Document Type
Article
Publication Date
2022
Publication Title
Proceedings of Machine Learning Research
Abstract
In this work, we are concerned with developing cognitive representations that may en- hance the ability for self-supervised learning systems to learn language as part of their world explorations. We apply insights from in-depth language understanding systems to the problem, specifically representations which decompose language inputs into language-free structures that are complex combinations of primitives representing cognitive abstractions such as object permanence, movement, and spatial relationships. These decompositions, performed by a system traditionally called a conceptual analyzer, link words with complex non-linguistic structures that engender the rich relations between language expressions and world exploration that are a familiar aspect of intelligence.
We focus on improving and extending both the Conceptual Dependency (CD) representation system, its primitive decompositions, and its conceptual analyzer, choosing as our corpus the ProPara (“Process Paragraphs”) dataset, which consists of paragraphs describing biological, chemical, and physical processes of the kind that appear in grade- school science textbooks (e.g., photosynthesis, erosion). In doing so, we avoid the significant challenges of decomposing concepts involving communication, thought, and complex social interactions. To meet the challenges of this dataset, we contribute a mental motion pictures representation system with important innovations, such as using image schemas in place of CD primitives and decoupling containment relationships into separate primitives.
Keywords
Self-Supervised Learning, Natural Language Understanding, Conceptual Primitives, Conceptual Representations
Volume
192
First Page
22
Last Page
34
Rights
© 2022 M. Zhou, B. Duah & J.C. Macbeth.
Recommended Citation
Zhoou, Mackie; Duah, Bridget; and Macbeth, Jamie C., "Novel Primitive Decompositions for Real-World Physical Reasoning" (2022). Computer Science: Faculty Publications, Smith College, Northampton, MA.
https://scholarworks.smith.edu/csc_facpubs/371
Comments
Archived as published.
International Workshop on Self-Supervised Learning