A Generative-Artificial-Intelligence-Powered Platform for an Educational Story-teller

Educators believe that using technology in education will be more effective for the new generation of learners. Therefore, we propose an educational technology platform called Hakawati that supports educators’ teaching methods and students’ learning skills by utilizing artificial intelligence capabilities based on text-to-speech (TTS) and text generation technologies. For the story-generation component of Hakawati, we employed a pretrained transformer model called Turbo 3.5. This model creates coherent narratives from various prompts, perfectly satisfying the goal of transforming lessons into stories. To ensure the effectiveness of the models, subjective evaluations were conducted using measures, such as relevance, coherence, story completeness, language, and grammar. To convert the generated stories into speech, Coqui_XTTS version 2 and StyleTTS were evaluated using a subjective mean opinion score for naturalness and understandability. Coqui_XTTS outperformed its counterpart, with the language model scoring 4.44 out of 5, whereas StyleTTS scored 1.7 out of 5. This demonstrates the ability of Coqui_XTTS to generate high-quality natural speech, add an immersive auditory element to the story, and enhance the user experience.