Research map / Label-efficient and robust learning / Recognition and long-tail
Disambiguating Spatial Prepositions Using Deep Convolutional Networks
Kaveh Hassani, Won-Sook Lee
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 pre-trained universal embeddings instead of corpus-trained embeddings (ablation) | all models · PDEP spatial prepositions | accuracy and F1: 4.85% accuracy, 5.39% F1 | Improved |
| use hybrid features instead of other feature sets (ablation) | all classifiers · PDEP spatial prepositions | accuracy and F1 | Improved |
| combine linguistic features with universal embeddings (ablation) | classifiers · PDEP spatial prepositions | accuracy and F1 | Improved |
| increase context window size beyond 5 (ablation) | DNN · PDEP spatial prepositions | F1-score | Hurt |
| use window size 3 instead of 5 (ablation) | DNN · PDEP spatial prepositions | F1-score | Hurt |
| reduce word vector dimensionality from 300 with PCA to about 100 (ablation) | DNN · PDEP spatial prepositions | F1-score | Improved |
| use GloVe instead of Word2Vec pre-trained embeddings with classic classifiers (ablation) | classic classifiers · PDEP spatial prepositions | accuracy | Improved |
| use Word2Vec instead of GloVe embeddings with DNN (ablation) | DNN · PDEP spatial prepositions | F1-score | Improved |
| use deep learning models instead of classic classifiers (ablation) | classifiers · PDEP spatial prepositions | accuracy and F1 | Improved |
| use ensemble classifiers instead of other classic classifiers (ablation) | classic classifiers · PDEP spatial prepositions | accuracy and F1 | Improved |
| hybrid features with CNN versus Kordjamshidi et al. 2011 | Hybrid + CNN · PDEP spatial prepositions | accuracy: 94.21% accuracy, +7.06% | Improved |
Closest papers by meaning
- Acquiring Common Sense Spatial Knowledge Through Implicit Spatial Templates (AAAI 2018 · similarity 0.85)
- Multi-Scale Representation Learning for Spatial Feature Distributions using Grid Cells (ICLR 2020 · similarity 0.84)
- Multi-Level Multimodal Common Semantic Space for Image-Phrase Grounding (CVPR 2019 · similarity 0.83)
- Syntax-Aware Neural Semantic Role Labeling (AAAI 2019 · similarity 0.83)
- SIRI: Spatial Relation Induced Network For Spatial Description Resolution (NeurIPS 2020 · similarity 0.83)
- Predicting the Argumenthood of English Prepositional Phrases (AAAI 2019 · similarity 0.83)
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.
