This lesson puts the Google Natural Language API to work through the first of the course’s browser tools. It starts with one page, adds the single input that turns the tool from descriptive to diagnostic, and then scales to the whole corpus built in the previous lesson.
What this lesson covers
- The Extractor’s six steps from a pasted page to two exported files
- The intended-entity field and why it has to come first
- How to read the entity table and the page-profile verdict
- Running a full corpus within the free tier
- Recovering from the common API failures
- What raw API output does and does not tell you
The lesson opens with a result from the demo site: on a page about entity strategy, the top entity was the word “entities”. It then works through the Entity Extractor & Salience Profiler step by step. You paste one page first, so a few units of free allowance confirm the extraction is sane before hundreds of calls are made. Before pressing the button you write down what the page is meant to be about — a commitment that has to come before the output, because once you have seen the machine’s answer it is very hard not to be influenced by it.
The intended entity is what makes the verdict possible: a red mismatch means a page is about something other than what you intended, which is a finding a stakeholder can act on immediately. Scaling up, the full demo-site corpus of 637 documents ran inside the free tier at a cost of $0.00 and produced two exports — every entity on every page, and one profile row per URL — that the rest of the course reads. The lesson also covers the five errors that stop a run, the reason version 2 of the API cannot be used for prominence, and why raw output is not yet an audit.
This lesson continues Module 2, Diagnostics: Extracting the Truth About Your Site, in The AI-powered Entity Audit, and leads into page profiles and prominence scoring. It suits anyone running entity extraction for the first time, including those who have only used entity tools that hide the underlying API response.