FoodKG

FoodKG enriches knowledge graphs in the Food, Energy, and Water (FEW) domains by adding semantically related triples, relations, and images to improve decision-making and knowledge discovery.


Key Features:

  • Knowledge Graph Enrichment: Enriches input knowledge graphs constructed from raw FEW datasets with semantically related triples, relations, and images.
  • Graph Embedding: Uses a graph embedding technique trained on AGROVOC, the controlled vocabulary published by the Food and Agriculture Organization of the United Nations, to provide semantic context for terms and classes.
  • Semantic Similarity Scoring: Computes semantic similarity scores to quantify relatedness between entities and classes.
  • Relation Inference and Classification: Establishes relations between different classes within FEW datasets and refines classification of existing entities using inferred relations.
  • Evaluation: Compared performance against state-of-the-art word embedding and knowledge graph embedding models trained on the same dataset using the Spearman Correlation Coefficient, reporting superior correlation.

Scientific Applications:

  • Decision Support: Supports decision-making in FEW domains by revealing semantic relationships across heterogeneous datasets.
  • Knowledge Discovery: Improves knowledge discovery by adding semantically related triples and scored relations for FEW concepts.
  • Data Integration: Enables integration of heterogeneous FEW datasets into semantically coherent knowledge graphs.
  • Semantic Annotation and Classification: Facilitates precise classification and the use of scientific terminology for entities and classes within FEW datasets.

Methodology:

Enriches input knowledge graphs with semantically related triples, relations, and images grounded in original dataset terms and classes; employs a graph embedding technique trained on AGROVOC; computes semantic similarity scores and establishes inter-class relations; evaluated against word embedding and knowledge graph embedding models using the Spearman Correlation Coefficient.

Topics

Details

Tool Type:
command-line tool
Programming Languages:
Python
Added:
9/8/2021
Last Updated:
9/20/2021

Operations

Publications

Gharibi M, Zachariah A, Rao P. FoodKG: A Tool to Enrich Knowledge Graphs Using Machine Learning Techniques. Frontiers in Big Data. 2020;3. doi:10.3389/fdata.2020.00012. PMID:33693387. PMCID:PMC7931944.

Links