spatial-centrality 0.1.0

Spatial centrality scores for syntactic dependency trees

spatial_centrality

This package provides spatial centrality scores (D, coverage, and D') for syntactic dependency trees, as introduced in this publication:

Ferrer-i-Cancho, R., & Arias, M. (2025). Who is the root in a syntactic dependency structure? arXiv:2501.15188

The scores are designed for use with undirected syntactic dependency trees, combined with the linear order of words in a sentence. These measures help identify the root of a dependency tree using spatial and topological information.

This package is an independent implementation based on the algorithms described in the paper. All credit for the concepts and theoretical work goes to the original authors. This implementation is provided under an open-source license to facilitate use and experimentation by the NLP and computational linguistics community.

Installation

pip install .

Usage

import networkx as nx
from spatial_centrality import d_centrality, coverage_centrality, d_prime_centrality

# Create an undirected syntactic dependency tree
G = nx.Graph()
G.add_edges_from([(0, 1), (1, 2), (1, 3)])  # edges represent syntactic relations

# Define the linear order of words in the sentence
linear_order = [0, 1, 2, 3]  # e.g., token indices from left to right

# Compute scores
print(d_centrality(G, linear_order))
print(coverage_centrality(G, linear_order))
print(d_prime_centrality(G, linear_order))

License

MIT License

---

If you use this package in academic work, please cite the original paper and feel free to mention this implementation in your acknowledgments or supplementary materials.