CMF

This is the source code for paper Understanding Event Predictions via Contextualized Multilevel Feature Learning appeared in CIKM21

Data

We processed some country based datasets from the ICEWS data. Please find example datasets in this Google Drive Link. A brief introduction of the data file is as follows:

loc2id.txt location to id mapping (several locations are considered for each country).
loc_entity2id.txt entity to id mapping.
data_count.pkl event count data
data_label.pkl location, time, etc.
data_graph.bin DGL graphs.
loc_text_emb.pkl document embeddings.

Prerequisites

The code has been successfully tested in the following environment. (For older dgl versions, you may need to modify the code)

Python 3.7.9
PyTorch 1.7.0+cu92
dgl 0.5.2
Sklearn 0.23.2

Getting Started

Prepare your code

Clone this repo.

git clone https://github.com/amy-deng/CMF
cd CMF

Prepare your data

Download the dataset (e.g., EG) from the given link and store them in data filder. Or prepare your own dataset in a similar format. The folder structure is as follows:

- CMF
	- data
		- EG
		- your own dataset
	- src

Training and testing

Please run following commands for training and testing under the src folder. We take the dataset EG as the example.

Evaluate the event prediction model

python train_pred.py -sl 7 -s -m cmf -ho 1 --gpu 1 -d EG -hd 32 -nl 2 -td 64 --eid 13 --lr 0.003 -w -l 5

Evaluate the event prediction and explanation model

python train_pred_exp.py -sl 7 -s -m cmf -ho 1 --gpu 1 -d EG -hd 32 -nl 2 -td 64 --eid 13 --lr 0.003 -w -l 1

Cite

Please cite our paper if you find this code useful for your research:

@inbook{10.1145/3459637.3482309,
author = {Deng, Songgaojun and Rangwala, Huzefa and Ning, Yue},
title = {Understanding Event Predictions via Contextualized Multilevel Feature Learning},
year = {2021},
isbn = {9781450384469},
publisher = {Association for Computing Machinery},
address = {New York, NY, USA},
url = {https://doi.org/10.1145/3459637.3482309},
booktitle = {Proceedings of the 30th ACM International Conference on Information &amp; Knowledge Management},
pages = {342–351},
numpages = {10}
}