Multilingual Datasets for Custom Input Extraction and Explanation Requests Parsing in Conversational XAI Systems

2025-08-20

AI Summary

Current ConvXAI systems face two key challenges: (1) scarcity of training data for low-resource languages, hindering multilingual generalization; and (2) inability to robustly parse users’ free-form, ad-hoc inputs—i.e., non-predefined instances. To address these, we introduce MultiCoXQL and Compass, the first multilingual benchmark datasets supporting joint modeling of explanation intent identification and custom input extraction across Chinese, English, French, Spanish, and Arabic. Leveraging LLMs and BERT-family models, we systematically evaluate monolingual, cross-lingual, and multilingual parsing strategies. Experiments demonstrate substantial improvements in both intent recognition and input structuring performance under multilingual settings. Our work establishes a scalable data foundation and methodological framework for low-resource ConvXAI, enabling effective adaptation to diverse linguistic and user-input contexts.

Technology Category

• Natural Language Processing: Language Grounding & Multi-modal NLP
• Humans and AI: Explainable AI (XAI) for Human Understanding
• Machine Learning: Large Multimodal Models (LMMs)

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

Conversational explainable artificial intelligence (ConvXAI) systems based on large language models (LLMs) have garnered considerable attention for their ability to enhance user comprehension through dialogue-based explanations. Current ConvXAI systems often are based on intent recognition to accurately identify the user's desired intention and map it to an explainability method. While such methods offer great precision and reliability in discerning users' underlying intentions for English, a significant challenge in the scarcity of training data persists, which impedes multilingual generalization. Besides, the support for free-form custom inputs, which are user-defined data distinct from pre-configured dataset instances, remains largely limited. To bridge these gaps, we first introduce MultiCoXQL, a multilingual extension of the CoXQL dataset spanning five typologically diverse languages, including one low-resource language. Subsequently, we propose a new parsing approach aimed at enhancing multilingual parsing performance, and evaluate three LLMs on MultiCoXQL using various parsing strategies. Furthermore, we present Compass, a new multilingual dataset designed for custom input extraction in ConvXAI systems, encompassing 11 intents across the same five languages as MultiCoXQL. We conduct monolingual, cross-lingual, and multilingual evaluations on Compass, employing three LLMs of varying sizes alongside BERT-type models.

Problem

• Addressing multilingual training data scarcity in intent recognition
• Enhancing support for free-form custom user inputs
• Improving multilingual parsing performance in conversational XAI systems

Innovation

• Multilingual dataset extension for five diverse languages
• New parsing approach to enhance multilingual performance
• Dataset for custom input extraction across languages