The code is being organized and will be provided later...
In light of the differences between value mentions in questions and cells in databases brought by the diversity and ambiguity of user questions in real-world scenarios, targeting the mismatch of conditions problem mentioned in our paper, we construct the dataset from almost all the samples with string-type conditional clauses in Spider, Spider-SYN and Spider-DK. We apply the following disturbances to these samples:
| Disturbance | Example |
|---|---|
| Original question | When did the episode “A Love of a Lifetime” air? |
| Remove column | When did “A Love of a Lifetime” air? |
| Remove highlight | When did a love of lifetime air? |
| Replace common value | When did double down air? |