As we wrap up the AI for Impact Talent Program, we're thrilled to share the final details of our app, which has been designed to enhance communication and provide valuable information to users in various contexts. Here's a closer look at the features we've developed and how they work:

• Translation Feature

Our app’s translation feature allows seamless communication across languages. Users can speak into the app, which transcribes the speech into text, translates it into the selected language, and plays it back as audio. The entire conversation is visually displayed on the screen in speech bubbles, making it easy to follow and review. A reset button clears the conversation screen, allowing users to start fresh whenever needed. Importantly, users can choose to consent to data collection for statistical purposes, helping us improve the service, while still maintaining full functionality without consent.

• Community Guide

The Community Guide is an invaluable resource for users seeking information about local resources, integration tips, and recommendations for activities and attractions. Questions can be asked either by speaking or typing, and responses are provided both as text on the screen and as audio. The app offers the flexibility to toggle between two selected languages within the conversation, and users can choose to disable the audio if they prefer text-only responses.

• Admin Functionality

For administrators, the app includes powerful tools for analyzing user interaction data:

1. Statistics Dashboard: This feature provides visual insights into the number of conversations related to specific topics. It helps administrators understand user needs and common themes in discussions.
2. Chatbot with RAG Integration: An AI-powered chatbot, enhanced with Retrieval-Augmented Generation (RAG), can answer questions about saved conversations. It uses techniques to generate contextually relevant responses based on the stored conversation data.

• Technical Highlights

Our app leverages AI models through several API integrations:

1. Transcription: The OpenAI Whisper model transcribes spoken audio into text.
2. Translation and Response Generation: GPT-3.5-turbo translates text or generates responses to queries.
3. Text-to-Speech: Microsoft’s Text-to-Speech API converts text back into spoken language.
4. Data Analysis: OpenAI Embeddings API is used for vectorizing conversations, enabling in-depth analysis of user interactions.

• Data Protection and Privacy

User privacy is a top priority. The app’s translation feature allows users to consent to data storage for statistical analysis, but this is entirely optional. The app remains fully functional without data storage, and all other interactions within the app are processed in real-time without saving any data. This ensures that users, particularly those in vulnerable populations, can use the app without pressure to consent to data collection.