Azure AI Sentiment Analysis Tutorial

Introduction to Sentiment Analysis

Sentiment analysis, also known as opinion mining, is the process of computationally identifying and categorizing opinions expressed in a piece of text, especially in order to determine whether the writer's attitude towards a particular topic, product, etc., is positive, negative, or neutral.

Azure Cognitive Services provides a powerful and easy-to-use Text Analytics API that includes sentiment analysis as a core feature. This tutorial will guide you through the steps to integrate this capability into your applications.

Prerequisites

An Azure subscription. If you don't have one, you can create a free account.
An Azure Text Analytics resource. You can create one via the Azure portal.
The endpoint and API key for your Text Analytics resource.
A development environment with a programming language like Python, C#, or Node.js.

Step 1: Create an Azure Text Analytics Resource

1

Navigate to the Azure Portal

Log in to your Azure account at portal.azure.com.

2

Create a Text Analytics Resource

Click on "Create a resource", search for "Text Analytics", and select it. Click "Create".

Fill in the required details such as Subscription, Resource group, Region, and Name. Choose a pricing tier (e.g., F0 for free tier testing).

3

Get Endpoint and Key

Once the resource is deployed, navigate to its overview page. You will find your "Endpoint" and one of the "Keys" listed under "Resource Management". Keep these secure as they are required for authentication.

Step 2: Implement Sentiment Analysis with Python

This example uses the Azure Text Analytics client library for Python.

Install the SDK

pip install azure-ai-textanalytics

Python Code Example

                from azure.core.credentials import AzureKeyCredential
                from azure.ai.textanalytics import TextAnalyticsClient

                endpoint = "YOUR_TEXT_ANALYTICS_ENDPOINT"
                key = "YOUR_TEXT_ANALYTICS_KEY"

                text_analytics_client = TextAnalyticsClient(endpoint=endpoint, credential=AzureKeyCredential(key))

                documents = [
                    {"id": 1, "text": "This is a wonderful library!"},
                    {"id": 2, "text": "I am not satisfied with the service."},
                    {"id": 3, "text": "The weather is neutral today."},
                    {"id": 4, "text": "I love this product, it's amazing and so easy to use."}
                ]

                try:
                    response = text_analytics_client.analyze_sentiment(documents=documents, show_opinion_mining=True)

                    for doc in response:
                        print(f"Document ID: {doc.id}")
                        print(f"Overall Sentiment: {doc.sentiment}")
                        print(f"Confidence Scores: Positive={doc.confidence_scores.positive:.2f}, Neutral={doc.confidence_scores.neutral:.2f}, Negative={doc.confidence_scores.negative:.2f}")
                        if doc.mined_opinions:
                            print("Mined Opinions:")
                            for opinion in doc.mined_opinions:
                                print(f"  - Target: {opinion.target.text}")
                                print(f"    Sentiment: {opinion.sentiment}")
                                print(f"    Assessments: {', '.join([a.text for a in opinion.assessments])}")
                        print("-" * 20)

                except Exception as err:
                    print(f"Encountered an exception: {err}")

            

Remember to replace YOUR_TEXT_ANALYTICS_ENDPOINT and YOUR_TEXT_ANALYTICS_KEY with your actual credentials.

Note: The show_opinion_mining=True parameter enables more granular analysis, extracting sentiments associated with specific targets (e.g., product features).

Understanding the Output

The sentiment analysis result for each document typically includes:

Document ID: A unique identifier for the input document.
Overall Sentiment: The general sentiment of the document (e.g., positive, neutral, negative).
Confidence Scores: Probabilities for each sentiment category, indicating the model's certainty.
Mined Opinions (Optional): If requested, this provides details about sentiments expressed towards specific entities or targets within the text.

Sentiment Scoring

The API returns a sentiment score, which can be interpreted as follows:

Positive: Indicates a positive sentiment.
Neutral: Indicates a neutral sentiment.
Negative: Indicates a negative sentiment.

Best Practices and Further Exploration

Tip: For production applications, store your Text Analytics keys securely using Azure Key Vault. Avoid hardcoding them directly in your code.

Batch Processing: The Text Analytics API supports processing multiple documents in a single request, which is efficient for large volumes of text.
Language Detection: Combine sentiment analysis with the Text Analytics language detection feature to automatically identify the language of your text before analyzing sentiment.
Error Handling: Implement robust error handling to gracefully manage potential issues with API requests or invalid input.
Fine-tuning (Advanced): For domain-specific text (e.g., medical reviews, financial reports), consider custom text classification or named entity recognition to improve accuracy.
Other Azure AI Services: Explore other services like Key Phrase Extraction, Named Entity Recognition, and Language Understanding (LUIS) to build more sophisticated AI solutions.

Conclusion

You have successfully learned how to perform sentiment analysis using Azure Cognitive Services. This powerful feature can provide valuable insights into customer feedback, social media monitoring, and more. Continue exploring the Azure AI platform to unlock the full potential of artificial intelligence.

For more detailed information, please refer to the official Azure Text Analytics documentation.