Multi-Concept Representation Learning for Knowledge Graph Completion

Abstract

Knowledge Graph Completion (KGC) aims at inferring missing entities or relations by embedding them in a low-dimensional space. However, most existing KGC methods generally fail to handle the complex concepts hidden in triplets, so the learned embeddings of entities or relations may deviate from the true situation. In this paper, we propose a novel M ulti- c oncept R epresentation L earning method for KGC task (McRL), which mainly consists of a multi-concept representation module, a deep residual attention module and an interaction embedding module. Specifically, instead of the single-feature representation, the multi-concept representation module projects each entity or relation to multiple vectors to capture the complex conceptual information hidden in them. The deep residual attention module simultaneously explores the inter-connection and intra-connection between entities and relations to enhance the entity and relation embeddings corresponding to the current contextual situation. Moreover, the interaction embedding module further weakens the noise and ambiguity to obtain the optimal and robust embeddings. We conduct the link prediction experiment to evaluate the proposed method on several standard datasets, and experimental results show that the proposed method outperforms existing state-of-the-art KGC methods.

IME: Integrating Multi-curvature Shared and Specific Embedding for Temporal Knowledge Graph Completion

“…Finally, these identified features are employed for adjustable multi-curvature pooling in subsequent predictions. [38], TDN [40], and ConvE [10].…”

“…Semantic matching-based methods employ a similarity-based scoring function to evaluate the probabilities of triplets, such as Dist-Mult [47] and McRL [38]. DistMult employs matrix multiplication to model the interaction between the entity and relation.…”

“…TuckER [2] explores Tucker decomposition into the SKGC task. In addition, McRL [38] captures the complex conceptual information hidden in triplets to acquire accurate representations of entities and relations. MLI [37] simultaneously captures the coarse-grained and fine-grained information to enhance the information interaction.…”

“…In this section, we provide an overview of KGC methods from two perspectives [27,41]: Euclidean embedding-based methods and Non-Euclidean embedding-based methods.…”

Filling the gaps: leveraging large language models for temporal harmonization of clinical text across multiple medical visits for clinical prediction

“…Recent works have leveraged LLMs for clinical applications, including named entity recognition, label generation, relation extraction, etc. 21,22 We propose that LLMs, by harnessing knowledge from their extensive training data and the existing data in the EHR, will be capable of making meaningful observations that consider the temporal aspect of EHR data. LLMs have been found to be capable of providing information about a hypothetical datapoint given information about its embedding space 23 .…”