A paraphrase dictionary for predicates in DBpedia, which takes advantage of existing datasets and combines the technologies of machine mining and crowdsourcing. It is composed of 2284 distinct predicates in DBpedia and over 30,000 paraphrase pairs in total, in the form of <predicate, paraphrase words or phrases, score>.
Here are some examples:
academicAdvisor again work 19
academicAdvisor also study under 53
academicAdvisor as assistant 11
academicAdvisor be teach from 58
hometown bear be raise in 53
hometown bear grow up 108
hometown be left at 53
hometown be live 102
You can find all the data in "dic.txt".
"AAAI-XueB.5511.pdf" is the camera-ready version of out paper for AAAI2020.
This dataset is annotated and released by pkumod group. If you use this dataset in your research work, please cite the following paper.
// To be updated
@inproceedings{
title={The Value of Paraphrase for Knowledge Base Predicates},
author={Bingcong Xue, Sen Hu, Lei Zou, Jiashu Cheng},
booktitle={The Thirty-Fourth {AAAI} Conference on Artificial Intelligence, {AAAI}
2020, The Thirty-Second Innovative Applications of Artificial Intelligence
Conference, {IAAI} 2020, The Tenth {AAAI} Symposium on Educational To
Advances in Artificial Intelligence, {EAAI} 2020, Hilton New York Midtown, New York,
USA, February 7-12, 2020},
volume={},
pages={},
year={2020}
}