Created by: Aza Allsop and Nilesh Arnaiya at Aza Lab at Yale University
Comprehensive qualitative research analysis tool powered by advanced AI models. A complementary tool to your human-level qualitative research pipeline. Customize prompts for optimal results.
Multi-perspective analysis with auto-save and instant downloads. Performs 6 independent runs (one per seed) and compares them using cosine similarity with all-MiniLM-L6-v2 (You can use other Similarity methods if you desire, see code!) for reliability assessment. Customize runs, seeds, and temperature as needed.
GitHub: The whole process and code is open source at our GitHub repository
Azure OpenAI Configured: Endpoint: alright.openai.azure.com | Deployment: gpt-4o | API Version: 2024-12-01-preview
Choose the AI model to use for analysis. Each model has different strengths and capabilities.
Get your API key from Azure Portal (Azure OpenAI Service)
Upload transcript dialogue data and customize prompts to steer the model for best outputs. Available placeholders: {seed} for the run seed, {text_chunk} or {text} for the transcript text. If no text placeholder is used, text is appended at the end. Leave empty to use default prompt.
Controls randomness: 0 = deterministic, 1 = balanced, 2 = creative. Applies to all runs.
Seeds control variation across runs. Each seed creates one analysis run. Use {seed} in your prompt to reference the current seed value (e.g., for "Run ID: {seed}" instructions).
File Size Warning: Do not upload files above 5MB. For larger files, chunk them into multiple smaller files (under 5MB each) and combine the outputs later for best results.