AI Paper Atlas

Research map / Label-efficient and robust learning / Recognition and long-tail

All-but-the-Top: Simple and Effective Postprocessing for Word Representations

Jiaqi Mu, Pramod Viswanath

ICLR 2018 · Label-efficient and robust learning › Recognition and long-tail · approach: Multilingual language representation pretraining · 129 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
Center embeddings and remove the top principal componentsWORD2VEC and GLOVE · Word similarity, seven datasetsSpearman correlation: average improvement of 2.3%Improved
Center embeddings and remove the top principal componentsWORD2VEC and GLOVE · Concept categorization, three datasetspurity: average improvement of 2.5%Improved
Center embeddings and remove the top principal componentsWORD2VEC · Word analogy, allaccuracy: 72.93 -> 73.44Improved
Center embeddings and remove the top principal componentsGLOVE · Semantic textual similarity, allPearson correlation: 49.19 -> 56.76Improved
Postprocess embeddings before classificationFour neural-network architectures with WORD2VEC or GLOVE · Text classification, five datasetsclassification performance: 34 out of 40; 2.32% on averageMixed
Center vectors and remove top principal componentsTSCCA · Word similarity, German and Spanish datasetsSpearman correlation: 60.33 -> 60.37; 61.75 -> 64.39; 44.91 -> 46.59; 30.37 -> 32.92Improved
Set D according to the d/100 ruleSkip-gram embeddings across dimensions 300 to 1000 · Word similaritySpearman correlation: 2.27%Improved
Set D according to the d/100 ruleSkip-gram embeddings across dimensions 300 to 1000 · Word analogyaccuracy: 0.01No clear effect
Set D according to the d/100 ruleSkip-gram embeddings across dimensions 300 to 1000 · Concept categorizationpurity: 3.37%Mixed
Set D according to the d/100 ruleSkip-gram embeddings across dimensions 300 to 1000 · Semantic textual similarityPearson correlation: 1.92%Improved
Remove the top principal componentsRAND-WALK and TSCCA · Word similaritySpearman correlation: 2.23% average improvement for RAND-WALK and TSCCAMixed
Remove the top principal componentsRAND-WALK and TSCCA · Semantic textual similarityPearson correlation: 0.61% average improvement for RAND-WALK and TSCCAMixed

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