DIN, VDE
Optimization of the search process within standards using Natural Language Processing methods
Users of standards documents often have to invest a lot of time in research in order to extract and use relevant information from standards. The large number of potentially relevant standards complicates the research effort. Initiatives such as the Semantic Standards Information Framework (SNIF) offer an initial remedy here and simplify the search. However, applications such as SNIF are based on fixed rules and keywords and are therefore limited to these. The aim of the project is to train a language model based on international and German standards. This pre-trained model can be refined for various applications. One of these applications is the extraction of relevant standard contents for a component according to a product data standard. In the course of the project, a data set is created that contains suitable text passages from relevant standards for a component and its characteristics from a product data standard (e.g. ECLASS). Using this data set, the pre-trained model can be refined so that it learns to extract suitable text passages from standards for a component and its features.
Project partner: IDiS - Initiative Digitale Standards,
Funding provider: DIN, VDE,
Duration: 31.01.2021 - 31.12.2022