Syntactic analysis of natural languages

Language-specific syntactic models

I’m using DSPy to develop Python packages that leverage language models to analyze the syntax of historical languages:

• arsgrammatica: a Python package leveraging LLMs with dspy to analyze the syntax of passages of Latin.
• grammatike: a Python package leveraging LLMs with dspy to analyze the syntax of passages of ancient Greek.
• diqduq: a Python package leveraging LLMs with dspy to analyze the syntax of passages of Biblical Hebrew.

Single-page web apps for exploring syntactic analyses

• syntaxexplorer: visualize syntactic analyses encoding tokens and verbal units. Currently tested against arsgrammatica.

Universal syntax models

A universal model of natural language syntax:

• aat: an Agent-Action-Target model of natural-language syntax in Python.

A package for extracting Universal Dependency (UD) information from nlp analysis with spaCy:

• udsyntax: get Universal Dependency (syntactic) data from spaCy into a simple syntax graph format.