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

Multimodal knowledge graph completion (MKGC) aims to predict missing entitiesin MKGs. Previous works usually share relation representation acrossmodalities. This results in mutual interference between modalities duringtraining, since for a pair of entities, the relation from one modality probablycontradicts that from another modality. Furthermore, making a unifiedprediction based on the shared relation representation treats the input indifferent modalities equally, while their importance to the MKGC task should bedifferent. In this paper, we propose MoSE, a Modality Split representationlearning and Ensemble inference framework for MKGC. Specifically, in thetraining phase, we learn modality-split relation embeddings for each modalityinstead of a single modality-shared one, which alleviates the modalityinterference. Based on these embeddings, in the inference phase, we first makemodality-split predictions and then exploit various ensemble methods to combinethe predictions with different weights, which models the modality importancedynamically. Experimental results on three KG datasets show that MoSEoutperforms state-of-the-art MKGC methods. Codes are available athttps://github.com/OreOZhao/MoSE4MKGC.

MoSE: Modality Split and Ensemble for Multimodal Knowledge Graph Completion