Abstract

Large natural language models (such as GPT-2 or T5) demonstrate impressive abilities across a range of general NLP tasks. Here, we show that the knowledge embedded in such models provides a useful inductive bias, not just on traditional NLP tasks, but also in the nontraditional task of training a symbolic reasoning engine. We observe that these engines learn quickly and generalize in a natural way that reflects human intuition. For example, training such a system to model block-stacking might naturally generalize to stacking other types of objects because of structure in the real world that has been partially captured by the language describing it. We study several abstract textual reasoning tasks, such as object manipulation and navigation, and demonstrate multiple types of generalization to novel scenarios and the symbols that comprise them. We also demonstrate the surprising utility of compositional learning, where a learner dedicated to mastering a complicated task gains an advantage by training on relevant simpler tasks instead of jumping straight to the complicated task.

Degree

MS

College and Department

Physical and Mathematical Sciences; Computer Science

Rights

https://lib.byu.edu/about/copyright/

Date Submitted

2020-12-23

Document Type

Thesis

Handle

http://hdl.lib.byu.edu/1877/etd12007

Keywords

natural language processing, deep learning, machine learning, reasoning

Language

english

Share

COinS