Text analysis solutions · 10–15%
Language analysis & translation
Azure Language (Foundry Tools)
Azure Language is a Foundry Tool with prebuilt APIs for analyzing text: detecting its language, finding entities, and spotting personal data — no model prompting needed.
Client: TextAnalyticsClient (azure-ai-textanalytics), key or Entra ID auth.
| Feature | Method | Notes |
|---|---|---|
| Language detection | detect_language() | Confidence 0–1; mixed-language text returns the dominant language with a lower score; unparseable text returns (unknown) with score 0. Doc ≤ 5,120 chars, ≤1,000 docs/request |
| Named Entity Recognition | recognize_entities() | Person, Location, DateTime, Organization, Address, Email, URL… |
| PII detection + redaction | recognize_pii_entities() | Categories like USSocialSecurityNumber; doc.redacted_text = masked with * |
Sentiment, summarization, key phrases exist but flagged deprecated/legacy in current content — modern route = generative prompting.
LLM-based text analysis
Study guide expects: entities/topics/summaries/structured JSON output via generative prompting (Responses API + JSON format instruction or structured outputs), sentiment/tone/safety detection with prompts, domain customization (compliance summarization) via prompt engineering or fine-tuning.
Azure Translator
Azure Translator is a neural machine-translation service in Foundry Tools.
- 90+ languages; text translation, transliteration, document translation (keeps structure), custom translation models for domain terms; default model or LLM-powered.
- Client:
TextTranslationClient(endpoint or region). translate(body=[InputTextItem...], to_language=["fr","en"])— omitfrom_languageand the service auto-detects the source language.transliterate(..., from_script="Jpan", to_script="Latn")— script change, same language (こんにちは becomes Kon’nichiwa).get_supported_languages()lists them.