---
title: How the NLP Analysis Works
description: "The deep linguistic analysis in Error Analyzer is not performed locally in your browser. Instead, it relies on resource-intensive AI models running on the backend. To keep the application responsive and to manage API usage efficiently, analyses are generated and…"
url: "https://erroranalyzer.com/docs/how-the-nlp-analysis-works/"
updated: "2025-09-11"
category: User Guide
---

# How the NLP Analysis Works

The deep linguistic analysis in Error Analyzer is not performed locally in your browser. Instead, it relies on resource-intensive AI models running on the backend. To keep the application responsive and to manage API usage efficiently, analyses are generated and retrieved **on a per-pair basis**.

![AI Chat at the pair level](https://erroranalyzer.com/wp-content/uploads/2025/09/image-12.png)

### When a New Analysis is Requested

- **Initial Load**
  When you first open the workspace, the system automatically loads the analysis for the first error pair in your dataset. During this process, you will see a *“Loading analysis…”* message in the AI Workspace.

![Loading NLP Analysis Message](https://erroranalyzer.com/wp-content/uploads/2025/09/image-39.png)

- **Switching Error Pairs**
  Every time you select a new pair from the Sources Panel, the system requests and displays the analysis for that specific pair. The AI Workspace will update and show that the data is being loaded.
- **Manual Refresh**
  The refresh icon in the top-right corner of the “NLP Analysis” card allows you to re-run the analysis for the currently selected pair. This is useful when a previous attempt failed or when you want to ensure the results reflect the latest AI model updates.
- **Changing the Language**
  If you switch the application’s display language, the interface and the analysis results are re-rendered. This action triggers a new analysis request for the active pair so that all generated content appears correctly in the chosen language.

### Best Practices to Manage API Requests

- **Be patient after switching pairs.** Allow the analysis to finish loading before moving to another pair. Rapid switching cancels ongoing requests and starts new ones, wasting both time and API calls.
- **Use refresh deliberately.** The refresh button is valuable for obtaining the latest analysis, but each click is a new request. Trigger it only when you have a clear reason to do so.
- **Rely on session caching.** Once an analysis has been generated for a pair, it is cached during your session. If you revisit that pair, it will reload instantly from memory without another API call. Keep in mind that refreshing the entire browser clears this cache.

### Why This Matters

By treating each analysis as a **separate, per-pair operation**, the application ensures that heavy computational tasks remain manageable. For researchers, this means smoother navigation through texts, reduced waiting times, and more intentional use of AI resources.