An Interactive Framework of Cross-Lingual NLU for In-Vehicle Dialogue

Area of Science:

• Natural Language Processing
• Artificial Intelligence
• Human-Computer Interaction

Background:

• Globalization increases linguistic diversity in in-vehicle communication.
• Existing systems struggle with cross-lingual natural language understanding (NLU) in vehicles.
• Need for enhanced NLU to support diverse language speakers in automotive environments.

Purpose of the Study:

• To propose an interactive attention-based contrastive learning framework (IABCL) for in-vehicle dialogue.
• To enhance cross-lingual NLU capabilities in automotive systems.
• To address challenges in cross-lingual interaction within in-vehicle dialogue.

Main Methods:

• Utilizes contrastive learning in the encoder to improve cross-lingual understanding by distinguishing similar meanings across languages.
• Applies an attention mechanism in the decoder to articulate slots and intents, enhancing NLU within language families.
• Constructed a multilingual in-vehicle dialogue (MIvD) dataset for evaluation.

Main Results:

• The IABCL framework demonstrated improved performance in cross-lingual dialogue tasks.
• Achieved a 2.42% improvement in intent recognition.
• Showed a 1.43% improvement in slot filling and a 2.67% overall enhancement compared to the latest models.

Conclusions:

• The IABCL framework effectively enhances cross-lingual NLU for in-vehicle dialogue systems.
• The proposed method offers a robust solution for handling linguistic diversity in automotive communication.
• IABCL shows significant potential for future development in multilingual intelligent vehicle systems.