Ioannis Dikeoulias

Temporal knowledge graph completion (TKGC) has become a popular approach for reasoning over the event and temporal knowledge graphs, targeting the completion of knowledge with accurate but missing information. In this context, tensor decomposition has successfully modeled interactions between entities and relations. Their effectiveness in static knowledge graph completion motivates us to introduce Time-LowFER, a family of parameter-efficient and time-aware extensions of the low-rank tensor… Mehr anzeigen

Temporal knowledge graph completion (TKGC) has become a popular approach for reasoning over the event and temporal knowledge graphs, targeting the completion of knowledge with accurate but missing information. In this context, tensor decomposition has successfully modeled interactions between entities and relations. Their effectiveness in static knowledge graph completion motivates us to introduce Time-LowFER, a family of parameter-efficient and time-aware extensions of the low-rank tensor factorization model LowFER. Noting several limitations in current approaches to represent time, we propose a cycle-aware time-encoding scheme for time features, which is model-agnostic and offers a more generalized representation of time. We implement our methods in a unified temporal knowledge graph embedding framework, focusing on time-sensitive data processing. The experiments show that our proposed methods perform on par or better than the state-of-the-art semantic matching models on two benchmarks. Weniger anzeigen

Artificial neural networks have emerged as some of the most promising computing systems in the past decade. They are particularly influential in natural language processing tasks such as Machine Translation, Question Answering, and Text Simplification, benefiting from the vast capabilities of statistical learning models. However, despite their advantages, these systems often depend on large volumes of human-annotated data and require thorough validation and evaluation. In this paper, we… Mehr anzeigen

Artificial neural networks have emerged as some of the most promising computing systems in the past decade. They are particularly influential in natural language processing tasks such as Machine Translation, Question Answering, and Text Simplification, benefiting from the vast capabilities of statistical learning models. However, despite their advantages, these systems often depend on large volumes of human-annotated data and require thorough validation and evaluation. In this paper, we introduce MoReViewer, a graphical tool designed for the error analysis of morphosyntactic learning models. MoReViewer is tailored to enable both computational linguists and non-technical researchers to efficiently analyze morphological inflections across various languages and grammatical structures. Weniger anzeigen

Knowledge bases (KBs) contain huge amounts of facts about entities, their properties, and relations between them. They are thus the key asset in any intelligent system for tasks such as structured search and question answering. However, due to dynamics in the real world, properties and relations change over time, and stored knowledge may become outdated. While KB information evolves steadily, there is no information whether or not a KB property might be subject to change with high probability… Mehr anzeigen

Knowledge bases (KBs) contain huge amounts of facts about entities, their properties, and relations between them. They are thus the key asset in any intelligent system for tasks such as structured search and question answering. However, due to dynamics in the real world, properties and relations change over time, and stored knowledge may become outdated. While KB information evolves steadily, there is no information whether or not a KB property might be subject to change with high probability or whether it is likely to be stable. Systems exploiting KB information, however, could benefit a lot if they had access to this kind of information. In this paper, we analyze and predict the stability of KB entries, which allows to accompany entries with stability scores. Our predictive model exploits entity-based features and learns through historic data. A particular challenge to determine stability scores is that KB entries are not only added or modified due to real-world changes but also to reduce the incompleteness of KBs in general. Nevertheless, our evaluation of sample properties demonstrates the effectiveness of our method for predicting the one-year stability of KB properties. Weniger anzeigen