Hasso-Plattner-Institut
Prof. Dr. Felix Naumann
 

Knowledge Graphs Representation and Reasoning

Knowledge Graphs (KGs) are widely used for systematic representation of real world data in form of entities and relations.  KGs rely on an underlying schema or ontology that consists of the concepts (that the entities are instances of) and their connections with each other. This ontology lays the ground rules for populating the KG with real world data. Ontologies encapsulate the necessary data semantics that can enable machine understanding of KG facts. Ontology design and construction is one of the first and most important steps for KG construction, yet it has largely remained a manual task, particularly for new domains. This project explores data-driven techniques for ontology learning and expansion that can in turn address the issue of lack of coverage of facts in KGs.

Subprojects

Learning Fine-Grained Semantics for Multi-Relational Data

The semantics of relations play a central role in the understanding and analysis of multi-relational data. Real-world relational datasets represented by knowledge graphs often contain polysemous relations between different types of entities, that represent multiple semantics. In this work, we present a data-driven method that can automatically discover the distinct semantics associated with high-level relations and derive an optimal number of sub-relations having fine-grained meaning. To this end, we perform clustering over vector representations of entities and relations obtained from knowledge graph embedding models. Details can be found in the paper and poster.

Knowledge Graph Embeddings and Semantics

Knowledge graph embeddings that generate vector space representations of knowledge graph triples, have gained considerable popularity in past years. Several embedding models have been proposed that achieve state-of-the-art performance for the task of triple completion in knowledge graphs. Relying on the presumed semantic capabilities of the learned embeddings, they have been leveraged for various other tasks such as entity typing, rule mining and conceptual clustering. However, a critical analysis of the utility as well as limitations of these embeddings for semantic representation of the underlying entities and relations has not been performed by previous work. In this work, we perform a systematic evaluation of popular knowledge graph embedding models to obtain a better understanding of their semantic capabilities as compared to a non-embedding based approach.

Further details can be found in the associated publication at ESWC 2021 conference.

Reasoning for Improving Knowledge Graph Embeddings

While KG embedding models might learn a good representation of the input KG, but due to the nature of machine learning approaches, they often lose the semantics of entities and relations, which might lead to nonsensical predictions. To address this issue we aim to improve the accuracy of embeddings using ontological reasoning. In this work, we propose a reasoning based approach to generate negative samples for knowledge graph embeddings.

This work was jointly done in collaboration with the researchers at Natural Language Processing and Semantic Reasoning Group of the Bosch Center for Artificial Intelligence (Renningen) during October 2020 - January 2021. The associated research paper titled 'Improving Knowledge Graph Embeddings with Ontological Reasoning' was accepted for presentation at the International Semantic Web Conference 2021 (ISWC 2021).

Associated Activities

  • SoSe 2021 - Knowledge Graphs (Seminar)
  • SoSe 2019 - Text Visualisation in Practice (Project Seminar)
  • WiSe 2021-22 - Relation Canonicalization in Knowledge Graphs (Research Project)
    • Student: Maria Lamaeva, MSc Data Science, University of Potsdam

 

Related Publications

  • Nitisha Jain : Knowledge Graph Representation with Embeddings. Featured talk at the the 2nd Workshop on Unstructured and Structured KBs at AKBC 2021 conference.
    [Abstract]
  • Nitisha Jain, Jan-Christoph Kalo, Wolf-Tilo Balke, Ralf Krestel: Do Embeddings Actually Capture Knowledge Graph Semantics? Recently published papers track of  the International Conference on Principles of Knowledge Representation and Reasoning (KR), 2021.
    [ExtendedAbstract]
  • Nitisha Jain, Trung-Kien Tran, Mohamed H. Gad-Elrab, Daria Stepanova: Improving Knowledge Graph Embeddings with Ontological Reasoning. Proceedings of the International Semantic Web Conference (ISWC), 2021.
    [Paper]
  • Nitisha Jain, Jan-Christoph Kalo, Wolf-Tilo Balke, Ralf Krestel: Do Embeddings Actually Capture Knowledge Graph Semantics?. Proceedings of the Extended Semantic Web Conference (ESWC), 2021
    [Paper][URL][DOI:10.1007/978-3-030-77385-4_9]
  • Nitisha Jain, Ralf Krestel: Learning Fine-Grained Semantics for Multi-Relational Data. Proceedings of the International Semantic Web Conference, Posters and Demos (ISWC), 2020
    [Paper][Poster]
  • Nitisha Jain: Domain-Specific Knowledge Graph Construction for Semantic Analysis. Proceedings of the Extended Semantic Web Conference (ESWC), 2020
    [Paper][URL][DOI:10.1007/978-3-030-62327-2_40]
  • Simon Razniewski, Nitisha Jain, Paramita Mirza, Gerhard Weikum: Coverage of Information Extraction from Sentences and Paragraphs. Proceedings of the Conference on Empirical Methods in Natural Language Processing and the International Joint Conference on Natural Language Processing (EMNLP-IJCNLP), 2019
  • [Paper][ACL Web][DOI:10.18653/v1/D19-1583]