WillJeynes
38 lines
1.6 KiB
Markdown
38 lines
1.6 KiB
Markdown
# AI models for identifying trigger events in disinformation analysis
|
|
Final Dissertation Submission Repository - Experiments with created dataset
|
|
|
|
This is a sister repository, for project background see [this link](https://git.host.jeynes.uk/jill/LLMsForDisinformationAnalysis)
|
|
|
|
|
|
# This Repository:
|
|
## Finetuned Model
|
|
Tinetuning a LLM to better predict possible disinformation claims arising from world event
|
|
|
|
|
|
|
|
Stats available [here](/finemodel/)
|
|
|
|
Final LoRa version available [on HuggingFace](https://huggingface.co/WillJeynes/LLMsForDisinformationPrediction)
|
|
|
|
## Graph Viz
|
|
A way to visualise the connections between claims and trigger events
|
|
|
|
|
|
|
|
[View Online Demo](https://jillweynes.github.io/LLMsForDisinformationPrediction-GraphVizBuilt/)
|
|
|
|
## Repository Structure
|
|
```
|
|
├── query_model.py # call final finetuned LLM from hugging face
|
|
├── finemodel/
|
|
| ├── eval*.py # Call APIs
|
|
| ├── lora*.py, full.py # Train models against dataset
|
|
| └── q_*.py # Expose trained models as API
|
|
├── graphviz/
|
|
| ├── frontend/ # React + Parcel + react-force-graph frontend to visualise results
|
|
| └── processing/ # Python scripts to generate clusters and titles
|
|
└── data/ # Holder from project data
|
|
├── dataset.jsonl # Collated dataset - in full format
|
|
└── dataset.csv # Collated dataset - in CSV format
|
|
```
|