Affiliation:
1. Computer Science & Engineering, University of Washington, Seattle, WA 98195,
Abstract
The context in which language is used provides a strong signal for learning to recover its meaning. In this paper, we show it can be used within a grounded CCG semantic parsing approach that learns a joint model of meaning and context for interpreting and executing natural language instructions, using various types of weak supervision. The joint nature provides crucial benefits by allowing situated cues, such as the set of visible objects, to directly influence learning. It also enables algorithms that learn while executing instructions, for example by trying to replicate human actions. Experiments on a benchmark navigational dataset demonstrate strong performance under differing forms of supervision, including correctly executing 60% more instruction sets relative to the previous state of the art.
Subject
Artificial Intelligence,Computer Science Applications,Linguistics and Language,Human-Computer Interaction,Communication
Cited by
53 articles.
订阅此论文施引文献
订阅此论文施引文献,注册后可以免费订阅5篇论文的施引文献,订阅后可以查看论文全部施引文献
1. Continuous and Interactive Language Learning and Grounding;Synthesis Lectures on Human Language Technologies;2024
2. Scene-Aware Activity Program Generation with Language Guidance;ACM Transactions on Graphics;2023-12-05
3. Text to Data;Natural Language Interfaces to Databases;2023-11-25
4. Generating Executable Action Plans with Environmentally-Aware Language Models;2023 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS);2023-10-01
5. GRIDS: Personalized Guideline Recommendations while Driving Through a New City;ACM Transactions on Recommender Systems;2023-07-17