Azure Language Overview

Natural language processing features for text analysis, entity recognition, sentiment analysis, and conversational AI

Azure Language

Azure Language is a cloud-based service providing Natural Language Processing (NLP) features for understanding and analyzing text. Build intelligent applications with pre-configured and customizable language models for text analytics, entity extraction, sentiment analysis, and conversational AI.

Key Capabilities

Text Analytics

Extract insights from unstructured text

Entity Recognition

Identify and categorize entities in text

Sentiment Analysis

Determine sentiment and opinions

Summarization

Generate summaries of documents and conversations

Q&A

Build question answering systems

Custom Models

Train models for your specific domain

Pre-configured Features

Ready-to-use NLP capabilities that require no training:

Named Entity Recognition (NER)

Identify and categorize entities in text:

• Person: Names of people
• Location: Cities, countries, landmarks
• Organization: Companies, agencies, institutions
• DateTime: Dates, times, durations
• Quantity: Numbers, measurements, percentages
• Email, URL, Phone Number: Contact information

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

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

documents = [
    "Microsoft was founded by Bill Gates and Paul Allen in Redmond, Washington."
]

result = client.recognize_entities(documents)
for doc in result:
    for entity in doc.entities:
        print(f"{entity.text} ({entity.category})")

Personally Identifiable Information (PII) Detection

Detect and redact sensitive information:

• Personal: Names, addresses, phone numbers
• Financial: Credit card numbers, bank accounts
• Medical: Health information (PHI)
• Government: SSN, passport numbers, tax IDs

result = client.recognize_pii_entities(
    documents=["My SSN is 123-45-6789 and my email is [e-mail]"],
    language="en"
)

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

Sentiment Analysis and Opinion Mining

Analyze sentiment at document and sentence level:

• Overall sentiment: Positive, negative, neutral, mixed
• Confidence scores: For each sentiment category
• Opinion mining: Link opinions to specific aspects

result = client.analyze_sentiment(
    documents=["The food was excellent but the service was slow."],
    show_opinion_mining=True
)

for doc in result:
    print(f"Document sentiment: {doc.sentiment}")
    print(f"Positive: {doc.confidence_scores.positive}")
    print(f"Negative: {doc.confidence_scores.negative}")
    
    for sentence in doc.sentences:
        print(f"  Sentence: {sentence.text}")
        print(f"  Sentiment: {sentence.sentiment}")
        
        for opinion in sentence.mined_opinions:
            print(f"    Target: {opinion.target.text} ({opinion.target.sentiment})")

Key Phrase Extraction

Extract main concepts from text:

• Identify key topics and themes
• Extract important phrases
• Summarize main points

result = client.extract_key_phrases(
    documents=["Azure Language provides powerful NLP capabilities for analyzing text."]
)

for doc in result:
    print("Key phrases:")
    for phrase in doc.key_phrases:
        print(f"  - {phrase}")

Language Detection

Detect language of text:

• Identify from 120+ languages
• Return language code and name
• Confidence score for detection

result = client.detect_language(
    documents=["Bonjour, comment allez-vous?"]
)

for doc in result:
    print(f"Language: {doc.primary_language.name}")
    print(f"ISO code: {doc.primary_language.iso6391_name}")
    print(f"Confidence: {doc.primary_language.confidence_score}")

Summarization

Generate summaries of text and conversations:

Extractive Summarization

Extract key sentences from document:

from azure.ai.textanalytics import ExtractiveSummaryAction

poller = client.begin_analyze_actions(
    documents=[long_document],
    actions=[ExtractiveSummaryAction(max_sentence_count=3)]
)

for result in poller.result():
    for summary_result in result:
        for sentence in summary_result.sentences:
            print(sentence.text)

Abstractive Summarization

Generate new summary text:

from azure.ai.textanalytics import AbstractiveSummaryAction

poller = client.begin_analyze_actions(
    documents=[long_document],
    actions=[AbstractiveSummaryAction()]
)

for result in poller.result():
    for summary_result in result:
        for summary in summary_result.summaries:
            print(summary.text)

Conversation Summarization

Summarize meetings and calls:

• Chapter summaries for long meetings
• Issue and resolution extraction (call centers)
• Follow-up actions and items

Entity Linking

Disambiguate entities and link to Wikipedia:

result = client.recognize_linked_entities(
    documents=["Microsoft was founded in Albuquerque."]
)

for doc in result:
    for entity in doc.entities:
        print(f"Name: {entity.name}")
        print(f"URL: {entity.url}")
        print(f"Data source: {entity.data_source}")

Text Analytics for Health

Extract medical information from clinical text:

• Conditions: Diagnoses, symptoms
• Medications: Drugs, dosages
• Procedures: Treatments, examinations
• Anatomy: Body structures
• Relations: Connections between entities

from azure.ai.textanalytics import HealthcareEntityRelation

poller = client.begin_analyze_healthcare_entities(
    documents=["Patient prescribed 50mg of ibuprofen for headache."]
)

for result in poller.result():
    for doc in result:
        for entity in doc.entities:
            print(f"{entity.text} ({entity.category})")
        
        for relation in doc.entity_relations:
            print(f"Relation: {relation.relation_type}")

