Azure Text Analytics: Customer Feedback Analyzer (Sentiment + Key Phrases)

This sample demonstrates how to use Azure Text Analytics to quickly understand customer satisfaction trends without manual effort. It:

• Detects sentiment (Positive, Neutral, Negative) for each feedback item.
• Extracts key phrases (e.g., product names, issues).
• Prints a concise console summary and writes an analysis-result.json report you can share in the repo.

What it looks like

Example console output:

Customer Feedback Analysis
--------------------------------
[0] Love the new noise-cancelling headphones, but the battery dies after 3 hours.
    Sentiment: Mixed (Pos 0.64, Neu 0.06, Neg 0.30)
    Key Phrases: noise-cancelling headphones, battery

[1] Customer support resolved my issue quickly. Very satisfied!
    Sentiment: Positive (Pos 0.98, Neu 0.01, Neg 0.01)
    Key Phrases: Customer support, issue

And a shareable JSON file: analysis-result.json.

Prerequisites

• Azure subscription with an Azure AI Language (Text Analytics) resource
  • Note the resource Endpoint and Key from the Azure portal.
• .NET 9 SDK

Setup

1. Add the Azure Text Analytics SDK

dotnet add package Azure.AI.TextAnalytics

2. Configure credentials (do NOT hardcode secrets)

Set environment variables:

• AZURE_TEXT_ANALYTICS_ENDPOINT — e.g., https://<your-resource>.cognitiveservices.azure.com/
• AZURE_TEXT_ANALYTICS_KEY — your resource key

Examples:

• Windows (PowerShell):

[Environment]::SetEnvironmentVariable("AZURE_TEXT_ANALYTICS_ENDPOINT","https://<your-resource>.cognitiveservices.azure.com/","User")
[Environment]::SetEnvironmentVariable("AZURE_TEXT_ANALYTICS_KEY","<your-key>","User")

• macOS/Linux (bash):

export AZURE_TEXT_ANALYTICS_ENDPOINT="https://<your-resource>.cognitiveservices.azure.com/"
export AZURE_TEXT_ANALYTICS_KEY="<your-key>"

In Visual Studio you can set per-profile environment variables via: Project > Properties > Debug > Environment variables.

3. Build and run

dotnet build
dotnet run

You’ll see a per-feedback sentiment summary in the console and a generated analysis-result.json file.

Customize

• Replace the feedback array in Program.cs with your real data source (CSV, database, etc.).
• Change language if your feedback isn’t English.
• For large volumes or multiple actions in one pass, consider long-running operations (LRO) with StartAnalyzeActions.

Troubleshooting

• 401/403: check AZURE_TEXT_ANALYTICS_KEY and that you’re using the matching endpoint.
• 404: verify the endpoint URL (region and resource name).
• 429: you are being throttled; back off and retry.

Repository contents

• Program.cs — main sample code.
• analysis-result.json — generated report.
• README.md — this file.