Detection of Relation Assertion Errors in Knowledge Graphs

Melo, André; Paulheim, Heiko

doi:10.1145/3148011.3148033

Cited by 51 publications

(31 citation statements)

References 24 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Therefore, noise detection is essential and significant in knowledge automatic construction and knowledge-driven intelligent applications. Most knowledge graph noise detection works happen when constructing knowledge graphs [ 15 , 25 , 26 ]. These approaches are usually involved with huge human efforts, which are extremely labor-intensive and time-consuming.…”

Section: Related Workmentioning

confidence: 99%

“…Paulheim and Bizer [ 31 ] propose the heuristic link-based type inference mechanism SDType, which can handle noisy and incorrect data. Melo and Paulheim [ 26 ] investigate the problem of error detection in relation assertions of knowledge graphs, and propose an error detection method which relies on path and type features used by a classifier for every relation in the graph exploiting local feature selection. Recently, Xie et al [ 23 ] propose an embedding method (CKRL) with confidence to deal with noise detection, however, it ignores the rich semantic information in external nonstructural information which is strong evidence to judge triple quality.…”

Section: Related Workmentioning

confidence: 99%

“…Knowledge graph refinement (KGR) is essential after automatic KG construction [ 11 ], since the result may never be perfect whichever approach is taken for constructing knowledge graph. Various methods for KGR have been proposed [ 25 ], which can differ along three distinct orthogonal dimensions: (i) the overall goal of the method, i.e., completion [ 17 , 19 ] vs. correction [ 15 , 26 ] of KG; (ii) the refinement target (e.g., relations between entities [ 19 ], entity types [ 47 ]), and (iii) the data used by the approach (i.e., only KG itself [ 17 ], or further external information [ 42 , 43 ]). However, most conventional approaches are only used for one goal as yet, while a combination between completion and error detection methods could be of great value [ 25 ].…”

Section: Related Workmentioning

confidence: 99%

See 2 more Smart Citations

Embedding Learning with Triple Trustiness on Noisy Knowledge Graph

Feng

Gallinari

2019

Entropy

View full text Add to dashboard Cite

Embedding learning on knowledge graphs (KGs) aims to encode all entities and relationships into a continuous vector space, which provides an effective and flexible method to implement downstream knowledge-driven artificial intelligence (AI) and natural language processing (NLP) tasks. Since KG construction usually involves automatic mechanisms with less human supervision, it inevitably brings in plenty of noises to KGs. However, most conventional KG embedding approaches inappropriately assume that all facts in existing KGs are completely correct and ignore noise issues, which brings about potentially serious errors. To address this issue, in this paper we propose a novel approach to learn embeddings with triple trustiness on KGs, which takes possible noises into consideration. Specifically, we calculate the trustiness value of triples according to the rich and relatively reliable information from large amounts of entity type instances and entity descriptions in KGs. In addition, we present a cross-entropy based loss function for model optimization. In experiments, we evaluate our models on KG noise detection, KG completion and classification. Through extensive experiments on three datasets, we demonstrate that our proposed model can learn better embeddings than all baselines on noisy KGs.

show abstract

Section: Related Workmentioning

confidence: 99%

Section: Related Workmentioning

confidence: 99%

Section: Related Workmentioning

confidence: 99%

See 1 more Smart Citation

Embedding Learning with Triple Trustiness on Noisy Knowledge Graph

Feng

Gallinari

2019

Entropy

View full text Add to dashboard Cite

show abstract

“…DBpedia [2] as well as Milne et al [29,30] extract content from Wikipedia on a large scale and make it available in structured (linked) form, while YAGO [27], UNIpedia [19,20] and BabelNet [31] add knowledge from other sources as well. Paulheim examines in [33] methods for rening knowledge graphs and suggests in [28] how to recognize relationship assertion errors in knowledge graphs.…”

