Building relatedness explanations from knowledge graphs

Pirrò, Giuseppe

doi:10.3233/sw-190348

Cited by 13 publications

(9 citation statements)

References 50 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…This measure is based on the Triple Frequency defined as T F (p i , p j )=log(1 + C i,j ), where C i,j counts the number of times the predicates p i and p j link the same subjects and objects. Moreover, it uses the Inverse Triple Frequency defined as IT F (p j , E)=log |E| |{pi:Ci,j >0}| (Pirrò 2019). Based on TF and ITF, for each pair of predicates p i and p j we can build a (symmetric) matrix C M where each element is C M (i, j)=T F (p i , p j ) × IT F (p j , E).…”

Section: Triple Line Graph Edge Weightingmentioning

confidence: 99%

“…This affects the dynamics of the graph thus having a direct impact on network embedding approaches like Deepwalk or node2vec that are based on the computation of random walks. To tackle this seconds challenge, we introduce a mechanism to weight the edges of G L based on predicate relatedness (Pirrò 2019). The weight of an edge between nodes of G L is equal to the semantic relatedness between the predicates in the triples of G represented by the two nodes.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Learning Triple Embeddings from Knowledge Graphs

Fionda

Pirrò

2020

AAAI

Self Cite

View full text Add to dashboard Cite

Graph embedding techniques allow to learn high-quality feature vectors from graph structures and are useful in a variety of tasks, from node classification to clustering. Existing approaches have only focused on learning feature vectors for the nodes and predicates in a knowledge graph. To the best of our knowledge, none of them has tackled the problem of directly learning triple embeddings. The approaches that are closer to this task have focused on homogeneous graphs involving only one type of edge and obtain edge embeddings by applying some operation (e.g., average) on the embeddings of the endpoint nodes. The goal of this paper is to introduce Triple2Vec, a new technique to directly embed knowledge graph triples. We leverage the idea of line graph of a graph and extend it to the context of knowledge graphs. We introduce an edge weighting mechanism for the line graph based on semantic proximity. Embeddings are finally generated by adopting the SkipGram model, where sentences are replaced with graph walks. We evaluate our approach on different real-world knowledge graphs and compared it with related work. We also show an application of triple embeddings in the context of user-item recommendations.

show abstract

Section: Triple Line Graph Edge Weightingmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Learning Triple Embeddings from Knowledge Graphs

Fionda

Pirrò

2020

AAAI

Self Cite

View full text Add to dashboard Cite

show abstract

“…There are approaches which provide explanations of such relationships to users. For example, [107,108] propose a method for generating explanations of relationships between entities in the form of most informative paths between the entities. The approach can be extended to explaining relationships among datasets which contain related entities.…”

Section: Analysis Of Semantic Relationshipsmentioning

confidence: 99%

Survey of tools for Linked Data consumption

Klímek

Škoda

Nečaský

2019

View full text Add to dashboard Cite

There is lots of data published as Linked (Open) Data (LOD/LD). At the same time, there is also a multitude of tools for publication of LD. However, potential LD consumers still have difficulty discovering, accessing and exploiting LD. This is because compared to consumption of traditional data formats such as XML and CSV files, there is a distinct lack of tools for consumption of LD. The promoters of LD use the well-known 5-star Open Data deployment scheme to suggest that consumption of LD is a better experience once the consumer knows RDF and related technologies. This suggestion, however, falls short when the consumers search for an appropriate tooling support for LD consumption. In this paper we define a LD consumption process. Based on this process and current literature, we define a set of 34 requirements a hypothetical Linked Data Consumption Platform (LDCP) should ideally fulfill. We cover those requirements with a set of 94 evaluation criteria. We survey 110 tools identified as potential candidates for an LDCP, eliminating them in 3 rounds until 16 candidates for remain. We evaluate the 16 candidates using our 94 criteria. Based on this evaluation we show which parts of the LD consumption process are covered by the 16 candidates. Finally, we identify 8 tools which satisfy our requirements on being a LDCP. We also show that there are important LD consumption steps which are not sufficiently covered by existing tools. The authors of LDCP implementations may use this survey to decide about directions of future development of their tools. LD experts may use it to see the level of support of the state of the art technologies in existing tools. Non-LD experts may use it to choose a tool which supports their LD processing needs without requiring them to have expert knowledge of the technologies. The paper can also be used as an introductory text to LD consumption.

show abstract

“…To tackle this seconds challenge, we introduce two mechanisms to weight the edges of G L . The first, specific for knowledge graphs, assigns weights on the basis of predicate relatedness [14]. The weight of an edge between nodes of G L is equal to the semantic relatedness between the predicates in the triples of G represented by the two nodes.…”

Section: Introductionmentioning

confidence: 99%

Triple2Vec: Learning Triple Embeddings from Knowledge Graphs

Fionda,

Pirró

2019

Preprint

View full text Add to dashboard Cite

Graph embedding techniques allow to learn high-quality feature vectors from graph structures and are useful in a variety of tasks, from node classification to clustering. Existing approaches have only focused on learning feature vectors for the nodes in a (knowledge) graph. To the best of our knowledge, none of them has tackled the problem of embedding of graph edges, that is, knowledge graph triples. The approaches that are closer to this task have focused on homogeneous graphs involving only one type of edge and obtain edge embeddings by applying some operation (e.g., average) on the embeddings of the endpoint nodes. The goal of this paper is to introduce Triple2Vec, a new technique to directly embed edges in (knowledge) graphs. Triple2Vec builds upon three main ingredients. The first is the notion of line graph. The line graph of a graph is another graph representing the adjacency between edges of the original graph. In particular, the nodes of the line graph are the edges of the original graph. We show that directly applying existing embedding techniques on the nodes of the line graph to learn edge embeddings is not enough in the context of knowledge graphs. Thus, we introduce the notion of triple line graph. The second is an edge weighting mechanism both for line graphs derived from knowledge graphs and homogeneous graphs. The third is a strategy based on graph walks on the weighted triple line graph that can preserve proximity between nodes. Embeddings are finally generated by adopting the SkipGram model, where sentences are replaced with graph walks. We evaluate our approach on different real world (knowledge) graphs and compared it with related work.

show abstract

Building relatedness explanations from knowledge graphs

Cited by 13 publications

References 50 publications

Learning Triple Embeddings from Knowledge Graphs

Learning Triple Embeddings from Knowledge Graphs

Survey of tools for Linked Data consumption

Triple2Vec: Learning Triple Embeddings from Knowledge Graphs

Contact Info

Product

Resources

About