Distant supervision, heuristically labeling a corpus using a knowledge base, has emerged as a popular choice for training relation extractors. In this paper, we show that a significant number of '' negative '' examples generated by the labeling process are false negatives because the knowledge base is incomplete. Therefore the heuristic for generating negative examples has a serious flaw. Building on a state-of-the-art distantly-supervised extraction algorithm, we proposed an algorithm that learns from only positive and unlabeled labels at the pair-of-entity level. Experimental results demonstrate its advantage over existing algorithms.
%0 Journal Article
%1 min_distant_2013
%A Min, Bonan
%A Grishman, Ralph
%A Wan, Li
%A Wang, Chang
%A Gondek, David
%D 2013
%K relationsextraktion
%T Distant supervision for relation extraction with an incomplete knowledge base
%U https://cs.nyu.edu/~min/index_files/BonanMin_NAACL2013.pdf
%X Distant supervision, heuristically labeling a corpus using a knowledge base, has emerged as a popular choice for training relation extractors. In this paper, we show that a significant number of '' negative '' examples generated by the labeling process are false negatives because the knowledge base is incomplete. Therefore the heuristic for generating negative examples has a serious flaw. Building on a state-of-the-art distantly-supervised extraction algorithm, we proposed an algorithm that learns from only positive and unlabeled labels at the pair-of-entity level. Experimental results demonstrate its advantage over existing algorithms.
@article{min_distant_2013,
abstract = {Distant supervision, heuristically labeling a corpus using a knowledge base, has emerged as a popular choice for training relation extractors. In this paper, we show that a significant number of '' negative '' examples generated by the labeling process are false negatives because the knowledge base is incomplete. Therefore the heuristic for generating negative examples has a serious flaw. Building on a state-of-the-art distantly-supervised extraction algorithm, we proposed an algorithm that learns from only positive and unlabeled labels at the pair-of-entity level. Experimental results demonstrate its advantage over existing algorithms.},
added-at = {2018-11-04T17:02:36.000+0100},
author = {Min, Bonan and Grishman, Ralph and Wan, Li and Wang, Chang and Gondek, David},
biburl = {https://www.bibsonomy.org/bibtex/26caab1a936d36677a98f49088a83db4e/lepsky},
interhash = {994244122f646c1f826af09bc023c8c4},
intrahash = {6caab1a936d36677a98f49088a83db4e},
keywords = {relationsextraktion},
timestamp = {2018-11-04T17:02:36.000+0100},
title = {Distant supervision for relation extraction with an incomplete knowledge base},
url = {https://cs.nyu.edu/~min/index_files/BonanMin_NAACL2013.pdf},
year = 2013
}