Section: Motivating Examplementioning

confidence: 99%

Enabling the Continuous Evolution of Ontologies for Ontology-Based Data Management

Pomp

Lipp

Meisen

2019

IJRC

View full text Add to dashboard Cite

Companies suer from heterogeneous data sources that are distributed across all business units and locations. Accessing, discovering, and understanding these data sources is challenging because data scientists have to deal with dierent protocols, data formats, and even company-specic organizational issues such as rewalls and privacy policies. Ontology-Based Data Management (OBDM) provides dierent aspects for reducing the barriers of integrating, accessing and managing heterogeneous data sources by using ontologies. For that, we establish a mapping between data sources and one or multiple target ontologies. However, the biggest challenges in ODBM are creation and administration of required ontologies. Usually, these ontologies are created in advance, for instance, by ontology engineers and domain experts that closely work together for manually designing and even maintaining the ontology. In order to enhance the process of designing and maintaining an ontology, we propose an approach consisting of an evolving knowledge graph that includes an internal ontology, which continuously evolves on demand as domain experts add new data sources and dene the mapping between the ontology and that data source. For this purpose, we develop an intuitive, user-oriented wizard and combine it with a semi-supervised evolution strategy that supports the user with the help of external knowledge databases. Moreover, we equip the system with additional logic that allows to automatically link related concepts of the ontology. We evaluate the accuracy and usability of our approach by conducting a user 20 André Pomp et al. study. The results show that mappings become more objective and consistent with our provided user wizard, resulting in a knowledge graph with higher connectivity and stability.

show abstract

“…However, as this method only relies on the type of the entity, it has limited error detection performance (Ringler and Paulheim, 2017). Another error detection approach, PaTyBRED (Paths and Types with Binary Relevance for Error Detection) (Melo and Paulheim, 2017), based on machine learning includes the connections between the entities in the subject and object positions of a predicate to detect the error. The inclusion of extra knowledge in error detection has been able to improve the predictions produced by this method, compared to SDValidate (Paulheim and Pan, 2012).…”

Section: Error Detectionmentioning

confidence: 99%

A Review of Machine Learning Applications in Semantic Web

Albannai¹,

Gültepe²,

Najih³

2019

JSTR

View full text Add to dashboard Cite

In recent years, the rapid growth of the semantic web has encouraged the use of knowledge graphs to store and exchange data among computers. The change in data storage and retrieval has imposed the need to integrate other techniques that interact with these data. Machine Learning (ML) is the most important fields of study that interacts with data, to extract the patterns and relations that can be used in future applications. However, these techniques are implemented to be used with structured data, where each data instance is characterized using a set of predefined features. Thus, to apply machine learning techniques to knowledge graphs, it is important to convert the formation of the information in these graphs. The knowledge graph represents a collection of interconnected descriptions of entities (objects, events, situations, or abstract concepts in the real world). This work reviews the knowledge graphs and how they are structured and stored. Then, the techniques being used to employ machine learning technique to improve the knowledge graphs are illustrated. Additionally, the integration of the information in the knowledge graphs, so that, they can be used in machine learning techniques are also reviewed. This review can provide a rigid overview for future researchers to provide more attention on the main drawbacks in the existing techniques. Moreover, the review shows that the existing integration approaches limit the performance of the machine learning techniques, as the knowledge extracted from the knowledge graph is trimmed to match the requirements of the ML technique. In conclusion, new machine learning techniques that can extract knowledge directly from the knowledge graph can provide significantly better performance.

show abstract

Detection of Relation Assertion Errors in Knowledge Graphs

Cited by 51 publications

References 24 publications

Embedding Learning with Triple Trustiness on Noisy Knowledge Graph

Embedding Learning with Triple Trustiness on Noisy Knowledge Graph

Enabling the Continuous Evolution of Ontologies for Ontology-Based Data Management

A Review of Machine Learning Applications in Semantic Web

Contact Info

Product

Resources

About