AI Paper Atlas

Research map / Vision-language and multimodal / Human and embodied modeling

Visualizing and Measuring the Geometry of BERT

Emily Reif, Ann Yuan, Martin Wattenberg, Fernanda B Viegas, Andy Coenen, Adam Pearce, Been Kim

NeurIPS 2019 · Vision-language and multimodal › Human and embodied modeling · approach: Scene graph generation · 179 citations (OpenAlex)

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 changedWhereMeasuredEffect
Apply the trained semantic probe to BERT embeddingsBERT-base nearest-centroid WSD · WSD test suiteF1 score: 71.1 -> 71.5Improved
Use BERT context embeddings for nearest-centroid WSD instead of ELMoWSD · WSD test suiteF1 score: 70.1 -> 71.1Improved
Use model-wide attention vectors as input to a linear dependency-existence classifierBERT-base attention · Penn Treebank dependency existenceaccuracy: 85.8%Improved
Use model-wide attention vectors as input to a linear dependency-type classifierBERT-base attention · Penn Treebank dependency relation typeaccuracy: 71.9%Improved
Reduce trained semantic-probe dimensionality from 768 to 512 (ablation)BERT-base semantic probe · WSD test suiteaccuracy: 71.26 -> 71.52Improved
Reduce trained semantic-probe dimensionality to 256 or 128 (ablation)BERT-base semantic probe · WSD test suiteaccuracy: 71.26; 71.29; 71.21No clear effect
Reduce trained semantic-probe dimensionality to 64, 32, or 16 (ablation)BERT-base semantic probe · WSD test suiteaccuracy: 70.19; 68.01; 65.34Hurt
Train the semantic probe rather than use a random probe of the same dimensionality (ablation)BERT-base semantic probe · WSD test suiteaccuracy: 71.26 vs 70.74; 71.52 vs 70.51; 71.29 vs 69.92; 71.21 vs 69.56; 70.19 vs 68.00; 68.01 vs 64.62; 65.34 vs 61.01Improved
Concatenate a sentence with a sentence using the same keyword in another senseBERT-base context embeddings · SemCor sentence pairsmisclassification rate: 2.43% -> 8.23%Hurt
Project keyword embeddings into the trained semantic subspace (ablation)BERT-base context embeddings · SemCor sentence pairssimilarity ratio: 1.28 -> 1.58 (concatenated); 1.43 -> 1.88 (individual)Improved
Project keyword embeddings into the trained semantic subspace (ablation)BERT-base context embeddings · SemCor sentence pairsmisclassification rate: 7.31% vs 8.23% (concatenated); 2.27% vs 2.43% (individual)Improved
Concatenate a random sentence instead of a different-sense sentence (ablation)BERT-base context embeddings · SemCor sentence pairssimilarity ratioNo clear effect

Closest papers by meaning

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy