Combining Distant and Partial Supervision for Relation Extraction

Abstract

Broad-coverage relation extraction either requires expensive supervised training data, or suffers from drawbacks inherent to distant supervision. We present an approach for providing partial supervision to a distantly supervised relation extractor using a small number of carefully selected examples. We compare against established active learning criteria and propose a novel criterion to sample examples which are both uncertain and representative. In this way, we combine the benefits of fine-grained supervision for difficult examples with the coverage of a large distantly supervised corpus. Our approach gives a substantial increase of 3.9% endto-end F 1 on the 2013 KBP Slot Filling evaluation, yielding a net F 1 of 37.7%.

Unsupervised Person Slot Filling based on Graph Mining

“…In Table 3 we compare our approach with two stateof-the-art English slot filling methods: a distant supervision method ) and a hybrid method that combines distant and partial supervision (Angeli et al, 2014b). Our method outperforms both methods dramatically.…”

“…Besides the methods based on distant supervision (e.g., (Surdeanu et al, 2010;Angeli et al, 2014b)) discussed in Section 6.2, pattern-based methods have also been proven to be effective in SF in the past years (Sun et al, 2011;Li et al, 2012;Yu et al, 2013). Dependency-based patterns achieve better performance since they can capture long-distance relations.…”

“…Considering any pair of query and candidate slot filler as an instance, these approaches train a classifier from manually labeled data through active learning (Angeli et al, 2014b) or noisy labeled data through distant supervision (Angeli et al, 2014a;Surdeanu et al, 2010) to predict the existence of a specific relation between them.…”

Extracting Relations between Non-Standard Entities using Distant Supervision and Imitation Learning

“…Most distant supervision research focuses on addressing the disadvantages of heuristic labelling, namely reducing false positive training data (Hoffmann et al, 2011;Surdeanu et al, 2012;Riedel et al, 2013;Alfonseca et al, 2012;Roth and Klakow, 2013;Takamatsu et al, 2012;Xu et al, 2013) and dealing with false negatives due to missing entries in the knowledge base (Min et al, 2013), as well as combining distant supervision with active learning (Angeli et al, 2014) Distant supervision has been researched for different domains, including newswire Riedel et al, 2013), Wikipedia (Mintz et al, 2009;Nguyen and Moschitti, 2011), the biomedical domain (Craven and Kumlien, 1999;Roller and Stevenson, 2014), the architecture domain (Vlachos and Clark, 2014) and the Web (Xin et al, 2014;Augenstein et al, 2014;Augenstein et al, 2015). To date, there is very little research on improving NERC for distant supervision to extract relations between non-standard entities such as musical artists and albums.…”

Incorporating Trustiness and Collective Synonym/Contrastive Evidence into Taxonomy Construction

“…The extraction of the triples includes six tasks: named entity recognition, part of speech tagging, dependency parsing, triple extraction, entity linkage (which maps mentions of proper nouns and their co-references to the corresponding entities in Freebase) and relation linkage. We use three information extraction (IE) tools (Angeli et al (2014), , MITIE 3 ) for the first four tasks, and develop a method similar to Hachey et al (2013) for the last two tasks of entity linkage and relation linkage.…”