Multilingual Semantic Parsing and Generation with Neural Models

Abstract

Meaning representation, an abstract formal language, addresses how sentence components combine to convey a coherent meaning. This involves defining rules for syntactic structures, semantic roles, logical operators, and more. By clarifying natural language, meaning representation potentially enhances text comprehension and processing for computers. In this dissertation, we delve into two important tasks related to meaning representations: semantic parsing and text generation. Our primary focus is to explore the potential and limitations of semantic parsing across multiple natural languages. Furthermore, we present a task --- generating text from meaning representations --- with the objective of assessing the challenges inherent in various languages.