python tool.
translator tool from the promptflow-azure-ai-language package.
pii_entity_recognition tool from the promptflow-azure-ai-language package.
abstractive_summarization tool from the promptflow-azure-ai-language package.
extractive_summarization tool from the promptflow-azure-ai-language package.
entity_recognition tool from the promptflow-azure-ai-language package.
sentiment_analysis tool from the promptflow-azure-ai-language package.

Connections used in this flow:

Custom connection (Azure AI Language).
Custom connection (Azure AI Translator).

Prerequisites

Prompt flow SDK:

Install promptflow sdk and other dependencies:

pip install -r requirements.txt

Note: when using the Prompt flow SDK, it may be useful to also install the Prompt flow for VS Code extension (if using VS Code).

Azure AI/ML Studio:

Start an compute session. Required packages will automatically be installed from the requirements.txt file.

Setup connections

To use the promptflow-azure-ai-language package, you must have an Azure AI Language Resource. Create a Language Resource if necessary. From your Language Resource, obtain its api_key and endpoint.

Create a connection to your Language Resource. The connection uses the CustomConnection schema:

Prompt flow SDK:

# Override keys with --set to avoid yaml file changes
pf connection create -f ./connections/azure_ai_language.yml --set secrets.api_key=<your_api_key> configs.endpoint=<your_endpoint> name=azure_ai_language

Ensure you have created the azure_ai_language connection:

pf connection show -n azure_ai_language

Azure AI/ML Studio:

If using Azure AI Studio, you will need to add two additional custom keys to the connection. Follow these instructions when creating a CustomConnection.

To use the translator tool, you must have an Azure AI Translator resource. Create a Translator resource if necessary. From your Translator Resource, obtain its api_key, endpoint, and region (if applicable).

Create a connection to your Translator Resource. The connection uses the CustomConnection schema:

Prompt flow SDK:

# Override keys with --set to avoid yaml file changes
pf connection create -f ./connections/azure_ai_translator.yml --set secrets.api_key=<your_api_key> configs.endpoint=<your_endpoint> configs.region=<your_region> name=azure_ai_translator

Ensure you have created the azure_ai_translator connection:

pf connection show -n azure_ai_translator

Azure AI/ML Studio:

If using Azure AI Studio, you will need to add two additional custom keys to the connection. Follow these instructions when creating a CustomConnection.

Note: if you already have an Azure AI Language or Azure AI Translator connection, you do not need to create additional connections and may substitute them in.

Run flow

Prompt flow SDK:

Run with single line input

# Test with default input values in flow.dag.yaml:
pf flow test --flow .
# Test with specific input:
pf flow test --flow . --inputs document_path=<path_to_txt_file> language=<document_language_code>

Run with multiple lines of data

pf run create --flow . --data ./data.jsonl --column-mapping document_path='${data.document_path}' language='${data.language}' --stream

You can also skip providing column-mapping if provided data has same column name as the flow. Reference here for default behavior when column-mapping not provided in CLI.

Azure AI/ML Studio:

Run flow.

Flow Description

The flow first reads in a text file and translates it to the input language. PII information is then redacted. From the redacted text, the flow generates summaries (extractive & abstractive) and extracts named entities. Finally, the flow analyzes the sentiment of the abstractive summary.

Note: you may remove all references to Azure AI Translator (connection and tool) if you do not wish to utilize those capabilities.

This flow showcases a variety of analyses to perform on documents. Consider extending it to summarize project documents or press releases, analyze and mine the sentiment of reviews, etc.

Contact

Please reach out to Azure AI Language (taincidents@microsoft.com) with any issues.