Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Autoencoder-Based Domain Learning for Semantic Communication with Conceptual Spaces

Dylan Wheeler, Balasubramaniam Natarajan

TL;DR

This work develops a framework for learning a domain of a conceptual space model using only the raw data with high-level property labels, and shows that the domains learned using the framework maintain semantic similarity relations and possess interpretable dimensions.

Abstract

Communication with the goal of accurately conveying meaning, rather than accurately transmitting symbols, has become an area of growing interest. This paradigm, termed semantic communication, typically leverages modern developments in artificial intelligence and machine learning to improve the efficiency and robustness of communication systems. However, a standard model for capturing and quantifying the details of "meaning" is lacking, with many leading approaches to semantic communication adopting a black-box framework with little understanding of what exactly the model is learning. One solution is to utilize the conceptual spaces framework, which models meaning explicitly in a geometric manner. Though prior work studying semantic communication with conceptual spaces has shown promising results, these previous attempts involve hand-crafting a conceptual space model, severely limiting the scalability and practicality of the approach. In this work, we develop a framework for learning a domain of a conceptual space model using only the raw data with high-level property labels. In experiments using the MNIST and CelebA datasets, we show that the domains learned using the framework maintain semantic similarity relations and possess interpretable dimensions.

Autoencoder-Based Domain Learning for Semantic Communication with Conceptual Spaces

TL;DR

This work develops a framework for learning a domain of a conceptual space model using only the raw data with high-level property labels, and shows that the domains learned using the framework maintain semantic similarity relations and possess interpretable dimensions.

Abstract

Communication with the goal of accurately conveying meaning, rather than accurately transmitting symbols, has become an area of growing interest. This paradigm, termed semantic communication, typically leverages modern developments in artificial intelligence and machine learning to improve the efficiency and robustness of communication systems. However, a standard model for capturing and quantifying the details of "meaning" is lacking, with many leading approaches to semantic communication adopting a black-box framework with little understanding of what exactly the model is learning. One solution is to utilize the conceptual spaces framework, which models meaning explicitly in a geometric manner. Though prior work studying semantic communication with conceptual spaces has shown promising results, these previous attempts involve hand-crafting a conceptual space model, severely limiting the scalability and practicality of the approach. In this work, we develop a framework for learning a domain of a conceptual space model using only the raw data with high-level property labels. In experiments using the MNIST and CelebA datasets, we show that the domains learned using the framework maintain semantic similarity relations and possess interpretable dimensions.
Paper Structure (13 sections, 6 equations, 5 figures, 1 algorithm)

This paper contains 13 sections, 6 equations, 5 figures, 1 algorithm.

Table of Contents

  1. Introduction
  2. Related Work
  3. Contributions
  4. Problem Definition
  5. Proposed Approach
  6. Encoder Module
  7. Decoder Module
  8. Classifier Module
  9. Putting it Together: Domain Learning Algorithm
  10. Experimental Results
  11. Experiment 1: MNIST
  12. Experiment 2: CelebA
  13. Conclusion

Figures (5)

  • Figure 1: Semantic similarity vs. semantic distortion.
  • Figure 2: Architecture of the domain learning network.
  • Figure 3: Architecture of the encoder and decoder used in the experiments.
  • Figure 4: The learned representations of the validation samples of the MNIST dataset for (a) a basic classifier (b) a basic autoencoder and (c) the proposed domain learning framework.
  • Figure 5: The learned representations of the validation samples of the CelebA dataset, and decoded representations at the edge of the domain. The domain encodes smiling/not smiling primarily along the x-axis and female/male primarily along the y-axis.

Theorems & Definitions (5)

  • Definition 1: Quality Dimension
  • Definition 2: Domain
  • Definition 3: Property
  • Definition 4: Semantic Distortion
  • Definition 5: Semantic Similarity