Skip to main content Skip to main navigation


Joint Proceedings of ISWC 2018 Workshops SemDeep-4 and NLIWOD-4

Key-Sun Choi; Luis Espinosa Anke; Thierry Declerck; Dagmar Gromann; Jin-Dong Kim; Axel-Cyrille Ngonga Ngomo; Muhammad Saleem; Ricardo Usbeck (Hrsg.)
Workshop on Semantic Deep Learning (SemDeep-2018), located at 17th International Semantic Web Conference, October 8, Monterey, USA, Vol. 2241, ISBN 1613-0073, CEURS, 10/2018.


Semantic Web (SW) Technologies and Deep Learning (DL) share the goal of creating intelligent artifacts. Both disciplines have had a remarkable impact in data and knowledge analysis, as well as knowledge representation, and in fact constitute two complementary directions for modeling expressible linguistic phenomena and solving semantically complex problems. In this context, and following the main foundations set in past editions, the 4th Workshop on Semantic Deep Learning (SemDeep-4) aims at bringing together SW and DL research as well as industrial communities. SemDeep is interested in contributions of DL to classic problems in semantic applications, such as: (semi-automated) ontology learning, ontology alignment, ontology annotation, duplicate recognition, ontology prediction, knowledge base completion, relation extraction, and semantically grounded inference, among many others. At the same time, we invite contributions that analyze the interaction of SW technologies and resources with DL architectures, such as knowledge-based embeddings, collocation discovery and classification, or lexical entailment, to name only a few. This workshop seeks to provide an invigorating environment where semantically challenging problems which appeal to both Semantic Web and Computational Linguistic communities are addressed and discussed. This half-day workshop consists of one invited talk by Marco Rospocher, one short paper presentation, and four long paper presentations. Marco Rospocher talks about results of applying Neural Networks to the task of learning expressive ontological concept descriptions from natural language text. A wide variety of topics related to the broader topic of combining Semantic Web technologies with Deep Learning techniques were submitted to this workshop. An extension of information considered when training knowledge graph embeddings to literals has been proposed as well as a new evaluation benchmark for knowledge graph embeddings based on link prediction methodologies. Within the domain of academic search, a new entity retrieval system combining paragraph and knowledge graph embeddings has been proposed and an ontology-based annotation system o academic publications utilizing deep learning forms part of this workshop. Finally, the best paper of this workshop was awarded to a neural network-based method to create semantic profiles of user interests emerging from pictures by combining object recognition methods with object category generalization.