It was the training data pruning too!
This finding highlights a crucial but overlooked factor in model performance for semantic parsing on tabular data, though it is incremental as it builds on existing work.
The paper reveals that the performance of the KDG model for question answering on tabular data depends significantly on a training data pruning step, with accuracy dropping from 43.3% to 36.3% when pruning is disabled.
We study the current best model (KDG) for question answering on tabular data evaluated over the WikiTableQuestions dataset. Previous ablation studies performed against this model attributed the model's performance to certain aspects of its architecture. In this paper, we find that the model's performance also crucially depends on a certain pruning of the data used to train the model. Disabling the pruning step drops the accuracy of the model from 43.3% to 36.3%. The large impact on the performance of the KDG model suggests that the pruning may be a useful pre-processing step in training other semantic parsers as well.