Interactive Semantic Parsing for If-Then Recipes via Hierarchical Reinforcement Learning
This addresses the problem of program synthesis from ambiguous text for users needing accurate semantic parsing, though it is incremental as it builds on existing interactive methods with a hierarchical reinforcement learning twist.
The paper tackles the challenge of ambiguous or incomplete text descriptions in semantic parsing by introducing an interactive approach where an agent asks clarification questions via dialogue, specifically for 'If-Then recipes', resulting in significantly improved parsing performance with minimal user questions as shown in simulation and human evaluations.
Given a text description, most existing semantic parsers synthesize a program in one shot. However, it is quite challenging to produce a correct program solely based on the description, which in reality is often ambiguous or incomplete. In this paper, we investigate interactive semantic parsing, where the agent can ask the user clarification questions to resolve ambiguities via a multi-turn dialogue, on an important type of programs called "If-Then recipes." We develop a hierarchical reinforcement learning (HRL) based agent that significantly improves the parsing performance with minimal questions to the user. Results under both simulation and human evaluation show that our agent substantially outperforms non-interactive semantic parsers and rule-based agents.