Research map / Label-efficient and robust learning / Recognition and long-tail
Unsupervised Transfer Learning for Spoken Language Understanding in Intelligent Agents
Aditya Siddhant, Anuj Goyal, Angeliki Metallinou
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (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 |
|---|---|---|---|
| Replace Fasttext embeddings with ELMo unsupervised transfer | SLU multi-task IC/ET model · Domain-A and Domain-B | SER: 38.76 -> 36.58; 29.30 -> 27.95 | Improved |
| Replace Fasttext embeddings with ELMo unsupervised transfer | SLU multi-task IC/ET model · ATIS and SNIPS | SER: 15.01 -> 12.65; 19.57 -> 14.57 | Improved |
| Use ELMoL instead of ELMo | SLU multi-task IC/ET model · Domain-A, Domain-B and ATIS | SER: 36.58 -> 36.86; 27.95 -> 28.42; 12.65 -> 13.38 | No clear effect |
| Use ELMoL instead of ELMo | SLU multi-task IC/ET model · SNIPS | SER: 14.57 -> 15.62 | Hurt |
| Add supervised transfer to ELMo | ELMo SLU model · Domain-A and Domain-B | SER: 36.58 -> 35.86; 27.95 -> 27.46 | Improved |
| Add supervised transfer to ELMoL | ELMoL SLU model · Domain-A and Domain-B | SER: 36.86 -> 36.38; 28.42 -> 28.11 | Improved |
| Add gradual unfreezing to ELMoL training (ablation) | ELMoL · ATIS, SNIPS, Domain-A and Domain-B | SER: 14.46 -> 13.79; 17.14 -> 16.39; 37.59 -> 37.26; 28.78 -> 28.72 | Improved |
| Add discriminative fine-tuning and slanted triangular learning rates after gradual unfreezing (ablation) | ELMoL · ATIS, SNIPS, Domain-A and Domain-B | SER: 13.79 -> 13.38; 16.39 -> 15.62; 37.26 -> 36.86; 28.72 -> 28.42 | Improved |
| Use ELMo embeddings rather than NoUT training | SLU multi-task IC/ET model · Domain-A, Domain-B, ATIS and SNIPS | SER: 39.79 -> 36.58; 30.93 -> 27.95; 17.13 -> 12.65; 24.14 -> 14.57 | Improved |
Closest papers by meaning
- Self-Taught Recognizer: Toward Unsupervised Adaptation for Speech Foundation Models (NeurIPS 2024 · similarity 0.85)
- UniWav: Towards Unified Pre-training for Speech Representation Learning and Generation (ICLR 2025 · similarity 0.84)
- MATS: An Audio Language Model under Text-only Supervision (ICML 2025 · similarity 0.84)
- MMSU: A Massive Multi-task Spoken Language Understanding and Reasoning Benchmark (ICLR 2026 · similarity 0.84)
- Generative Adapter: Contextualizing Language Models in Parameters with A Single Forward Pass (ICLR 2025 · similarity 0.84)
- On Effects of Steering Latent Representation for Large Language Model Unlearning (AAAI 2025 · similarity 0.84)
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.
