Startup Studio
Software for diagnosing language levels
Date:
Changed on 02/06/2025
The result of multidisciplinary academic research, A4LL software automatically analyzes texts written by students to build a profile of their language level, enabling teachers to personalize their teaching methods. A company project has just joined the Inria Startup Studio program at the Inria Center at the University of Rennes, with the aim of industrializing this prototype and turning it into an operational tool for English teachers.
Verbatim
Our platform supports language teachers in the diagnosis of writing. Students write their text. They click to send it to a Machine Learning pipeline that extracts measurements. These measurements give an estimate of each student's level. This gives the teacher a precise idea of what's going on in the classroom. He can see in detail why a particular student has been placed in level A1, A2, B1, B2, C1... Graphs can also be used to show which students have common characteristics, such as mastery of the simple past or a particular type of vocabulary. This makes it possible, for example, to set up work groups.
Auteur
Poste
A4LL business project sponsor
The application under development follows in the wake of multi-disciplinary work on natural language processing carried out by the University of Rennes 2 (Lidile), Université Paris Cité (Altae), the University of Le Mans (Lium), the University of Galway and UMR Irisa.
This collaboration involved researchers Thomas Gaillat, Nicolas Ballier (linguistics), Rémi Venant (learning computing environment) and Andrew Simpkin (statistics), as well as PhD students Patrick Li and Bernardo Stearns. Cyriel Mallart was the research engineer. Called A4LL (an acronym for Analytics for Language Learning), the project was supported by the French National Research Agency for two years.
An AI trained on the writings of thousands of students
The resulting software uses artificial intelligence trained on the writings of thousands of students at the two universities in Rennes. It produces lexical, syntactic and semantic metrics. A visual representation allows exploration of all these dimensions of language and an estimation of the level corresponding to these characteristics.
"In the end, we had a useful tool that not only worked well, but also appealed to teachers. So we thought it would be a shame to leave this academic prototype in a drawer. We might as well try to push it further and make it widely available."
A year accompanied by Inria Startup Studio
And that's the whole point of the Inria Startup Studio, an entrepreneurial support scheme for digital deep-tech projects.
Starting in January 2025 for a period of one year, this preparation period will comprise two parts running in parallel. "In the first, we'll carry out technological maturation. The software works, but it needs work on ergonomics. For this part, we have recruited Théo Blandin, a UX designer engineer, i.e. a user experience specialist. Next June, we'll be organizing a focus group during which we'll invite teachers who have never tried the solution to come and take the latest version in hand and give us their cold opinions."
To begin with, the project will focus solely on the English language. It should be noted in passing that the software will interface easily with Moodle, a distance learning platform widely used by universities.
The second part aims to refine the business model. "Universities, business schools and engineering schools will be able to acquire the solution in the form of licenses. But, initially, we'll be talking directly to teachers. Because they are our best ambassadors."
Titre