The use of artificial intelligence is nothing new for Egis. For several years now, our Group has been carrying out and implementing projects involving AI. Our activities already make use of certain AI capabilities, including the very promising Natural Language Processing technologies.
Natural language processing is one of the most developed areas of AI in the last twenty years. It is a multidisciplinary field involving computer science, linguistics and artificial intelligence, with the aim of creating tools capable of interpreting and synthesising text for various applications.
NLP is increasingly used by companies and organisations of all sizes, as it enables them to analyse and understand human language and propose appropriate responses.
NLP has many applications in business. It is used, for example, in marketing to identify potential buyers by analysing their online behaviour. Chatbots also use this technology to manage standard tasks such as providing information on products or services, answering questions, etc. NLP is widely used for text classification, character recognition, automatic correction and automatic summarisation.
How does NLP work?
There are two key stages in any NLP project: linguistic pre-processing and the application of Machine Learning or Deep Learning models.
Pre-processing includes cleaning, standardisation, tokenisation (dividing the text into smaller units called tokens), stemming (the process of reducing inflected words to their root), lemmatisation (isolating the canonical form of the word), and other operations to transform the text into a usable dataset.
The text is then converted into digital data. Various approaches exist, such as TF-IDF: a statistical measure used to evaluate the importance of a term contained in a document, relative to a corpus.
The learning stage involves applying one of three types of approach: either methods based on domain-specific rules, classic machine learning models that use statistical learning methods, or deep learning models. As a reminder, deep learning uses artificial neural networks to learn and perform complex tasks. It is an approach based on machine learning where deep learning models are able to extract hierarchical features from unstructured data, such as images, text or sound. These models are trained on large datasets to recognise patterns and make accurate predictions.
How in practice does Egis exploit the capabilities of natural language processing?
Egis goes further than NLP by using NLU (Natural Language Understanding, a branch of NLP). The difference is that NLP focuses on the literal interpretation of what a person says or writes, whereas NLU takes into account the intention and deeper meaning of what is said or written.
This technology, which can be used for a wide range of tasks from text classification to entity recognition, is used by Egis to extract concepts from text corpora and apply different treatments to them. Here are some examples of use cases:
A use case applied to project management
The eTag project, carried out in partnership with Fieldbox, a company specialising in industrial AI, was set up last summer by the Energy & Sustainable Cities Business Line with the aim of speeding up the reading and analysis of a building programme and automatically extracting the essential requirements.
Today, the platform offers the capability of tagging and sorting into specialities, providing a common interpretation and possibilities for annotation, prioritisation and validation between the project manager and the specialist. It then aggregates a list of these requirements and annotations in csv format.
eTag is a valuable tracking tool for the project manager. It helps to clarify requirements in exchanges with the project owner, the architect and all the competencies interfacing with Egis.
In the role of a productivity booster, this list will used as "input data " in the project management solution. The project leaders at Egis are Xavier Davy and Laurent Fuhs.
Another use of AI: mass-producing LCAs
The Energy & Sustainable Cities Business Line uses NLU to automate Life Cycle Analyses (LCAs) with the solution from SustainEcho, a start-up combining digital technology and environmental assessment which recently joined the Egis group (Contacts: Anatole Parre / Paul Lieberherr).
The LCAs are "mass produced" by simply "dragging and dropping" the contractors’ estimates. The platform can then be used to visualise and highlight the most emissive materials, and interact to find ways of optimising them.
In other words, by using artificial intelligence, SustainEcho can automate the carbon assessment of structures directly from the quantity surveys. This new tool enables Egis to develop an engineering and operating proposition that effectively meets the challenges of climate change, for less carbon-intensive equipment and infrastructure.
More and more AI-based technological solutions are emerging, and Egis remains at the forefront of innovation. These solutions are not limited to natural language processing. Many other artificial intelligence capabilities are being used. We will have the opportunity to come back to them...