Testbed for developing AI solutions for language learning
Reference number | |
Coordinator | HELSINGBORGS KOMMUN - Vuxenutbildning Helsingborg |
Funding from Vinnova | SEK 498 000 |
Project duration | May 2020 - December 2020 |
Status | Completed |
Venture | Test beds for society´s challenges |
Call | Test beds for the challenges of society - preparation project 2020 |
Important results from the project
Test bed for AI-driven language learning wanted to investigate whether an AI-based language development tool could give newly arrived and foreign-born residents better conditions to learn Swedish in order to be established more quickly in society and in the labor market. During the project, a dialogue application was developed with an associated interface and AI-based dialogue generator. Initial tests and surveys regarding the use of such a tool have given a positive response from the target group.
Expected long term effects
We wanted to create something that could be used anywhere and anytime, whose context can be applied in everyday life for our students and with high usability. During the project, a dialogue application was developed with an associated interface, via which a student can perform tasks together with a dialogue generator on a specific topic to shop in the grocery store. The topic is taken from conversations with SFI students and the application can be accessed via the website. Thus, an application has been developed that can be used anytime and anywhere, and which is rooted in everyday life.
Approach and implementation
During the autumn, a market analysis was carried out, where the company Talkamatic was judged to have the right conditions to develop a dialogue generator suitable for the purpose of this project. Through interviews with SFI students and teacher, a case was chosen for the dialogue generator - the "food store". This was followed by a design sprint to define an associated interface that would suit the target audience well. The last step was to develop the interface and the dialogue generator. This implementation was followed by continuous analysis of what would be needed in a test bed.