including large language model, named entity recognition, syntactic parsers, and others. These were appropriately combined to recognize design features and geometries from natural language documents taken from patents. In this paper we show both the methodological path to build the library and a demonstration of how to integrate libraries of geometric effects within dynamic pointers to existing physical effects
On Opportunities and Challenges of Large Language Models and GPT for Problem Solving and TRIZ Education
No full textThe advent of GPT has caused a real revolution in many application contexts. Even the TRIZ community has had to face up to this new technology, questioning the possible integrations with traditional paths and tools. Many problem-solving experts have for some time been proposing specific prompts based on the methodology’s tools such as functional analysis, reconstruction of cause-effect relationships, identification of Resources, 40 inventive principles, etc., in order to support the problem solver, or even replace him altogether, during the inventive process. The free generation of LLM content has been applied for very different purposes such as, for example, to contextualize general purpose heuristics in specific domains, or as a search engine to answer technical questions, to suggest creative ideas or improve the formulation and redefinition of a problem, or finally to find connections between different application contexts. This article proposes a critical analysis of the real effectiveness of these prompts according to the different needs of users. The analysis was carried out using a software application that was developed in-house and for which a testing phase was conducted on a variegated sample covering both the academic and industrial fields, with more experienced users and users who have been approaching TRIZ for less time
AI Based Search Engine to Deploy a TRIZ Pointer to Chemical Effects
No full textPointers to effects are a group of TRIZ tools which helps the inventor to master a greater knowledge of scientific phenomena and laws, so to suggest him or her different directions to reach possibilities of solution. Pointers to geometric, chemical, and technological effects have been theorized, but only those to physical effects have ever had concrete developments at the research level and as commercial applications.
The aim of this work is twofold: on the one hand, to bring the pointer back to chemical effects (CE), recovering little-known texts that are difficult to find but also difficult to interpret, as they have never been translated from Russian. The other aim is to contextualize these tools in the light of the recent achievements of artificial intelligence technologies in the field of information retrieval. A combination of AI tools, as NER (Named Entity Recognition), RAG (Retrieval Augmented Generation) and LLM (Large Language Model) have been combined in order to identify chemical features from several chemical sources, to index documents in order to answer user’s questions, to interact with this Knowledge-Base by a chatbot and finally to generate a complete and standardized output.
A comparison is presented between recent commercial applications of AI and traditional pointers to CE from TRIZ literature. In this paper it is explained how the system works, which are the potentialities according to the AI technologies evolution and a comparative study between a SW infrastructure developed by the authors in collaboration with university spin-off software house and others current AI commercial players like GPT or Gemini based applications
Strategic Potential of Patent-Based RAG Systems for Industrial R&D Applications: A Comparison with General-Purpose LLMs
No full textLarge language models (LLMs) have rapidly transformed how information is accessed and generated across domains by leveraging deep learning to produce human-like responses. Their applications have become powerful in supporting coding, law, medicine, and design tools. However, despite their capabilities, LLMs often suffer from critical limitations such as hallucinations, lack of domain specificity, and reliance on generalized internet-based knowledge. These limitations could pose risks for industrial research and development (R&D), where precision and innovation are essential. This study investigates the potential of a patent-based Retrieval-Augmented Generation (RAG) tool (Omnia) to support R&D activities more effectively than general-purpose LLMs (Google AI Studio). Omnia accesses patent databases in real-time to provide structured and validated data, offering reliable and domain-specific responses. Multiple research questions were generated to evaluate the responses from both Omnia and Google AI Studio, which are addressed through targeted case studies. Findings demonstrate that patent-based RAG systems can offer significant advantages in R&D scenarios, including semantic search accuracy, TRIZ-based problem-solving, technical failure analysis, prospective life cycle assessment, and identification of circular economy opportunities
- …
