Nuclear Data. Chapter 7

The production of high-quality atomic and nuclear data involves compilation, evaluation and validation. Nuclear and atomic reaction and transport models are an essential part in the evaluation and validation process. The section discusses how ML is already being used to support these activities. In the compilation process, publications with relevant measurements need to be identified, the data extracted from text and tables, and added to nuclear databases. For instance, the EXFOR library is a prominent nuclear database that contains cross section data and other nuclear reaction quantities, along with their metadata and uncertainties. This database is maintained and developed under the auspices of the IAEA by the International Network of Nuclear Reaction Data Centers (NRDCs). At present, these activities are performed manually. However, the use of natural language processing to find relevant literature and facilitate data extraction is under development. An example in the general nuclear domain is given. Obstacles to automation are the specific vocabulary used in nuclear physics, such as the notation to identify reaction systems, e.g., 181Ta(n,2n) and reports written in various languages. Furthermore, some historical reports need to be made text-searchable first by using Optical Character Recognition software and errors are usually introduced in the conversion process.