Research map / Vision-language and multimodal / Cross-modal learning
Cross-Lingual Propagation for Deep Sentiment Analysis
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 |
|---|---|---|---|
| concatenate Amazon sentiment embeddings to word vectors (ablation) | CNN with G/F · 9 languages | accuracy: Spanish 85.13 -> 84.56; English MR 85.99 -> 86.55 | Mixed |
| use separate sentiment channel (dual-channel) for Amazon embeddings instead of concatenation (ablation) | G/F + A · Japanese RR | accuracy: 75.91 -> 88.08 | Helped |
| dual-channel with Amazon embeddings vs plain G/F CNN | G/F CNN · English MR | accuracy: 85.99 -> 86.60 | Helped |
| dual-channel with Amazon embeddings vs plain G/F CNN | G/F CNN · Spanish RR | accuracy: 85.13 -> 85.93 | Helped |
| add Amazon transfer-learning sentiment embeddings instead of Polyglot lexicon (ablation) | DC-CNN · Japanese RR | accuracy: 76.64 -> 88.08 | Helped |
| freeze Amazon sentiment embeddings (static) instead of fine-tuning (ablation) | DC-CNN G||A · Japanese RR | accuracy: 88.08 -> 77.62 | Hurt |
| freeze Amazon sentiment embeddings (static) (ablation) | DC-CNN G||A · English MR | accuracy: 86.60 -> 86.82 | No clear effect |
| add Amazon sentiment channel to English model for cross-domain test | GloVe CNN · MR to FR cross-domain | accuracy: 69.24% -> 74.85% | Helped |
| use pretrained G/F vectors instead of random init | CNN · English MR | accuracy: 80.78 -> 85.99 | Helped |
| add sentiment embeddings when training data is small (ablation) | DC-CNN · English and Czech | accuracy | Helped |
| dual-channel with Amazon embeddings vs plain G/F CNN | G/F CNN · Dutch RR | accuracy: 77.57 -> 79.30 | Helped |
Closest papers by meaning
- Structural Correspondence Learning for Cross-Lingual Sentiment Classification with One-to-Many Mappings (AAAI 2017 · similarity 0.86)
- MNCN: A Multilingual Ngram-Based Convolutional Network for Aspect Category Detection in Online Reviews (AAAI 2019 · similarity 0.85)
- Simultaneous Learning of Pivots and Representations for Cross-Domain Sentiment Classification (AAAI 2020 · similarity 0.84)
- Massively Multilingual Corpus of Sentiment Datasets and Multi-faceted Sentiment Classification Benchmark (NeurIPS 2023 · similarity 0.84)
- DLF: Disentangled-Language-Focused Multimodal Sentiment Analysis (AAAI 2025 · similarity 0.82)
- Transformers Get Stable: An End-to-End Signal Propagation Theory for Language Models (ICML 2024 · similarity 0.81)
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.