Machine learning and data mining methods have become a prominent technology of our time, and are playing an increasingly important role for the Semantic Web. One of the reasons for this is the ability to deal with imperfect data, which is inherent for web data and data at scale in general. Now, we see a wide adoption of machine learning and data mining methods for tasks, such as knowledge extraction, entity (inter-)linking, rule mining, entity retrieval, knowledge graph construction, cleaning, and completion, as well as ontology management.
At the same time, semantic technologies offer semantically interpretable tools for reasoning on structured data and could facilitate explanations in machine learning systems making them more transparent and suitable for critical applications such as healthcare.
This track invites high quality submissions that focus on the one hand, on the definition, adoption and assessment of new machine learning methods for the Semantic Web, and on the other hand, on exploitation of semantic technologies for the improvement of machine learning models. Papers showing how machine learning methods and techniques have been applied to develop and improve the Semantic Web field or showing how Semantic Web technologies and resources have been used to enhance different machine learning tasks are very welcome. Reproducibility of all presented works is strictly required and should ideally be supported by providing links to used datasets, source code, queries/test cases or live deployments.
As a new theme in 2021, ESWC encourages the submission of negative results papers. Specific instructions for negative results papers can be found here.
Topics of Interest
Topics of interest include, but are not limited to:
- Machine learning-based methods and algorithms for knowledge graph construction and curation
- Statistical relational learning for knowledge graphs and ontologies
- Neuro-symbolic reasoning
- Inductive logic programming for the Semantic Web
- Data mining and knowledge discovery from knowledge bases, linked data and ontologies
- Knowledge graph embeddings
- Graph Neural Networks applied to Knowledge Graphs
- Semantic technologies for explainable machine learning
- Representation learning techniques for knowledge graphs and linked data, especially hybrid models that combine text, graph structure and semantics
- Machine learning for information extraction, retrieval and semantic search
- Learning from big data for large-scale knowledge graphs
- Machine learning approaches towards question and approximate query answering
- Exploratory data analysis in the context of knowledge graphs
Delineation from other tracks.
Knowledge Graphs: If you are proposing a large system composed of many techniques which work together to create a knowledge graph, you should submit to the Knowledge Graph track. If you, instead, are working on one specific technique which uses machine learning to aid in knowledge graph construction, the Machine Learning track is more suitable.
There are many cases where machine learning is used to achieve a specific aim in other fields, e.g., question answering, information retrieval, or ontology matching. As a rule of thumb: if the core innovative idea of your paper is a novel ML method or approach, then the Machine Learning track is more suitable. If, on the other hand, the core innovative idea is some way to improve information retrieval, ontology matching etc. by using some existing ML method or approach, then the respective NLP and Information Retrieval and Ontology Matching tracks are more suitable.
Information on deadlines and submission formats can be found here.
- Michael Cochez, Vrije Universiteit Amsterdam
- Daria Stepanova, Bosch Center for AI