kg-parser 0.1.1

A Python package to parse text into a knowledge graph using LLMs.

Knowledge Graph Parser

Overview

kg-parser is a Python package and CLI application that extracts structured knowledge graph triples from unstructured text using large language models (LLMs). The package supports multiple backends (HuggingFace, OpenAI, and local Jan) and offers flexible output formats—either as arrays of strings or as dictionaries.

Features

• Multi-backend Support: Use HuggingFace, OpenAI, or a local Jan server.
• Batch Processing: Process multiple texts efficiently.
• Flexible Output: Choose between list or dict representations for triples.
• CLI & API: Easily run as a command-line tool or integrate into your Python projects.

Installation

Using Conda

Create and activate the conda environment with the provided configuration:

conda env create -f environment.yml
conda activate kg-parser

Using Pip

Install the package directly from PyPI:

pip install kg-parser

Install the minimal dependencies from requirements.txt:

pip install -r requirements.txt

Alternatively, install the package in editable mode:

pip install -e .

Usage

As a CLI Application

Run the CLI tool from the command line:

python -m kg_parser.cli \
  --model-type huggingface \
  --model-name-or-path "google/flan-t5-small" \
  --input-file test_input.json \
  --output-file output_kg.json \
  --triple-format list

Arguments:

• --model-type: Choose from huggingface, openai, or jan_local.
• --model-name-or-path: Specify the model name or path.
• --input-file: Path to a JSON file containing an array of text strings.
• --output-file: Path where the output JSON will be saved.
• --triple-format: Output format for triples (list for arrays or dict for dictionaries).

As a Python Package

Import and use kg-parser in your own Python scripts:

from kg_parser.config import ModelConfig, ModelType
from kg_parser.core import KGParser

# Configure the model
model_config = ModelConfig(
    model_type=ModelType.HUGGINGFACE,
    model_name_or_path="google/flan-t5-small"
)
parser = KGParser(model_config)

# Process texts
texts = [
    "Mount Everest is the highest mountain in the world. It's located in Nepal."
]
results = parser.parse_batch(texts)

# Save results with triples as lists
parser.save_to_json(results, "output_kg.json", triple_format="list")

License

This project is licensed under the MIT License. See the LICENSE file for details.

Acknowledgements

• The knowledge graph extraction prompt is adapted from the paper GraphEval: A Knowledge-Graph Based LLM Hallucination Evaluation Framework. The original unmodified prompt can be found in Appendix A ("A. KG Construction Prompt").