Confidence-Aware Negative Sampling Method for Noisy Knowledge Graph Embedding

Shan, Yingchun; Bu, Chenyang; Liu, Xiaojian; Ji, Shengwei; Li, Lei

doi:10.1109/icbk.2018.00013

Cited by 22 publications

(16 citation statements)

References 13 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…However, the relevant methods only aim to detect the anomalies without iterative learning by using the detected anomalies to obtain more accurate representation. Although confidence-based approaches [28,42] were proposed to incorporate triple confidence into knowledge graph representation learning model to jointly detect the noise and learn the embedding, this kind of methods cannot be easily used in our entity alignment scenario. The reason is that our task is to detect the noisy links between two knowledge graphs rather than triples within a single knowledge graph.…”

Section: Anomaly Detection and Robust Representation Learning On Graphsmentioning

confidence: 99%

“…It is inevitable that noises exist in the labeled entity pairs owing to different labeled pair collection processes as discussed before. Inspired by confidence-based methods [28,42], we introduce a trust score into our proposed model to describe the likelihood of a labeled entity pair (e x , e y ) being real. Hence, the margin-based ranking objective function of our noise-aware entity alignment can be defined with the trust score as follows:…”

Section: 22mentioning

confidence: 99%

See 1 more Smart Citation

REA: Robust Cross-lingual Entity Alignment Between Knowledge Graphs

Pei

et al. 2020

Proceedings of the 26th ACM SIGKDD International Conference on Knowledge Discovery &Amp; Data Mining

View full text Add to dashboard Cite

Cross-lingual entity alignment aims at associating semantically similar entities in knowledge graphs with different languages. It has been an essential research problem for knowledge integration and knowledge graph connection, and been studied with supervised or semi-supervised machine learning methods with the assumption of clean labeled data. However, labels from human annotations often include errors, which can largely affect the alignment results. We thus aim to formulate and explore the robust entity alignment problem, which is non-trivial, due to the deficiency of noisy labels. Our proposed method named REA (Robust Entity Alignment) consists of two components: noise detection and noise-aware entity alignment. The noise detection is designed by following the adversarial training principle. The noise-aware entity alignment is devised by leveraging graph neural network based knowledge graph encoder as the core. In order to mutually boost the performance of the two components, we propose a unified reinforced training strategy to combine them. To evaluate our REA method, we conduct extensive experiments on several real-world datasets. The experimental results demonstrate the effectiveness of our proposed method and also show that our model consistently outperforms the state-of-theart methods with significant improvement on alignment accuracy in the noise-involved scenario.

show abstract

Section: Anomaly Detection and Robust Representation Learning On Graphsmentioning

confidence: 99%

Section: 22mentioning

confidence: 99%

REA: Robust Cross-lingual Entity Alignment Between Knowledge Graphs

Pei

et al. 2020

Proceedings of the 26th ACM SIGKDD International Conference on Knowledge Discovery &Amp; Data Mining

View full text Add to dashboard Cite

show abstract

“…We find that there are some novel negative sampling methods that cannot be simply classified into the above three categories, such as confidence-aware negative sampling [20] and Markov chain Monte Carlo negative sampling [21].…”

Section: Other Novel Approachesmentioning

confidence: 99%

“…Noise and conflicts are inevitably involved due to the auto-construction, explosive growth and frequent updates of typical KGs. Xie et al [19] initially proposes a novel confidence-aware KRL framework (CKRL), and Shan et al [20]extends this idea to negative sampling in noisy KRL (NKRL). CKRL detects noises but applies uniform negative sampling that easily causes zero loss problems and false detection issues.…”

Section: Nkrlmentioning

confidence: 99%

See 1 more Smart Citation

Negative Sampling in Knowledge Representation Learning: A Mini-Review

Qian¹,

Li²,

Atkinson³

et al. 2020

Computer Science &Amp; Information Technology (CS &Amp; IT)

View full text Add to dashboard Cite

Knowledge representation learning (KRL) aims at encoding components of a knowledge graph (KG) into a low-dimensional continuous space, which has brought considerable successes in applying deep learning to graph embedding. Most famous KGs contain only positive instances for space efficiency. Typical KRL techniques, especially translational distance-based models, are trained through discriminating positive and negative samples. Thus, negative sampling is unquestionably a non-trivial step in KG embedding. The quality of generated negative samples can directly influence the performance of final knowledge representations in downstream tasks, such as link prediction and triple classification. This review summarizes current negative sampling methods in KRL and we categorize them into three sorts, fixed distribution-based, generative adversarial net (GAN)-based and cluster sampling. Based on this categorization we discuss the most prevalent existing approaches and their characteristics.

show abstract