Research map / Vision-language and multimodal / CLIP and zero-shot
Attention-Informed Mixed-Language Training for Zero-Shot Cross-Lingual Task-Oriented Dialogue Systems
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Use attention-selected word pairs for MLT instead of ontology word pairs (ablation) | MLT on Multilingual WOZ 2.0 German · Multilingual WOZ 2.0 German | joint goal accuracy: 30.61 -> 36.51 | Helped |
| Use attention-selected word pairs for MLT instead of ontology word pairs (ablation) | MLT on Multilingual WOZ 2.0 Italian · Multilingual WOZ 2.0 Italian | joint goal accuracy: 36.88 -> 39.35 | Helped |
| Use attention-selected word pairs for MLT instead of human-selected word pairs (ablation) | MLT on Multilingual NLU Spanish · Multilingual NLU Spanish | intent accuracy: 77.51 -> 86.54 | Helped |
| Use attention-selected word pairs for MLT instead of human-selected word pairs (ablation) | MLT on Multilingual NLU Spanish · Multilingual NLU Spanish | slot F1: 74.51 -> 74.43 | Mixed |
| Add a transformer encoder to map subword outputs to word-level representations (ablation) | XLM (MLM) with MLT_A · Multilingual NLU Spanish | intent accuracy: 83.95 -> 85.63 | Helped |
| Add a transformer encoder to map subword outputs to word-level representations (ablation) | Multi. BERT with MLT_A · Multilingual NLU Thai | slot F1: 28.47 -> 27.12 | Hurt |
| Apply MLT rather than direct zero-shot transfer | MUSE on Multilingual WOZ 2.0 German · Multilingual WOZ 2.0 German | joint goal accuracy: 21.57 -> 36.51 | Helped |
| Apply MLT rather than direct zero-shot transfer | RCSLS on Multilingual NLU Spanish · Multilingual NLU Spanish | intent accuracy: 37.67 -> 87.05 | Helped |
| Apply MLT rather than direct zero-shot transfer | RCSLS on Multilingual NLU Thai · Multilingual NLU Thai | slot F1: 8.72 -> 30.42 | Helped |
| Use attention-based word selection rather than human word selection with five word pairs (ablation) | MLT on Spanish NLU · Spanish test data | intent prediction: +17.69% | Helped |
| Use attention-based word selection rather than human word selection with five word pairs (ablation) | MLT on Spanish NLU · Spanish test data | slot filling: +21.45% | Helped |
Closest papers by meaning
- Unsupervised Interlingual Semantic Representations from Sentence Embeddings for Zero-Shot Cross-Lingual Transfer (AAAI 2020 · similarity 0.89)
- BiToD: A Bilingual Multi-Domain Dataset For Task-Oriented Dialogue Modeling (NeurIPS 2021 · similarity 0.88)
- Cross-lingual Alignment vs Joint Training: A Comparative Study and A Simple Unified Framework (ICLR 2020 · similarity 0.87)
- Dialogue for Prompting: A Policy-Gradient-Based Discrete Prompt Generation for Few-Shot Learning (AAAI 2024 · similarity 0.86)
- MCIF: Multimodal Crosslingual Instruction-Following Benchmark from Scientific Talks (ICLR 2026 · similarity 0.86)
- Zero-Shot Continuous Prompt Transfer: Generalizing Task Semantics Across Language Models (ICLR 2024 · similarity 0.85)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.