Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Neural Class Expression Synthesis

N'Dah Jean Kouagou, Stefan Heindorf, Caglar Demir, Axel-Cyrille Ngonga Ngomo

TL;DR

This work introduces neural class expression synthesis (NCES), a translation-inspired approach to learn $ALC$ class expressions from sets of positive and negative examples, avoiding costly search over the concept space. NCES leverages embeddings of knowledge graphs and three neural architectures (LSTM, GRU, Set Transformer) to generate class-expression tokens, with ensemble variants improving robustness and scalability on web-scale knowledge bases. Across four benchmarks, NCES achieves fast synthesis (roughly one second per problem) and competitive or superior F-measures on large datasets, while maintaining feasible training times. The results suggest NCES is well-suited for large-scale ontology learning scenarios where traditional search-based methods struggle, and the authors provide public code and pretrained models for reproducibility.

Abstract

Many applications require explainable node classification in knowledge graphs. Towards this end, a popular ``white-box'' approach is class expression learning: Given sets of positive and negative nodes, class expressions in description logics are learned that separate positive from negative nodes. Most existing approaches are search-based approaches generating many candidate class expressions and selecting the best one. However, they often take a long time to find suitable class expressions. In this paper, we cast class expression learning as a translation problem and propose a new family of class expression learning approaches which we dub neural class expression synthesizers. Training examples are ``translated'' into class expressions in a fashion akin to machine translation. Consequently, our synthesizers are not subject to the runtime limitations of search-based approaches. We study three instances of this novel family of approaches based on LSTMs, GRUs, and set transformers, respectively. An evaluation of our approach on four benchmark datasets suggests that it can effectively synthesize high-quality class expressions with respect to the input examples in approximately one second on average. Moreover, a comparison to state-of-the-art approaches suggests that we achieve better F-measures on large datasets. For reproducibility purposes, we provide our implementation as well as pretrained models in our public GitHub repository at https://github.com/dice-group/NeuralClassExpressionSynthesis

Neural Class Expression Synthesis

TL;DR

This work introduces neural class expression synthesis (NCES), a translation-inspired approach to learn class expressions from sets of positive and negative examples, avoiding costly search over the concept space. NCES leverages embeddings of knowledge graphs and three neural architectures (LSTM, GRU, Set Transformer) to generate class-expression tokens, with ensemble variants improving robustness and scalability on web-scale knowledge bases. Across four benchmarks, NCES achieves fast synthesis (roughly one second per problem) and competitive or superior F-measures on large datasets, while maintaining feasible training times. The results suggest NCES is well-suited for large-scale ontology learning scenarios where traditional search-based methods struggle, and the authors provide public code and pretrained models for reproducibility.

Abstract

Many applications require explainable node classification in knowledge graphs. Towards this end, a popular ``white-box'' approach is class expression learning: Given sets of positive and negative nodes, class expressions in description logics are learned that separate positive from negative nodes. Most existing approaches are search-based approaches generating many candidate class expressions and selecting the best one. However, they often take a long time to find suitable class expressions. In this paper, we cast class expression learning as a translation problem and propose a new family of class expression learning approaches which we dub neural class expression synthesizers. Training examples are ``translated'' into class expressions in a fashion akin to machine translation. Consequently, our synthesizers are not subject to the runtime limitations of search-based approaches. We study three instances of this novel family of approaches based on LSTMs, GRUs, and set transformers, respectively. An evaluation of our approach on four benchmark datasets suggests that it can effectively synthesize high-quality class expressions with respect to the input examples in approximately one second on average. Moreover, a comparison to state-of-the-art approaches suggests that we achieve better F-measures on large datasets. For reproducibility purposes, we provide our implementation as well as pretrained models in our public GitHub repository at https://github.com/dice-group/NeuralClassExpressionSynthesis
Paper Structure (31 sections, 7 equations, 2 figures, 6 tables)

This paper contains 31 sections, 7 equations, 2 figures, 6 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Background
  4. Notation
  5. Description Logics
  6. Refinement Operators
  7. Class Expression Learning
  8. Knowledge Graph Embedding
  9. Permutation-Invariant Network Architectures for Set Inputs
  10. Neural Class Expression Synthesis
  11. Learning Problem
  12. Learning Approach (NCES)
  13. Neural Network Architectures
  14. Recurrent Networks (LSTM and GRU)
  15. Set Transformer
  16. ...and 16 more sections

Figures (2)

  • Figure 1: NCES architecture. DNN stands for deep neural network that produces a sequence of tokens in the vocabulary (e.g., a sequence-to-sequence or a set-to-sequence model). The input consists of positive examples (upper left, dotted green box) and negative examples (bottom left, dotted red box).
  • Figure 2: Training accuracy curves.

Theorems & Definitions (3)

  • definition thmcounterdefinition: lehmann2010concept
  • definition thmcounterdefinition
  • definition thmcounterdefinition