Research map / Label-efficient and robust learning / Recognition and long-tail
Emu: Enhancing Multilingual Sentence Embeddings with Semantic Specialization
Wataru Hirota, Yoshihiko Suhara, Behzad Golshan, Wang-Chiew Tan
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 |
|---|---|---|---|
| Fine-tune LASER with semantic classification using L2-constrained softmax and center loss, plus language-adversarial training | LASER · Cross-lingual intent classification, TrustYou, ATIS, and Quora | Acc@1 | Improved |
| Add a language discriminator to Emu (ablation) | Emu · TrustYou | Acc@1: 1.45 percentage points | Improved |
| Add a language discriminator to Emu (ablation) | Emu · ATIS | Acc@1: 2.81 percentage points | Improved |
| Add a language discriminator to Emu (ablation) | Emu · Quora | Acc@1: 1.05 percentage points | Improved |
| Add center loss (ablation) | Emu without center loss · TrustYou | Acc@1: -0.44 percentage points | Hurt |
| Add center loss (ablation) | Emu without center loss · ATIS | Acc@1: 0.04 percentage points | No clear effect |
| Add center loss (ablation) | Emu without center loss · Quora | Acc@1: 6.00 percentage points | Improved |
| Replace randomly sampled non-parallel adversarial sentences with parallel sentences (ablation) | Emu · TrustYou, ATIS, and Quora | Acc@1: -0.5 points, -1.2 points, and -5.9 points | Hurt |
| Fine-tune with common contrastive, N-pair, or softmax losses instead of the proposed losses | LASER · Cross-lingual intent classification across the three datasets | Acc@1 | Hurt |
| Use the semantic classifier directly for intent detection instead of nearest-neighbor similarity | Emu · Cross-lingual intent classification | classification performance | Hurt |
Closest papers by meaning
- A Simple but Tough-to-Beat Baseline for Sentence Embeddings (ICLR 2017 · similarity 0.89)
- Unsupervised Interlingual Semantic Representations from Sentence Embeddings for Zero-Shot Cross-Lingual Transfer (AAAI 2020 · similarity 0.89)
- A Comparison of Architectures and Pretraining Methods for Contextualized Multilingual Word Embeddings (AAAI 2020 · similarity 0.88)
- Narrowing the Gap between Supervised and Unsupervised Sentence Representation Learning with Large Language Model (AAAI 2024 · similarity 0.86)
- MMTEB: Massive Multilingual Text Embedding Benchmark (ICLR 2025 · similarity 0.86)
- The Scandinavian Embedding Benchmarks: Comprehensive Assessment of Multilingual and Monolingual Text Embedding (NeurIPS 2024 · similarity 0.86)
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.
