The application is a Flask-based web server for qualitative error analysis of texts. It allows users to upload pairs of wrong and corrected texts, visualize differences, apply custom tags as annotations, and perform deep linguistic analysis with Google’s NLP and Generative AI APIs.
Core Workflows #
- Project Setup: A user creates a project, uploads a CSV of text pairs, and the backend processes the data.
diff_handler.py
calculates textual differences, andgoogle_nlp.py
enriches the texts with syntactic and semantic features, storing everything in a project-specific SQLite database. - Annotation: In the analyzer UI, users select spans of text and apply tags. This creates annotation records linking the tag, text, and location in the database.
- Analysis: The app provides per-pair AI-driven linguistic analysis and a project-wide dashboard (
/tag_report
) with aggregate charts and reports generated by LLMs.