architxt 0.7.0

ArchiTXT is a tool for structuring textual data into a valid database model. It is guided by a meta-grammar and uses an iterative process of tree rewriting.

ArchiTXT: Text-to-Database Structuring Tool

ArchiTXT is a Python library and CLI tool that automatically converts unstructured text corpora into structured, database-ready data. It infers database schemas directly from text and generates corresponding structured instances using a meta-grammar and iterative tree-rewriting process.

ArchiTXT is designed for researchers, data engineers, and NLP practitioners who need a transparent and auditable process to transform raw textual data into storable, queryable and machine-learning-ready datasets.

Why ArchiTXT?

Working with unstructured text becomes complex when you need:

• Structured storage
• Queryable entities and relations
• Reproducible data modeling

ArchiTXT bridges this gap by:

• Discovering latent structural patterns in annotated corpora
• Automatically generating database schemas
• Producing structured instances aligned with the inferred schema
• Ensuring transparency through rule-based rewriting

Installation

To install ArchiTXT, make sure you have Python 3.10+ and pip installed. Then, run:

pip install architxt

For the development version, you can install it directly through GIT using

pip install git+https://github.com/Neplex/ArchiTXT.git

Usage

ArchiTXT is built to work seamlessly with BRAT-annotated corpora that includes pre-labeled named entities. It can parse the texts using either CoreNLP or SpaCy, depending on your preference and setup. See the documentation for more information.

For CoreNLP, it requires access to a CoreNLP server, which you can set up using the Docker Compose configuration available in the source repository. To deploy it, you can use the following command:

docker compose up -d corenlp

After parsing the annotated texts into ArchiTXT's internal representation, you can infer a database schema and instance based on the annotated entities and generate structured instances accordingly. See the documentation for more information.

The result can be exported as a relational or property graph database. See the documentation for more information.

ArchiTXT is available as a Python library but also provides a command-line interface (CLI) for users who prefer working in the terminal. You can run the CLI using:

architxt --help

Sponsors

This work has received support under the JUNON Program, with financial support from Région Centre-Val de Loire (France).