Dataset splits for Yahoo Answers used in SIGIR 2017, AAAI 2018 and WSDM 2018 papers. Check the papers below for model comparisons on this dataset.
Original data comes from https://webscope.sandbox.yahoo.com/catalog.php?datatype=l
You will find a .pkl file containing a dictionary object. The data
is split into train, test and dev, which are by itself, dictionaries
of the format train_QA[question] = [[ans1,0],[ans2,1],[ans3],0] etc..
Please contact me at ytay017@e.ntu.edu.sg if there are any issues. If I am not supposed to be publicly releasing my dataset splits, please let me know as well.
If you use our dataset splits, please cite our paper:
@inproceedings{DBLP:conf/sigir/TayPLH17,
author = {Yi Tay and
Minh C. Phan and
Anh Tuan Luu and
Siu Cheung Hui},
title = {Learning to Rank Question Answer Pairs with Holographic Dual {LSTM}
Architecture},
booktitle = {Proceedings of the 40th International {ACM} {SIGIR} Conference on
Research and Development in Information Retrieval, Shinjuku, Tokyo,
Japan, August 7-11, 2017},
pages = {695--704},
year = {2017},
crossref = {DBLP:conf/sigir/2017},
url = {http://doi.acm.org/10.1145/3077136.3080790},
doi = {10.1145/3077136.3080790},
timestamp = {Sun, 06 Aug 2017 18:21:32 +0200},
biburl = {http://dblp.org/rec/bib/conf/sigir/TayPLH17},
bibsource = {dblp computer science bibliography, http://dblp.org}
}
@article{DBLP:journals/corr/abs-1711-07656,
author = {Yi Tay and
Luu Anh Tuan and
Siu Cheung Hui},
title = {Cross Temporal Recurrent Networks for Ranking Question Answer Pairs},
journal = {CoRR},
volume = {abs/1711.07656},
year = {2017},
url = {http://arxiv.org/abs/1711.07656},
archivePrefix = {arXiv},
eprint = {1711.07656},
timestamp = {Sun, 03 Dec 2017 12:38:15 +0100},
biburl = {http://dblp.org/rec/bib/journals/corr/abs-1711-07656},
bibsource = {dblp computer science bibliography, http://dblp.org}
}