A Python Azure Function to process and enrich output from Azure Batch Speech-to-text with a Language Understanding (LUIS) model.
Use this function to parse transcribed text to identify entities and intents from audio.
This repo assumes the user has already set up:
- Clone this repo into your development environment
- Install the required python libraries in your environment by running
make install
- Update the input binding in
BatchTranscriptionLuisEnrichment/function.json
to reference the folder name in your blob storage account where your Batch Transcriptions are written to. - Update the
local.settings.json
with the LUIS secrets and Storage account connection string - Test the function is working by running the function locally and uploading an example transcription using Azure Storage Explorer.
- Check your storage account for the enriched document in the
transcriptions-enriched
folder. You can now publish the function to your Function App.
- The Azure Functions blob storage trigger does not scale to high-throughput scenarios, as acknowledged in the documentation. For scenarios with higher than 100 blob updates per second, Event Grid trigger should be used.
- LUIS currently has a 500 character limit per REST call. Transcribed utterances that are longer than this are currently truncated in the function.