🐾 Wildlife Trafficking Detection: Using LLMs to Automate the Creation of Classifiers for Data Triage

This repository contains the code used in the paper:

"A Cost-Effective LLM-based Approach to Identify Wildlife Trafficking in Online Marketplaces" Proceedings of the ACM on Management of Data, Volume 3, Issue 3. Article No.: 119, Pages 1 - 23.
https://dl.acm.org/doi/10.1145/3725256 https://arxiv.org/html/2504.21211v1

---

📚 Table of Contents

1. Requirements
2. Setup
3. Use Cases & Reproducibility

---

📦 Requirements

Experiments were conducted using Python 3.11.2. All required dependencies are listed in requirements.txt and can be installed via pip.

Before running the experiments, you need to set your OpenAI key on labeling.py (line 116)

---

⚙️ Setup

1. Create a Virtual Environment (Recommended)

Use a virtual environment to avoid dependency conflicts:

python -m venv venv
source venv/bin/activate  # For Unix/macOS
venv\Scripts\activate     # For Windows

2. Install dependencies

pip install -r requirements.txt

🧪 Use Cases & Reproducibility To reproduce experiments from the paper, run main_cluster.py with the appropriate flags:

The data needed to run all experiments can be found on: https://drive.google.com/drive/folders/1UO4OYjBmvgKcFz71YeB1kefXqQhMvXGA?usp=sharing

🔧 Required Parameters:

• -sample_size: Number of samples per iteration.

• -filename: Path to the dataset CSV file. Must contain a text column named 'title'.

• -val_path: Path to the validation dataset.

• -balance: Whether to balance the dataset via undersampling (bool).

• -sampling: Sampling strategy (string: "thompson" or "random").

• -filter_label: Whether to filter out negative samples. (bool)

• -model_finetune: Model name for fine-tuning in the first iteration (string: e.g., "bert-base-uncased").

• -labeling: Source of labels (string: gpt, llama, or file).

• -model: Choose model type (string: text, multi-modal).

• -metric: Evaluation metric used to compare models between iterations (string: "f1", "accuracy", "recall", "precision").

• -baseline: Initial baseline metric score for first iteration.

• -cluster_size: Number of clusters to use.

👜 Use Case 1: Leather Products

python main_cluster.py \
  -sample_size 200 \
  -filename "data_use_cases/data_leather" \
  -val_path "data_use_cases/leather_validation.csv" \
  -balance False \
  -sampling "thompson" \
  -filter_label True \
  -model_finetune "bert-base-uncased" \
  -labeling "gpt" \
  -model "text" \
  -baseline 0.5 \
  -metric "f1" \
  -cluster_size 10

🦈 Use Case 2: Shark Products

python main_cluster.py \
  -sample_size 200 \
  -filename "data_use_cases/shark_trophy" \
  -val_path "data_use_cases/validation_sharks.csv" \
  -balance True \
  -sampling "thompson" \
  -filter_label True \
  -model_finetune "bert-base-uncased" \
  -labeling "gpt" \
  -model "text" \
  -baseline 0.5 \
  -metric "f1" \
  -cluster_size 5

Use Case 3: Animal Products

python main_cluster.py \
  -sample_size 200 \
  -filename "data_use_cases/animals" \
  -val_path "data_use_cases/validation_animals.csv" \
  -balance True \
  -sampling "thompson" \
  -filter_label True \
  -model_finetune "bert-base-uncased" \
  -labeling "gpt" \
  -model "text" \
  -baseline 0.5 \
  -metric "f1" \
  -cluster_size 10

📫 Contact For questions or feedback, please open an issue or reach out via the contact information provided in the paper.