Development of a dedicated software for the treatment of infants with phonological disorder, through the application of natural language processing algorithms.
Limited access to speech therapy in rural areas of Ecuador affects children's phonological development. This study presents "FoneKid", a system based on natural language processing (NLP) that utilizes recurrent convolutional neural networks (CRNN) for speech recognition and deep neura...
Saved in:
| 主要作者: | |
|---|---|
| 格式: | article |
| 出版: |
2025
|
| 主題: | |
| 在線閱讀: | https://repositorio.espe.edu.ec/handle/21000/43113 |
| 標簽: |
添加標簽
沒有標簽, 成為第一個標記此記錄!
|
添加標簽 | 總結: | Limited access to speech therapy in rural areas of Ecuador affects children's phonological development. This study presents "FoneKid", a system based on natural language processing (NLP) that utilizes recurrent convolutional neural networks (CRNN) for speech recognition and deep neural network-based models specialized in phonological processing for phonological analysis. With 94% accuracy, the system dynamically adapts exercises using reinforcement learning. Developed in Unity and supported by Google cloud storage, "FoneKid" enables real-time evaluation and generates intelligent feedback based on phonetic patterns. Its modular architecture optimizes processing on mobile devices with intermittent internet access. Gamification enhances child retention, achieving a 95% interaction rate and a usability score of 84.2 on the SUS Scale. Results from tests with 22 children demonstrate significant improvements in phonetic articulation, with a 30% increase in correct pronunciation after two weeks of use. "FoneKid" emerges as a scalable alternative for speech therapy in resource-limited environments, combining artificial intelligence and playful strategies to optimize phonological learning. |
|---|