Warning This site is not recommended for Internet Explorer browsers. Please use another web browser to get a better experience.

Natural Language Processing: Safran Engineering Services is developing new digital methods


Natural Language Processing (NLP) is a technology that enables machines to understand human language. It is currently a key area of development and innovation for Safran Engineering Services’ Data department. The idea is to create new industrial applications for the entities making up the Safran group. Guillaume Besseron, head of the Data department and Marc Lutrot, head of innovation at Safran Engineering Services in Saclay discuss this technology and the potential it has for Safran.

Natural Language Processing - Safran Engineering Services

NLP at Safran

Safran's tools, methods and processes are increasingly being digitized, meaning that more and more data is computer-entered and can be processed and then leveraged. This includes free field text boxes and the contents of our emails, for example”, says Guillaume Besseron.
Safran Engineering Services has therefore decided to use NLP in order to make use of this type of textual data: “We can use NLP to perform lexical analysis, extract key information from a document, classify it and summarize it, as well as even generating language in our chatbots. And all that can be done automatically and on a massive scale”, adds Marc Lutrot.

Safran Engineering Services’ packaged offering

From gauging a particular requirement to rolling out analysis tools across business processes, Safran Engineering Services uses a dedicated process underpinned by robust expertise that its Data team has developed using NLP over the last few years. An "active learning" tool for ongoing certification and learning has been developed internally to meet the needs of Safran's companies. It is user-friendly and reduces the number of iterations required, while at the same time significantly improving the way in which AI algorithms learn. Based on the various application scenarios encountered, a dictionary for the Group’s business lines has been designed and improved by Safran Engineering Services. It is yielding helpful results more rapidly, with greater relevance and accuracy than standard dictionaries.

Concrete application scenarios

Given that it works closely alongside the business of Safran companies at their request, Safran Engineering Services has implemented a number of pilot application scenarios. One of them focuses on a particular process: producing electronic cards. “Some of them contain more than a thousand components and operatives can sometimes encounter a problem when they are being assembled – a fault. When they do, they enter the information – in natural language, more often than not – in their management tool”. If the same problem on a particular component were to be reported several times in the tool’s description field, the factory quality team would not be able to detect that it was a recurring problem. “With NLP, we can cross-reference thousands of technical facts, classify them on a per-component basis and crosscheck them. The result is that we can be more responsive by detecting early weak signals, and this helps us to prevent further occurrences. Information compiled this way can therefore be used to identify a solution, decide on corrective or preventive action, or detect underlying trends, all as part of an ongoing approach to improving the performance of our companies”, says Marc Lutrot.

Collaborating effectively with business lines

The algorithms that we create are improved to increase their accuracy. This can only be done with input from operatives working out in the field who are familiar with the products and requirements. We are also aware that for that, we need to take up as little of their time as possible”, says Guillaume. Regarding this particular electronic card application, Safran Engineering Services has therefore applied its efficient process and used the active learning tool. “In a few exchanges with the Safran Electronics & Defense factory team in Fougères, we have already been able to process 2000 technical events which occurred during production, and then identify and automatically classify 90% of faults. This will enable the quality department to take the necessary action and address them in the long run”, says Marc.

Innovation prospects at Safran

Nowadays, there are numerous, varied application prospects for NLP technologies in an industrial group such as Safran. Files containing details of technical production events can be analyzed, client feedback data for our engines and equipment can be analyzed (Customer Support Center), derogations can be analyzed, AOG (Aircraft on Ground) risks can be detected in client requests, and maintenance reports can be analyzed. Furthermore, CVs can be analyzed and invitations to tender can be classified. Essentially, anything can be analyzed and interpreted automatically and on a massive scale. “We can take action in all of these areas to improve business line performance – in production workshops, offices and in support roles. This will mean higher levels of satisfaction for our clients. We are developing our skills, we are intertwining digitization technologies to improve performance and are launching numerous innovation projects to explore solutions for creating value involving new services. And what we enjoy above all else is coming up with these solutions for and with our clients”, says Guillaume.

Safran Engineering Services is involved in the Group’s digital transformation, and on the back of its achievements with NLP applications, it has developed an offering for effectively meeting – in compliance with security standards – all of the Group's needs regarding the leveraging of natural-language text. “We are ready to embrace the challenges laid down by the Group!”, say Marc and Guillaume.