Picture a product launch. Within a few days there are 8,000 reviews, 2,000 support tickets, and hundreds of social posts. Someone reads fifty of them, decides customers seem happy, and moves on. Buried in the rest is the same battery complaint, over and over.
AI can read all of them before lunch. The fourth video in my AI-901 Microsoft Azure AI Fundamentals series shows how, using four questions you can ask of any text: what is it about, who and what does it mention, how does the writer feel, and what’s the short version? A fifth question covers what to do when the content isn’t text yet.
How AI Turns Words Into Data
Computers don’t read words, they work with numbers. The video starts with the natural language processing (NLP) steps that prepare text for analysis, including tokenization, normalization, stop-word removal, stemming and lemmatization, and part-of-speech tagging. From there it compares the two main approaches: statistical methods like TF-IDF and bag-of-words, and semantic methods that turn tokens into embeddings, the same idea behind modern language models.
What Is It About, and Who Does It Mention?
Key phrase extraction answers the first question by pulling out the main concepts in a piece of text, and the video explains why a simple word count isn’t enough. Entity detection, or named entity recognition (NER), answers the second by finding specific people, places, dates, and products and labeling each one with a type. The video also covers PII detection and redaction, entity linking, and custom NER, plus the subtle difference between a keyword and an entity that exam questions like to test.
How Does the Writer Feel?
Sentiment analysis labels text as positive, neutral, or negative, with confidence scores, at both the document and sentence level. Opinion mining goes a step further and links each opinion to the thing it’s about. That’s how a four-star review that says the battery is terrible still shows up as a battery problem.
What’s the Short Version?
Summarization comes in two flavors, and the difference matters. Extractive summarization returns the most important sentences word for word, while abstractive summarization writes new sentences that capture the meaning. The video explains when each one fits, and why extractive is the safer choice when exact wording matters.
Two Ways to Analyze Text in Microsoft Foundry
Every technique in the video can be run with a general-purpose model and a prompt, or with Azure Language in Foundry Tools. The video compares the two, explains which Azure Language features are core and which are now legacy, and gives a simple rule of thumb for choosing between them.
When It Isn’t Text Yet
Not every source starts as text. The video shows how OCR and field extraction pull data from scanned documents, how audio and video are transcribed and analyzed, and how Azure Content Understanding brings it all together in one service. Sarah, the series’ AI-generated learner, connects it back to her scanned invoices from the last video.
Watch the Video
If you’re preparing for the AI-901 exam or building text analysis into an app, these four questions give you a simple way to understand what AI can pull out of text. Watch the video now, and you’ll never look at a pile of customer feedback the same way.
And if you find it helpful, subscribe to the channel and turn on notifications so you don’t miss the next video in the AI-901 series, where the focus turns to how AI sees and hears.
Links:
A Beginner’s Guide to the AZ-900
Zero to Hero with Azure Virtual Desktop
Hybrid Identity with Windows AD and Azure AD
Windows 365 Enterprise and Intune Management
Exam AI-901: Microsoft Azure AI Fundamentals
Study guide for Exam AI-901: Microsoft Azure AI Fundamentals
What is Azure Language in Foundry Tools?
What is sentiment analysis and opinion mining?
What is summarization?
What is Azure Content Understanding in Foundry Tools?
