AIDA(AI Dialogic Reading Aid)

Did you know that early literacy plays a pivotal role in future academic success? However, early literacy arises even when kids start reading. Intervention in early age can be most beneficial before these gaps become too large. A child's language development is heavily influenced by their parents, and shared book reading is identified as an effective method to boost literacy in young children. However, the effectiveness of these interventions often varies based on the family's socioeconomic status (SES).

The dialogic reading is identified as one of the effective methods to enhance early reading skills. Dialogic reading ("DR") turns shared book reading into a conversation about the story. However, there are some difficulties posed by dialogic reading. Parents would come up with their own questions on the spot and be considerate about what types of questions to ask and the appropriate time to ask (Shavlik, Margaret). Here’s where the AIDA project steps in. With its strong, proven effect of dialogic reading, AIDA aims to automate the process and be easier and more accessible to any parents regardless of SES.

In the AIDA(A project, my primary responsibilities centered on UI development and prompt engineering. For the UI, I successfully integrated the Whisper API into the front end, ensuring seamless interaction. Further, I implemented a dynamic question generation mechanism using LLM, which tailored questions based on factors like the child's age, the frequency with which a book was read, and the child's reading level. On the prompt engineering side, I worked on evaluating the difficulty of generated questions. This involved considering the child's age and how often they've read the book, using benchmarks like child reading milestones and the mean length of utterance (MLU). Additionally, I delved deep into understanding how the context of the book influenced the nature of the questions generated, ensuring they were relevant and engaging.