Within EDF R&D, the PRISME department provides expertise and support to EDF’s various power generation business lines, particularly on topics related to engineering and information systems. Following the partnership with Semios.ai, Yasmina Quatrain and Uta Hosokawa share their feedback.
What were your motivations behind the project with Semios?
We were looking for a tool that could help us detect ambiguities, and more specifically implicit wording, in our operating procedures for nuclear power plants.
The key challenges were, first, to improve the quality of instructions, and also to contribute to automating the verification of our procedures.
The aim of this project was to automate the detection and characterization of implicit wording.
Why did you choose us?
We had known Semios for a long time, as the company originated from a research collaboration with IRIT and CNAM.
When choosing a partner, what first made the difference was that the Semios software suite seemed more adaptable to our needs compared with other “off-the-shelf” products.
We also appreciated the fact that the Semios team was sensitive to R&D issues and open to carrying out a proof of concept.
Following this partnership, what is your feedback on the collaboration with the Semios team?
The collaboration was very smooth. We greatly appreciated the clarity and transparency of our exchanges: whenever we encountered an issue, we were able to discuss it openly, and each time the Semios team provided a solution or proposed longer-term improvements.
This enabled us to clearly identify both the potential of the solution and its limitations.
The satisfactory results achieved were first and foremost due to this partnership-based approach, which is not about simply using software, but about working together to create value.
In addition, the Semios team adapted well to our working pace, which was highly appreciated.
What is your feedback on the Semios technology and software?
Regarding the existing product — rule checking, vocabulary control, and so on — the results were of high quality, while requiring adaptation to EDF’s specific context.
However, on the new component developed specifically for us, which includes a graphical analysis of operating procedures, we were impressed by the results !
As a result, the solution will soon be tested on other types of operating procedures.
What potential do you see for Semios’ generative AI features ?
The Semios team had given us a choice in how the solution would be developed. We decided to build the proof of concept using classical or symbolic AI, which seemed more robust to us.
That said, we encountered some technical barriers, and Semios’ generative AI features appear to offer promising ways to overcome them.
For example, to help ensure consistency across a set of operating documents, in order to verify compliance with the “specifications” of the overall documentation set.
You have a question ? Contact us.
