AI for nuclear: Information and safety management

Natural Language Processing

The adoption of AI technologies in the nuclear sector has a huge role to play in both ensuring the industry is competitive and in improving safety. Following work to help the nuclear energy sector with the adoption of novel AI technologies, AI for Nuclear has produced a report showcasing the role Natural Language Processing (NLP) can play in modernising the industry.

What does it say?

The AI For Nuclear report highlights how NLP can be utilised to bring benefits including improved efficiency and effectiveness, enhancement of safety, R&D and regulatory compliance and streamlined operational management. The report identifies areas where NLP could be deployed including:

• Decommissioning
• Business support
• Within the nuclear power plant

It also features case studies surfaced during a workshop which offered evidence of the effectiveness of NLP and, by showcasing real-world scenarios and outcomes, they demonstrate how NLP can enhance various processes. Examples showcased include how AI reduced the administrative burden by 90% and how NLP was used to extract information from 40 years’ worth of archives; half a million pages of information was converted into a digital ‘Knowledge Base’ by a propriety software.

These real-world demonstrations of the benefits of AI help to build confidence among stakeholders and decision-makers regarding the potential of NLP to improve operational efficiency, safety and decision-making.

What are the key recommendations?

The workshop and case studies demonstrate that NLP is a pivotal technology capable of bolstering safety measures in the nuclear sector. It offers valuable support in managing a broad spectrum of information-related tasks, including regulatory document handling, public communication, training initiatives and event detection and monitoring.

The delegates did sound a note of caution recognising that, whilst NLP brings demonstrable benefits, tools such as those based on generative AI still have limitations and require human oversight when being used.