Summary of “what’s My Model Inside Of?”: Exploring the Role Of Environments For Grounded Natural Language Understanding, by Ronen Tamari

“What’s my model inside of?”: Exploring the role of environments for grounded natural language understanding

by Ronen Tamari

First submitted to arxiv on: 4 Feb 2024

GrooveSquid.com Paper Summaries

GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!

Summary difficulty	Written by	Summary
High	Paper authors	High Difficulty Summary Read the original abstract here
Medium	GrooveSquid.com (original content)	Medium Difficulty Summary This thesis adopts an ecological approach to grounded natural language understanding (NLU) research, focusing on the role of environment and body in shaping cognition. Unlike classic NLU studies, which design new models and optimize methods while treating environments as given, this work explores the potential of environment design for improving data collection and model development. Novel training and annotation approaches were developed for procedural text understanding based on text-based game environments, leveraging embodied cognitive linguistics to inform a roadmap for grounded NLP research. A new benchmark was proposed for measuring large language models’ progress on challenging commonsense reasoning tasks. Breakpoint Transformers, a novel approach to modeling intermediate semantic information in long narrative or procedural texts, was developed by leveraging the richer supervision provided by text-based game environments.
Low	GrooveSquid.com (original content)	Low Difficulty Summary Imagine trying to understand a story without seeing what’s happening around you. That’s like how most language understanding research is done – in isolation. This thesis takes a different approach. It looks at how our bodies and surroundings help us make sense of language. By designing special “game-like” environments, researchers can collect more helpful data and create better models for understanding natural language. They even came up with new ways to train and label this data. The goal is to make computers better at understanding everyday conversations and stories.