Generative Adversarial Zero-Shot Relational Learning for Knowledge Graphs

Qin, Pengda; Wang, Xin; Chen, Wenhu; Zhang, Chunyun; Wang, Xu; Wang, William Yang

doi:10.48550/arxiv.2001.02332

Cited by 5 publications

(6 citation statements)

References 17 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The RE/C task is formulated as a textual entailment problem [37]. [171] also learns the semantic features of relations from their text description, it proposes to leverage GAN to establish the connection between text and knowledge graph domain.…”

Section: IIImentioning

confidence: 99%

“…The RE/C performance will be greatly affected by the quality and relevance of auxiliary knowledge. However, the benefits of prior knowledge are still incorporated into other LSL RE/C methods [171,200,240,242,265]. For example, in semi-supervised methods [240,265], fresh materials are extracted from the external knowledge library to join the training.…”

Section: IIImentioning

confidence: 99%

“…Application LSL type Data resource Data format FewRel [84] Relation classification FSL Wikipedia [220] Sentence & Relation name FewRel 2.0 [74] Relation classification FSL Wikipedia [220], PubMed, SemEval-2010 task-8 [90] Sentence & Relation name FewEvent [41] Event detection FSL ACE-2005 corpus, TAC-KBP-2017 Event Track Data, Freebase [15], Wikipedia [220] Sentence & Mention zero-shot EL [134] Entity linking ZSL Wikia Sentence & < Mention, Description > WikiLinks NED Unseen-Mentions [160] Entity linking ZSL WikilinksNED dataset [54] Sentence & < Mention, Categories > ManyEnt [49] Entity typing FSL FewRel [84] Sentence & Mention, Type NELL-ZS [171] Relation learning ZSL NELL [241] Sentence & < Triple, Description > NELL-One [242] Relation learning FSL NELL [241] Triple Wiki-ZS [171] Relation learning ZSL Wikipedia [220] Sentence & < Triple, Description > Wiki-One [242] Relation learning FSL Wikipedia [220] Triple UW-RE [116] Relation classification ZSL Wikipedia [220] Sentence & < Relation name, Description > FB20K [237,274] Graph representation ZSL Freebase [15], FB15K [214] Triple & Entity description for Train/Val/Test). Its difference from the other regular entity linking datasets is that the entities in the validation and test sets are from completely different domains than the train set.…”

Section: Dataset Namementioning

confidence: 99%

“…NELL-ZS [171] and NELL-One [242]. These two datasets are constructed from the original KG dataset -NELL [241], for zero-shot and one-shot setting respectively.…”

Section: Dataset Namementioning

confidence: 99%

“…Wiki-ZS [171] and Wiki-One [242]. They are similar with above NELL-ZS and NELL-one datasets and based on Wikipedia data [220].…”

Section: Dataset Namementioning

confidence: 99%

See 4 more Smart Citations

What Can Knowledge Bring to Machine Learning? -- A Survey of Low-shot Learning for Structured Data

Hu¹,

Chapman²,

Wen³

et al. 2021

Preprint

View full text Add to dashboard Cite

Supervised machine learning has several drawbacks that make it difficult to use in many situations. Drawbacks include: heavy reliance on massive training data, limited generalizability and poor expressiveness of high-level semantics. Low-shot Learning attempts to address these drawbacks. Low-shot learning allows the model to obtain good predictive power with very little or no training data, where structured knowledge plays a key role as a high-level semantic representation of human. This article will review the fundamental factors of low-shot learning technologies, with a focus on the operation of structured knowledge under different low-shot conditions. We also introduce other techniques relevant to low-shot learning. Finally, we point out the limitations of low-shot learning, the prospects and gaps of industrial applications, and future research directions.

show abstract

Section: IIImentioning

confidence: 99%

Section: IIImentioning

confidence: 99%

Section: Dataset Namementioning

confidence: 99%

“…NELL-ZS [171] and NELL-One [242]. These two datasets are constructed from the original KG dataset -NELL [241], for zero-shot and one-shot setting respectively.…”

Section: Dataset Namementioning

confidence: 99%

“…Wiki-ZS [171] and Wiki-One [242]. They are similar with above NELL-ZS and NELL-one datasets and based on Wikipedia data [220].…”

Section: Dataset Namementioning

confidence: 99%

See 3 more Smart Citations