Subgraph Retrieval Toolkit (SRTK)

The Subgraph Retrieval Toolkit (SRTK) is a Python-based toolkit for retrieving subgraphs from large-scale knowledge graphs. SRTK provides a unified interface to access multiple knowledge graphs, including Wikidata and Freebase, and ships with state-of-the-art subgraph retrieval algorithms.

SRTK CLI Interface supports the following subcommands:

• preprocess: Preprocess a dataset for training a subgraph retrieval model.

• train: Train a subgraph retrieval model on a preprocessed dataset.

• link: Link entity mentions in a text to a knowledge graph. Currently only Wikidata is supported out of the box.

• retrieve: Retrieve a semantic-relevant subgraph from a knowledge graph with a trained retriever. It can also be used to evaluate a trained retriever.

• visualize: Visualize a retrieved subgraph using a graph visualization tool.

Contents:

• Getting Started
  • Installation
  • Setup Wikidata
  • Setup Freebase
• Command Line Interface
  • srtk preprocess
  • srtk train
  • srtk link
  • srtk retrieve
  • srtk visualize

Besides, srtk can also be used as a python library.

Python API Reference

• Python API
  • Subpackages
  • srtk.link_wikidata module
  • srtk.preprocess module
  • srtk.retrieve module
  • srtk.train module
  • srtk.visualize module

Tutorials

• Tutorials