A Decomposable Attention Model for Natural Language Inference

EMNLP 2016 Ankur P. ParikhOscar TäckströmDipanjan DasJakob Uszkoreit

We propose a simple neural architecture for natural language inference. Our approach uses attention to decompose the problem into subproblems that can be solved separately, thus making it trivially parallelizable... (read more)

PDF Abstract

Evaluation results from the paper


Task Dataset Model Metric name Metric value Global rank Compare
Natural Language Inference SNLI 200D decomposable attention model % Test Accuracy 86.3 # 26
Natural Language Inference SNLI 200D decomposable attention model % Train Accuracy 89.5 # 36
Natural Language Inference SNLI 200D decomposable attention model Parameters 380k # 1
Natural Language Inference SNLI 200D decomposable attention model with intra-sentence attention % Test Accuracy 86.8 # 21
Natural Language Inference SNLI 200D decomposable attention model with intra-sentence attention % Train Accuracy 90.5 # 32
Natural Language Inference SNLI 200D decomposable attention model with intra-sentence attention Parameters 580k # 1