Research map / Large language models / Language, safety and interpretability
Predicting the Argumenthood of English Prepositional Phrases
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 |
|---|---|---|---|
| Encode the verb-preposition pair with a BiLSTM rather than bag-of-words (ablation) | Binary PP argumenthood classifier · VerbNet binary PP argumenthood test | accuracy: 93.7 -> 95.5 | Helped |
| Preserve input order with concatenated vectors rather than bag-of-words (ablation) | Binary PP argumenthood classifier · VerbNet binary PP argumenthood test | accuracy: 93.7 -> 94.4 | Helped |
| Use ELMo embeddings rather than other tested embeddings | BiLSTM + MLP · VerbNet binary PP argumenthood test | accuracy: 94.6 -> 95.5 | Helped |
| Pretrain the sentence encoder on PP argumenthood | Sentence encoder for SRL · SRL-CoNLL2005 (WSJ) | F1: 81.7 -> 83.9 | Helped |
| Pretrain the sentence encoder on PP argumenthood | Sentence encoder for SRL · SRL-CoNLL2012 (OntoNotes) | F1: 77.3 -> 80.2 | Helped |
| Pretrain the sentence encoder on PP argumenthood | Sentence encoder for PP attachment · PP attachment | accuracy: 87.5 -> 88.2 | No clear effect |
| Add mutual information, direct-object, and diagnostic features to word embeddings (ablation) | Gradient argumenthood predictor (word2vec-wiki) · Gradient PP argumenthood | Pearson's r: 0.430 -> 0.624 | Helped |
| Add mutual information to word embeddings (ablation) | Gradient argumenthood predictor (word2vec-wiki) · Gradient PP argumenthood | Pearson's r: 0.430 -> 0.464 | Helped |
| Add a direct-object feature to word embeddings (ablation) | Gradient argumenthood predictor (word2vec-wiki) · Gradient PP argumenthood | Pearson's r: 0.430 -> 0.575 | Helped |
| Add traditional diagnostic scores to word embeddings (ablation) | Gradient argumenthood predictor (word2vec-wiki) · Gradient PP argumenthood | Pearson's r: 0.430 -> 0.449 | Helped |
| Use linguistic features without embeddings (ablation) | Gradient argumenthood predictor · Gradient PP argumenthood | Pearson's r: 0.624 -> 0.516 | Hurt |
| Increase the source embedding dimensionality before PCA reduction | Gradient argumenthood predictor · Gradient PP argumenthood | Pearson's r: 0.609 -> 0.624 | Helped |
Closest papers by meaning
- Span-Based Semantic Role Labeling with Argument Pruning and Second-Order Inference (AAAI 2022 · similarity 0.86)
- Dependency or Span, End-to-End Uniform Semantic Role Labeling (AAAI 2019 · similarity 0.85)
- Frame Semantic Role Labeling Using Arbitrary-Order Conditional Random Fields (AAAI 2024 · similarity 0.83)
- Underspecification in Language Modeling Tasks: A Causality-Informed Study of Gendered Pronoun Resolution (AAAI 2024 · similarity 0.83)
- Evaluating the Zero-shot Robustness of Instruction-tuned Language Models (ICLR 2024 · similarity 0.82)
- Learning the Wrong Lessons: Syntactic-Domain Spurious Correlations in Language Models (NeurIPS 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.