Research map / Label-efficient and robust learning / Recognition and long-tail
Have We Solved The Hard Problem? It’s Not Easy! Contextual Lexical Contrast as a Means to Probe Neural Coherence
Wenqiang Lei, Yisong Miao, Runpeng Xie, Bonnie Webber, Meichun Liu, Tat-Seng Chua, Nancy F. Chen
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 |
|---|---|---|---|
| Use contextual BERT embeddings instead of FastText embeddings | CLC recognition model · Cont2Lex | accuracy: 66.3 -> 70.0 | Improved |
| Use contextual embeddings rather than static embeddings | Out-of-context lexical contrast recognition · Out-of-context lexical contrast | accuracy: 84.1 -> 79.5 | Hurt |
| Add a BiLSTM encoder to contextual BERT embeddings | CLC recognition model · Cont2Lex | accuracy: 70.0 -> 69.2; 70.0 -> 69.1 | No clear effect |
| Use BERT rather than BERT.Lex, which lacks surrounding-context input | CLC recognition model · Cont2Lex | accuracy: 66.4 -> 70.0 | Improved |
| Use BERT rather than FastText | CLC recognition model · Same-sentence pairs | accuracy: 60.4 -> 67.4 | Improved |
| Use BERT rather than FastText | CLC recognition model · Adjacent-sentence pairs | accuracy: 69.8 -> 71.4 | Improved |
| Use full BERT instead of BERT.Lex | CLC recognition model · Same-sentence pairs | accuracy: 60.7 -> 67.4 | Improved |
| Use full BERT instead of BERT.Lex | CLC recognition model · Adjacent-sentence pairs | accuracy: 69.8 -> 71.4 | Improved |
| Use BERT rather than FastText | CLC recognition model · Instances with word repetition | accuracy: 61.1 -> 68.7 | Improved |
| Use BERT rather than FastText | CLC recognition model · Instances without word repetition | accuracy: 68.8 -> 70.7 | Improved |
| Use full BERT instead of BERT.Lex | CLC recognition model · Instances with word repetition | accuracy: 58.7 -> 68.7 | Improved |
| Use full BERT instead of BERT.Lex | CLC recognition model · Instances without word repetition | accuracy: 69.9 -> 70.7 | Improved |
Closest papers by meaning
- Joint Word Representation Learning Using a Corpus and a Semantic Lexicon (AAAI 2016 · similarity 0.87)
- BBScore: A Brownian Bridge Based Metric for Assessing Text Coherence (AAAI 2024 · similarity 0.86)
- Joint Semantic Analysis with Document-Level Cross-Task Coherence Rewards (AAAI 2021 · similarity 0.86)
- Deep contextualized word representations (ICLR 2018 · similarity 0.86)
- What do you learn from context? Probing for sentence structure in contextualized word representations (ICLR 2019 · similarity 0.86)
- Less is More: Local Intrinsic Dimensions of Contextual Language Models (NeurIPS 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.
