Research map / Label-efficient and robust learning / Recognition and long-tail
Multi-view Recurrent Neural Acoustic Word Embeddings
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 |
|---|---|---|---|
| symmetrized objective obj0 + obj2 instead of obj0 alone (ablation) | multi-view LSTM · acoustic word discrimination dev | AP: 0.659 -> 0.702 | Helped |
| obj0 + obj2 instead of obj2 alone (ablation) | multi-view LSTM · acoustic word discrimination dev | AP: 0.675 -> 0.702 | Helped |
| obj0 + obj2 instead of obj0 alone (ablation) | multi-view LSTM · cross-view word discrimination dev | AP: 0.791 -> 0.814 | Helped |
| sum of all four objectives instead of obj0 + obj2 (ablation) | multi-view LSTM · acoustic word discrimination dev | AP: 0.702 -> 0.672 | Hurt |
| cost-sensitive margin objective instead of obj0 + obj2 (ablation) | multi-view LSTM · acoustic word discrimination dev | AP: 0.702 -> 0.671 | Hurt |
| cost-sensitive margin instead of fixed margin (ablation) | obj0 · word similarity acoustic | Spearman rho: 0.179 -> 0.240 | Helped |
| cost-sensitive margin instead of fixed margin (ablation) | obj0 · word similarity text | Spearman rho: 0.207 -> 0.270 | Helped |
| cost-sensitive margin instead of fixed margin (ablation) | obj0 · cross-view discrimination dev | AP: 0.791 -> 0.802 | No clear effect |
| bidirectional instead of unidirectional 1-layer LSTM (ablation) | obj0 embedding · acoustic word discrimination dev | AP: 0.379 -> 0.466 | Helped |
| 2-layer instead of 1-layer bidirectional LSTM (ablation) | obj0 embedding · acoustic word discrimination dev | AP: 0.466 -> 0.659 | Helped |
| 2-layer instead of 1-layer bidirectional LSTM (ablation) | obj0 embedding · cross-view word discrimination dev | AP: 0.690 -> 0.791 | Helped |
| more than two LSTM layers (ablation) | obj0 embedding · word discrimination dev | AP | No clear effect |
Closest papers by meaning
- Training and Evaluating Multimodal Word Embeddings with Large-scale Web Annotated Images (NeurIPS 2016 · similarity 0.87)
- All Word Embeddings from One Embedding (NeurIPS 2020 · similarity 0.87)
- Improving Word Embeddings with Convolutional Feature Learning and Subword Information (AAAI 2017 · similarity 0.87)
- Slim Embedding Layers for Recurrent Neural Language Models (AAAI 2018 · similarity 0.86)
- Word-Level Speech Recognition With a Letter to Word Encoder (ICML 2020 · similarity 0.86)
- Unified Generative and Discriminative Training for Multi-modal Large Language Models (NeurIPS 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.