OpenEP

This repository contains the data and code for the paper "OpenEP: Open-Ended Future Event Prediction". OpenEP aims to generate flexible and diverse predictions aligned with real-world scenarios. To facilitate the study of this task, we first construct OpenEPBench, an open-ended future event prediction dataset. Consequently, we propose StkFEP, a stakeholder-enhanced future event prediction framework that incorporates event characteristics for open-ended settings.

1. Quick Start

The code repository is based on Pytorch and Transformers. Please use the following command to install the necessary dependcies. install -r requirements.txt.

2. Data

We collect daily hot topics to generate predictive questions from seven persectives. The data samples are shown in data/question_samples.json and data/answer_samples.json.

3. Run Experiments

We use GPT-3.5 and GLM-4 as the backbone models. To run the experiments, please follow the steps outlined below.

1. Place your api keys in file: src/api_keys.py.
2. Specify data_path, response_path, and model_name in files script/run_gpt35_cn.sh or script/run_glm4_cn.sh.

4. Evaluation

We use GPT-4 as the backbone model to evaluate the model predictions. Please follow the steps below.

1. Place your api keys in file: src/api_keys.py.
2. Specify predictions_path, questions_path, ground_truth_path, and model_name in file script/run_evaluation.sh.