https://github.com/aksw/llmdatasetgenerator
LLM based datatset generator for KGQA on user defined knowledge graphs
Science Score: 13.0%
This score indicates how likely this project is to be science-related based on various indicators:
-
○CITATION.cff file
-
✓codemeta.json file
Found codemeta.json file -
○.zenodo.json file
-
○DOI references
-
○Academic publication links
-
○Academic email domains
-
○Institutional organization owner
-
○JOSS paper metadata
-
○Scientific vocabulary similarity
Low similarity (8.0%) to scientific vocabulary
Keywords
Repository
LLM based datatset generator for KGQA on user defined knowledge graphs
Statistics
- Stars: 0
- Watchers: 26
- Forks: 0
- Open Issues: 0
- Releases: 2
Topics
Metadata Files
README.md
Queryfy - Generating datasets for LLM finetuning from Knowledge Graphs
This tool takes a knowledge graph in ttl format as input and generates KGQA datasets from that. It does so by using multiple LLMs (see below) to generate appropriate questions for that specific knowledge graph, including the expected answers for reference and corresponding SPARQL queries.
This work was done for research purposes because by the time of this writing, there was no way to automatically generate datasets for training/finetuning from arbitrary knowledge graphs.
We hope to open up new areas of research by providing this prototype and are looking forward to contributions.
Execution
Requirements: * python * transformers * capable GPU
- Edit the
config.yamlto your liking - run
python pipeline.py
The script generates folders for each step with the results.
Owner
- Name: AKSW Research Group @ University of Leipzig
- Login: AKSW
- Kind: organization
- Location: Leipzig
- Website: http://aksw.org
- Repositories: 358
- Profile: https://github.com/AKSW
GitHub Events
Total
- Release event: 2
- Watch event: 1
- Push event: 10
- Fork event: 1
- Create event: 2
Last Year
- Release event: 2
- Watch event: 1
- Push event: 10
- Fork event: 1
- Create event: 2