Neural Network Models for Natural Language Inference Fail to Capture the Semantics of Inference
share
Neural network models have been very successful for natural language inference, with the best models reaching 90 the success of these models turns out to be largely task specific. We show that models trained on one inference task fail to perform well in others, even if the notion of inference assumed in these tasks is the same or similar. We train four state-of-the-art neural network models on different datasets and show that each one of these fail to generalize outside of the respective task. In light of these results we conclude that the current neural network models are not able to generalize in capturing the semantics of natural language inference, but seem to be overfitting to the specific dataset.
READ FULL TEXT