# It was the training data pruning too!

## TL;DR

This paper reveals that data pruning during training significantly impacts the performance of the KDG question answering model on tabular data, highlighting the importance of data preprocessing in semantic parsing.

## Contribution

It demonstrates that data pruning is a crucial factor in model performance, challenging previous attributions solely to model architecture.

## Key findings

- Disabling pruning reduces accuracy from 43.3% to 36.3%.
- Pruning is a potentially valuable pre-processing step for semantic parsers.
- Performance is highly sensitive to training data preprocessing.

## Abstract

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.