Developing a tool for automatic transcription and analysis of children’s language samples
Yvonne Wren, Bristol Speech and Language Therapy Research Unit, North Bristol NHS Trust/University of Bristol, UK (yvonne.wren@bristol.ac.uk)
Rebecca Bright, Therapy Box, UK (rbright@therapybox.co.uk)
Swapnil Gadgil, Therapy Box, UK (sgadgil@therapybox.co.uk)
Cristina McKean, University of Newcastle, UK (Cristina.mckean@newcastle.ac.uk)
Geraldine Bates, Sirona Care and Health, UK (geraldine.bates@nhs.net)
Sam Harding, Bristol Speech and Language Therapy Research Unit, North Bristol NHS Trust (samantha.harding@nbt.nhs.uk)
Miriam Seifert, Bristol Speech and Language Therapy Research Unit, North Bristol NHS Trust(miriam.seifert@nbt.nhs.uk)
Background: Language sampling provides speech-language pathologists (SLPs) with information about a child’s use of spoken language within a naturalistic communication environment. However, a survey of 257 SLPs in Australia showed that language samples were generally short, often not recorded, and analysed informally, meaning that management decisions are being made on insufficient data. The main barrier to more detailed language sample analysis is time.
Aim: To determine acceptability to children and parents of new technology which provides automatic language sample transcription and analysis.
Method: The app Language Explorer, developed using machine learning, records narrative samples of children retelling a story. A citizen science approach was used to promote public participation in the collection of data for both automatic and manual transcription and analysis to determine the reliability of the tool. Parents using the app were asked to complete a survey on their experiences of using the app.
Results: Over 1000 parents downloaded the app and contributed narrative language samples and 432 of these completed the survey. Children using the app were aged between less than 3 and up to 6-years-old and most had typically developing speech, language and communication skills (91%). The majority of parents reported that the app was easy to use (95%) and that their child enjoyed using it (91%). Comments from parents on the survey were overwhelmingly positive but included some suggestions for how the app could be improved.
Conclusions: Technology has the capability to significantly enhance our methods for assessment in the field of SLP. These preliminary findings show that a new app, which provides automatic transcription and analysis of language samples, is acceptable to parents of young children. Work is ongoing to determine the reliability of the tool and to investigate its usefulness with a clinical population of children with speech, language and communication needs.
Implications for children: We have developed a new app which asks you to retell a story after you have heard it. We record you as you tell us the story and use that recording to see how we can help you with your talking.
Implications for families: We have developed an app which will record your child as they retell a story. We will use that recording to make measurements of their language and to work out what areas they need help with. We will also use this app to measure change in their language skills following intervention from a speech language pathologist.
Implications for practitioners: Language sampling is challenging because of the time required for transcription and analysis. A new app, Language Explorer, has the potential to significantly reduce the time this takes. Initial investigations have found that this app is acceptable to parents and children.
Funding: UK National Institutes of Health Research
Key words: children’s voices, families’ voices, workforce issues, communication, community services
This presentation relates to the following United Nations Sustainable Development Goals:)
