Resources on Semantic Table Interpretation
Welcome to the website for all things related to Semantic Table Interpretation (STI). Here, you'll find a wealth of information and resources to help you understand, implement, and innovate in the field of STI. Whether you're a researcher, developer, or enthusiast, our site offers detailed insights, cutting-edge research, and practical tools to support your work in making sense of tabular data
What's Semantic Table Interpretation?
Semantic Table Interpretation, as defined by the the SemTab challenge, involves annotating relational tables with information from a Knowledge Graph (KG). This process includes associating each column in a table with one or more KG types, known as Column Type Annotation (CTA). Additionally, Cell Entity Annotation (CEA) is applied to annotate each cell in named entity columns with a KG entity or mark it as Not In Lexicon (NIL) if it does not exist in the KG. Columns Property Annotation (CPA) involves annotating pairs of columns with a KG property. The result of this annotation process is a table enriched with semantic information
TUTSTI @ISWC2024
Discover the comprehensive world of Semantic Table Interpretation (STI) in this tutorial, which covers both theoretical and practical aspects, and trace the evolution of STI from heuristic-based methods to machine learning (ML) techniques and the latest large language model (LLM) innovations. By examining the unique characteristics, advantages, and limitations of each approach you will understand their optimal contexts of use
Our Approaches, Datasets, Tools and UIs
Our Projects
MammoTab
MammoTab is a unique dataset consisting of 1 million Wikipedia tables, extracted from over 20 million Wikipedia pages, and annotated using Wikidata. This dataset fills a gap in the current state-of-the-art resources, making it an excellent tool for testing and training Semantic Table Interpretation approaches. MammoTab is specifically designed to address several key challenges, including disambiguation, homonymy, and NIL-mentions, providing a comprehensive resource for advancing STI research and applications
s-elBat
s-elBat is a Semantic Table Interpretation approach designed to perform Entity Linking (EL) on tables. It uses an iterative process to link entities within tables and includes an innovative and optimised lookup approach for generating candidate entities for annotation. This method enhances the accuracy and efficiency of STI.
MantisTable UI
MantisTable UI is the ultimate web interface for managing Semantic Table Interpretation (STI) approaches, providing an intuitive, ready-to-use platform tailored anyone who is interested in STI world.
stEELlm
stEELlm is an innovative Semantic Table Interpretation tool that transforms data analysis. Employing the precisely tailored Mixtral 8x7B model, stEELlm excels at producing exact semantic annotations across varied datasets.
LamAPI
LamAPI supports string-based retrieval but also hard and soft filters based on an input entity type (i.e., rdf:type for DBpedia and Property:P31 for Wikidata). Hard type filters remove non matching results, while soft type filters promote or demote results when an exact match is not feasible.