Error Analyzer is a free, open-source web application designed for qualitative error analysis. It helps researchers, educators, and students analyze texts more systematically by providing tools to compare text versions, annotate errors with custom tags, and gain deeper insights using AI. Because it runs locally on your computer, it ensures complete data privacy and user control. 

The application is designed for:

• Linguists and Researchers studying language acquisition, error typologies, or revision practices.
• Educators and Teachers who want to diagnose learner difficulties and identify recurring error patterns in student writing.
• Students in linguistics or applied language studies who need a tool for hands-on practice in textual analysis and annotation. 

The key advantages include:

• Streamlined Workflow: Organizes your data, annotations, and reports into self-contained projects.
• Efficient Tagging: Allows for precise and consistent annotation using custom, hierarchical tagging schemas.
• AI-Powered Insights: Integrates AI to provide automated suggestions, summaries, and linguistic analysis, reducing repetitive tasks.
• Data-Driven Reporting: Generates statistical summaries and visualizations to help you identify trends and patterns in your data. 

To run Error Analyzer, you’ll need:

• Operating System: Windows 10 or newer, macOS 10.14 (Mojave) or newer, or a mainstream Linux distribution.
• Web Browser: A modern browser like Google Chrome, Mozilla Firefox, or Microsoft Edge.
• Python (for manual installation): If not using the Windows installer, you must have Python 3.8 or newer installed. 

There are two methods:

1. Windows Installer (Recommended): Download and run the ErrorAnalysisTaggingApp-Setup.exe file. It includes everything needed and creates shortcuts for easy access.
2. Manual Installation (macOS, Linux, Advanced Windows): Download the project’s source code, create a Python virtual environment, and install the required libraries using the command pip install -r requirements.txt. 

Yes. To activate the AI-powered features like automated NLP analysis and the AI Chat, you must perform a one-time setup. The application will guide you to its Settings page, where you’ll need to enter a Google API Key. You can obtain this free key from the Google Cloud Console by creating a project and enabling the “Generative Language API”. 

1. Create a Project: Give your project a name, a description, and specify the language of the texts.
2. Prepare Your Data: Create a CSV file with two columns: ErrorText for the original or erroneous version and CorrectedText for the revised version.
3. Upload and Process: Upload your CSV file into the project. The application will then run an initial linguistic analysis to prepare your texts, creating a “linguistic blueprint” for each pair. Once complete, you’ll be taken to the main analysis screen. 

A project is a self-contained research workspace for a specific study or dataset. It keeps all your components—texts, the tagging schema, annotations, and reports—organized in one place. This ensures your analysis remains structured, contextually grounded, and reproducible. 

Annotation is a two-step process:

1. Build Your Codebook: In the “Noteboard” panel, you create a library of custom tags. Each tag represents an analytical category (e.g., Subject-Verb Agreement Error) and can have a description, color, and be organized into hierarchies.
2. Apply Tags: In the main text comparison view, you highlight a segment of text in either the original or corrected version and apply the appropriate tag from your library. The annotated text will be underlined with the tag’s color. 

The Tag Insights page serves as your main reporting dashboard. Here you can review a statistical summary of your tags, see visualizations of linguistic patterns, and synthesize your notes. You can also export an entire project as a .zip archive for backup or collaboration, or export your annotations to various formats like CSV, Excel, XML, or HTML for use in other software. 

Also called the “Linguistic Blueprint,” this is a deep grammatical and semantic analysis that runs automatically on each text pair. It provides detailed information through several tabs, including:

• Part-of-Speech Tagging: Identifies the grammatical role of every word.
• Dependency Parsing: Maps the grammatical relationships between words in a sentence.
• Named Entity Recognition: Identifies proper nouns like people, places, and organizations.
• Delta View: Quantifies the linguistic changes between the ErrorText and CorrectedText.

This analysis provides a structured, computational foundation to support your manual annotations. 

The AI Chat is an interactive research assistant that adapts to your workflow.

• Pair-Level (Microscope): When viewing a single text pair, you can ask detailed linguistic questions about it.
• Project-Level (Telescope): In the Tag Report view, you can ask “big picture” questions to find trends across your entire dataset (e.g., “Which error tags are most frequent?”). It can also perform web searches to help contextualize your findings. 

The Auto-Tag tool, available within the AI Chat, automates the annotation process for the currently viewed text pair. You activate it and give a clear command (e.g., “Find every word with incorrect capitalization in the Error Text and apply the ‘Capitalization Error’ tag”). The AI will first present a plan for your approval. Once you confirm, it applies the annotations, which you can then review and modify. 

The documentation notes that users may occasionally encounter issues such as installation errors, performance slowdowns when working with very large corpora, annotation glitches, or the AI analysis not responding. The “Troubleshooting” section of the official documentation provides guidance on identifying and fixing these problems. 

The “Reference” section of the documentation is for developers, contributors, and advanced users. It provides in-depth technical details on the application’s architecture, technology stack (Flask, SQLite), database schema, and a full API reference for all available endpoints.