Getting Started with ArchiTXT UI

Getting Started with ArchiTXT UI#

ArchiTXT provides a web-based user interface built with Streamlit, allowing for interactive exploration of your corpora, schema inference, and data simplification. This UI is particularly useful for visualizing the transformation process and fine-tuning parameters without writing code.

Launching the Application#

The ArchiTXT UI is accessible via a dedicated Command Line Interface (CLI). To start the server, run:

$ architxt ui

By default, the application will be available at http://localhost:8501.

Core Interface Overview#

The interface is divided into several functional areas designed to guide you through the data lifecycle:

  1. Global Metrics: Located at the top of the page, this section provides real-time counts of your current Forest (Total Trees, Entities, Groups, and Relations).

  2. Navigation: Use the top navigation bar to switch between File operations (Import/Export) and Tools (Visualize/Transform).

Data Lifecycle Workflows#

The UI works on one forest at a time, allowing you to iteratively refine your data. The metrics at the top will update dynamically as you import new data and apply transformations. The clear data button allows you to reset the current session and start fresh with a new dataset. The typical workflow involves:

Importing Data#

Navigate to File > Import to load data into the ArchiTXT environment. The UI supports five primary sources:

  • Text Corpus: Process raw text files using NLP parsers (CoreNLP or Benepar). You can configure entity resolution (UMLS, MeSH, etc.) and filter specific relations.

  • SQL/Graph Databases: Connect to relational databases (via SQLAlchemy URI) or Neo4j graph instances.

  • Document Files: Support for structured formats including .json, .toml, .csv, and .xlsx.

  • JSONL File: Direct import of ArchiTXT native JSON Lines format.

Visualization#

Navigate to Tools > Visualize to inspect your data:

  • Schema View: A dynamic graph visualization showing the relationships between entities and groups.

  • Instance Explorer: Select specific groups to view their underlying data in a searchable tabular format.

Transformation & Simplification#

Navigate to Tools > Simplify to refine your data structures. Two methods are available:

  • Rule-Based: Uses similarity (Tau) and support thresholds to merge or rewrite tree structures.

  • LLM-Based: Leverages Large Language Models (via local HuggingFace models or APIs like OpenAI) to perform intelligent rewriting and refinement.

Exporting Results#

Once processed, navigate to File > Export to save your session. You can export back to:

  • SQL Databases

  • Graph Databases (Neo4j)

  • JSONL files (available for direct browser download)