Trainable Methods for Surface Natural Language Generation

Abstract

We present three systems for surface natural language generation that are trainable from annotated corpora. The first two systems, called NLG1 and NLG2, require a corpus marked only with domain-specific semantic attributes, while the last system, called NLG3, requires a corpus marked with both semantic attributes and syntactic dependency information. All systems attempt to produce a grammatical natural language phrase from a domain-specific semantic representation. NLG1 serves a baseline system and uses phrase frequencies to generate a whole phrase in one step, while NLG2 and NLG3 use maximum entropy probability models to individually generate each word in the phrase. The systems NLG2 and NLG3 learn to determine both the word choice and the word order of the phrase. We present experiments in which we generate phrases to describe flights in the air travel domain.

LaTeX, 8 pages

Cited by in corpus (11)

• Multimodal Machine Learning: A Survey and Taxonomy
• Generating Natural Language Descriptions from OWL Ontologies: the NaturalOWL System
• From Captions to Visual Concepts and Back
• CycleGT: Unsupervised Graph-to-Text and Text-to-Graph Generation via Cycle Training
• Universal Model for Paraphrasing -- Using Transformation Based on a Defined Criteria --
• Learning to Order Facts for Discourse Planning in Natural Language Generation
• Strategies for Structuring Story Generation
• Semantic Refinement GRU-based Neural Language Generation for Spoken Dialogue Systems
• Bootstrapping Lexical Choice via Multiple-Sequence Alignment
• Extracting Linguistic Resources from the Web for Concept-to-Text Generation
• Modeling informational novelty in a conversational system with a hybrid statistical and grammar-based approach to natural language generation