Custom Features

Train models specific to your domain:

Custom Named Entity Recognition

Train models to extract domain-specific entities:

• Define custom entity categories
• Label training data
• Train and deploy models
• 50+ labeled documents recommended

from azure.ai.textanalytics import RecognizeCustomEntitiesAction

poller = client.begin_analyze_actions(
    documents=["Order #12345 shipped to warehouse A."],
    actions=[
        RecognizeCustomEntitiesAction(
            project_name="my-project",
            deployment_name="production"
        )
    ]
)

for result in poller.result():
    for entity_result in result:
        for entity in entity_result.entities:
            print(f"{entity.text} ({entity.category})")

Custom Text Classification

Classify documents into custom categories:

• Single-label: Each document gets one category
• Multi-label: Documents can have multiple categories
• Define your own classification schema
• 50+ labeled documents per class

from azure.ai.textanalytics import SingleLabelClassifyAction

poller = client.begin_analyze_actions(
    documents=["Customer requesting refund for damaged product."],
    actions=[
        SingleLabelClassifyAction(
            project_name="ticket-classifier",
            deployment_name="production"
        )
    ]
)

for result in poller.result():
    for classification_result in result:
        for classification in classification_result.classifications:
            print(f"Category: {classification.category}")
            print(f"Confidence: {classification.confidence_score}")

Conversational Language Understanding (CLU)

Build conversational AI models:

• Intents: What the user wants to do
• Entities: Key information to extract
• Utterances: Example phrases
• Train with Language Studio

from azure.ai.language.conversations import ConversationAnalysisClient

clu_client = ConversationAnalysisClient(endpoint, credential)

result = clu_client.analyze_conversation(
    task={
        "kind": "Conversation",
        "analysisInput": {
            "conversationItem": {
                "text": "Book a flight to Seattle for next Friday",
                "id": "1",
                "participantId": "user"
            }
        },
        "parameters": {
            "projectName": "travel-assistant",
            "deploymentName": "production"
        }
    }
)

print(f"Intent: {result['result']['prediction']['topIntent']}")
for entity in result['result']['prediction']['entities']:
    print(f"Entity: {entity['category']} = {entity['text']}")

Question Answering

Build custom Q&A systems:

• Import from documents, URLs, FAQs
• Add question-answer pairs manually
• Train conversational models
• Multi-turn conversations
• Chitchat personalities

from azure.ai.language.questionanswering import QuestionAnsweringClient

qa_client = QuestionAnsweringClient(endpoint, credential)

response = qa_client.get_answers(
    question="What is Azure Language?",
    project_name="product-faq",
    deployment_name="production"
)

for answer in response.answers:
    print(f"Answer: {answer.answer}")
    print(f"Confidence: {answer.confidence}")

Orchestration Workflow

Connect multiple language models:

• Route to CLU, Q&A, or LUIS
• Combine different project types
• Single endpoint for multiple models

Language Support

Azure Language supports 100+ languages for various features:

• NER: 20+ languages
• Sentiment Analysis: 90+ languages
• Key Phrase Extraction: 120+ languages
• Language Detection: 120+ languages

Use Cases

Customer Support

• Analyze customer feedback sentiment
• Extract key issues from support tickets
• Build Q&A chatbots
• Summarize support conversations
• Detect PII in customer communications

Content Analysis

• Categorize articles and documents
• Extract key topics and themes
• Summarize long documents
• Detect language of content
• Link entities to knowledge bases

Healthcare

• Extract medical entities from notes
• Analyze clinical documentation
• Identify medications and conditions
• Support clinical decision systems
• Ensure PHI compliance

Business Intelligence

• Analyze survey responses
• Monitor brand sentiment
• Extract insights from feedback
• Categorize business documents
• Track customer satisfaction

SDK Support

Python

pip install azure-ai-textanalytics

C#

dotnet add package Azure.AI.TextAnalytics

Java

<dependency>
  <groupId>com.azure</groupId>
  <artifactId>azure-ai-textanalytics</artifactId>
</dependency>

JavaScript

npm install @azure/ai-text-analytics

Getting Started

1. Create Resource

   Create an Azure Language resource in the Azure Portal

2. Try Language Studio

   Test features with sample data at language.cognitive.azure.com

3. Install SDK

   Install the Text Analytics SDK for your language

4. Analyze Text

   Start extracting insights from your text data

Pricing

• Free Tier (F0): 5,000 text records per month
• Standard Tier (S): Pay per 1,000 text records
• Custom models: Additional training and hosting costs
• Different pricing for different features

Next Steps

• Try Language Studio
• View Quickstart Guide
• Explore Custom Models
• Learn about Question Answering