English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Knowledge graph embedding models (KGEMs) are used for various tasks relatedto knowledge graphs (KGs), including link prediction. They are trained withloss functions that are computed considering a batch of scored triples andtheir corresponding labels. Traditional approaches consider the label of atriple to be either true or false. However, recent works suggest that allnegative triples should not be valued equally. In line with this recentassumption, we posit that negative triples that are semantically valid w.r.t.domain and range constraints might be high-quality negative triples. As such,loss functions should treat them differently from semantically invalid negativeones. To this aim, we propose semantic-driven versions for the three main lossfunctions for link prediction. In an extensive and controlled experimentalsetting, we show that the proposed loss functions systematically providesatisfying results on three public benchmark KGs underpinned with differentschemas, which demonstrates both the generality and superiority of our proposedapproach. In fact, the proposed loss functions do (1) lead to better MRR andHits@10 values, (2) drive KGEMs towards better semantic awareness as measuredby the Sem@K metric. This highlights that semantic information globallyimproves KGEMs, and thus should be incorporated into loss functions. Domainsand ranges of relations being largely available in schema-defined KGs, thismakes our approach both beneficial and widely usable in practice.

Treat Different Negatives Differently: Enriching Loss Functions with Domain and Range Constraints for Link Prediction