Multilingual Dialogue Generation and Localization with Dialogue Act Scripting

2025-09-26

AI Summary

Non-English conversational data is scarce, prompting existing models to rely heavily on translation for constructing training and evaluation sets—leading to unnatural outputs and cultural misalignment. To address this, we propose the Dialogue Act Scripting (DAS) framework, which employs structured dialogue acts as an intermediate representation to decouple intent modeling from linguistic realization, thereby circumventing culture-specific distortions inherent in direct translation. DAS is the first approach to integrate dialogue act encoding–decoding with multilingual natural language generation (NLG), enabling culturally grounded cross-lingual dialogue generation. Human evaluations across Italian, German, and Chinese demonstrate that DAS-generated dialogues significantly outperform both machine-translated and human-translated baselines in cultural relevance, coherence, and contextual appropriateness—effectively mitigating the “translationese” artifact.

Technology Category

• Natural Language Processing: Generation
• Machine Learning: Large Multimodal Models (LMMs)
• Planning, Routing, and Scheduling: Planning with Language Models

Application Category

• Search and Retrieval-Augmented AI: Multilingual and cross-lingual Web search
• Web Mining and Content Analysis: Mining multimedia, multimodal, multilingual, cross-lingual Web data
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions

Abstract

Non-English dialogue datasets are scarce, and models are often trained or evaluated on translations of English-language dialogues, an approach which can introduce artifacts that reduce their naturalness and cultural appropriateness. This work proposes Dialogue Act Script (DAS), a structured framework for encoding, localizing, and generating multilingual dialogues from abstract intent representations. Rather than translating dialogue utterances directly, DAS enables the generation of new dialogues in the target language that are culturally and contextually appropriate. By using structured dialogue act representations, DAS supports flexible localization across languages, mitigating translationese and enabling more fluent, naturalistic conversations. Human evaluations across Italian, German, and Chinese show that DAS-generated dialogues consistently outperform those produced by both machine and human translators on measures of cultural relevance, coherence, and situational appropriateness.

Problem

• Addresses scarcity of natural multilingual dialogue datasets
• Generates culturally appropriate dialogues from abstract intents
• Mitigates translation artifacts through structured act scripting

Innovation

• Structured framework for multilingual dialogue generation
• Generates culturally appropriate dialogues from abstract intents
• Uses dialogue act representations to avoid translation artifacts