Azure AI Language — Text Analytics Features

Quick Answer: Azure AI Language provides pre-built NLP features through a single API: sentiment analysis (with opinion mining), NER, key phrase extraction, language detection, PII detection/redaction, entity linking, and text summarization. These features work out-of-the-box with no training required.

Pre-Built Features Overview

Sentiment Analysis and Opinion Mining

Basic Sentiment Analysis

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

client = TextAnalyticsClient(
    endpoint="https://my-language.cognitiveservices.azure.com/",
    credential=AzureKeyCredential("<your-key>")
)

documents = [
    "The hotel room was spacious and clean, but the restaurant food was terrible.",
    "I absolutely loved the customer service experience!"
]

response = client.analyze_sentiment(documents, show_opinion_mining=True)

for doc in response:
    print(f"Document sentiment: {doc.sentiment}")
    print(f"  Positive: {doc.confidence_scores.positive:.2f}")
    print(f"  Negative: {doc.confidence_scores.negative:.2f}")
    print(f"  Neutral: {doc.confidence_scores.neutral:.2f}")

    for sentence in doc.sentences:
        print(f"  Sentence: '{sentence.text}'")
        print(f"    Sentiment: {sentence.sentiment}")

        # Opinion mining - aspect-level sentiment
        if sentence.mined_opinions:
            for opinion in sentence.mined_opinions:
                target = opinion.target
                print(f"    Target: {target.text} ({target.sentiment})")
                for assessment in opinion.assessments:
                    print(f"      Assessment: {assessment.text} "
                          f"({assessment.sentiment})")

Output Example

Document sentiment: mixed
  Positive: 0.45
  Negative: 0.50
  Neutral: 0.05
  Sentence: 'The hotel room was spacious and clean...'
    Sentiment: positive
    Target: hotel room (positive)
      Assessment: spacious (positive)
      Assessment: clean (positive)
  Sentence: '...but the restaurant food was terrible.'
    Sentiment: negative
    Target: restaurant food (negative)
      Assessment: terrible (negative)

On the Exam: Sentiment analysis returns four possible values: positive, negative, neutral, and mixed. "Mixed" appears when the document contains both positive and negative sentiments. Opinion mining adds target-assessment pairs to identify WHAT is being praised or criticized.

Named Entity Recognition (NER)

documents = [
    "Microsoft was founded by Bill Gates and Paul Allen in Albuquerque on April 4, 1975."
]

response = client.recognize_entities(documents)

for doc in response:
    for entity in doc.entities:
        print(f"Entity: {entity.text}")
        print(f"  Category: {entity.category}")
        print(f"  Subcategory: {entity.subcategory}")
        print(f"  Confidence: {entity.confidence_score:.2f}")
        print(f"  Offset: {entity.offset}")

NER Entity Categories

PII Detection and Redaction

documents = [
    "My SSN is 123-45-6789 and my email is john@example.com. "
    "Call me at (555) 123-4567."
]

response = client.recognize_pii_entities(documents)

for doc in response:
    print(f"Redacted text: {doc.redacted_text}")
    for entity in doc.entities:
        print(f"  PII: {entity.text} ({entity.category})")

# Output:
# Redacted text: My SSN is *********** and my email is ****************.
#   Call me at **************.
# PII: 123-45-6789 (USSocialSecurityNumber)
# PII: john@example.com (Email)
# PII: (555) 123-4567 (PhoneNumber)

PII Categories

Key Phrase Extraction

documents = [
    "Azure AI Language provides powerful natural language processing "
    "capabilities for text analytics, sentiment analysis, and entity recognition."
]

response = client.extract_key_phrases(documents)

for doc in response:
    print(f"Key phrases: {doc.key_phrases}")
# Output: ['Azure AI Language', 'natural language processing',
#          'text analytics', 'sentiment analysis', 'entity recognition']

Language Detection

documents = [
    "Bonjour le monde",
    "Hello world",
    "Hola mundo"
]

response = client.detect_language(documents)

for doc in response:
    lang = doc.primary_language
    print(f"Text: '{doc.id}' → {lang.name} ({lang.iso6391_name})"
          f" confidence: {lang.confidence_score:.2f}")
# Output:
# Text → French (fr) confidence: 1.00
# Text → English (en) confidence: 1.00
# Text → Spanish (es) confidence: 1.00

Text Summarization

Azure AI Language offers two types of summarization:

from azure.ai.textanalytics import (
    ExtractiveSummaryAction,
    AbstractiveSummaryAction
)

# Extractive summarization
poller = client.begin_analyze_actions(
    documents=["Long document text here..."],
    actions=[ExtractiveSummaryAction(max_sentence_count=3)]
)

for result in poller.result():
    for summary in result:
        for sentence in summary.sentences:
            print(f"  - {sentence.text} (rank: {sentence.rank_score:.2f})")