Connecting Language and Knowledge with Heterogeneous Representations for Neural Relation Extraction
Paper Published in NAACL 2019: NFETC
- tensorflow >= r1.2
- hyperopt
- gensim
- sklearn
To download the dataset used:
cd ./data
python prepare_data.py
Construct the knowledge graph:
python create_kg.py
Preprocessing the data:
python preprocess.py -p -g
Copy the fils in ./kg to the data folder in tensorflow-efe and run the following commands to obtain the complex embeddings:
python preprocess.py --data fb3m
python train.py --model best_Complex_tanh_fb3m --data fb3m --save
python get_embeddings.py --embed complex --model best_Complex_tanh_fb3m --output <repo_path>/kg
Then copy e2id.txt and r2id.txt in the tensorflow-efe/data/fb3m to ./kg and run the following command:
python get_embeddings.py
python task.py --model <model_name> --eval <max_number_of_search> --runs <number_of_runs_per_setting>
model_name can be found in model_param_space.py. You can also define the search space by yourself.
python eval.py --model <model_name> --prefix <file_prefix> --runs <number_of_runs> [--eval]
model_name can be found in model_param_space.py.
Option --eval is used to save the data for plotting; otherwise, it'll run multiple times and calculate the means and stds of P@N.
If you found this codebase or our work useful, please cite:
@InProceddings{xu2019connecting,
author = {Xu, Peng and Barbosa, Denilson},
title = {Connecting Language and Knowledge with Heterogeneous Representations for Neural Relation Extraction}
booktitle = {The 17th Annual Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies (NAACL 2019)},
month = {June},
year = {2019},
publisher = {ACL}
}