Research map / Label-efficient and robust learning / Recognition and long-tail
Beyond Word Importance: Contextual Decomposition to Extract Interactions from LSTMs
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 contextual decomposition instead of word-level attribution baselines | LSTM sentiment interpretation · Stanford Sentiment Treebank | Pearson correlation with logistic regression coefficients: 0.724 -> 0.758 | Helped |
| Use contextual decomposition instead of cell decomposition | LSTM sentiment interpretation · Yelp Polarity | Pearson correlation with logistic regression coefficients: 0.560 -> 0.520 | Hurt |
| Use contextual decomposition to score dissenting subphrases | LSTM sentiment interpretation · Yelp Polarity | separation of positive and negative dissenting-subphrase score distributions | Helped |
| Use contextual decomposition to score long phrases of opposing sentiment | LSTM sentiment interpretation · Stanford Sentiment Treebank | one-sided Kolmogorov-Smirnov statistic: 0.61 -> 0.74 | Helped |
| Use integrated gradients to score positive phrases in mixed-sentiment reviews | LSTM sentiment interpretation · Stanford Sentiment Treebank | positive phrases labeled negative: 87% | Hurt |
| Extract phrase interactions with contextual decomposition rather than leave-one-out | LSTM sentiment interpretation · Stanford Sentiment Treebank negation phrases | distinction between positive and negative negation interactions | Helped |
| Use contextual decomposition to assign phrase-level sentiment within mixed-polarity reviews | LSTM sentiment interpretation · Yelp Polarity | identification of contrasting subphrase sentiment | Helped |
Closest papers by meaning
- How recurrent networks implement contextual processing in sentiment analysis (ICML 2020 · similarity 0.86)
- CEBaB: Estimating the Causal Effects of Real-World Concepts on NLP Model Behavior (NeurIPS 2022 · similarity 0.85)
- Paloma: A Benchmark for Evaluating Language Model Fit (NeurIPS 2024 · similarity 0.83)
- Sparsity-Guided Holistic Explanation for LLMs with Interpretable Inference-Time Intervention (AAAI 2024 · similarity 0.83)
- Language Ranker: A Lightweight Ranking framework for LLM Decoding (NeurIPS 2025 · similarity 0.82)
- DLF: Disentangled-Language-Focused Multimodal Sentiment Analysis (AAAI 2025 · similarity 0.82)
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.