TORQUE: A Reading Comprehension Dataset of Temporal Ordering Questions
This addresses a gap in reading comprehension benchmarks for AI systems, enabling better temporal understanding, though it is incremental as it focuses on dataset creation.
The authors tackled the lack of temporal reasoning in machine reading comprehension by introducing TORQUE, a dataset with 21k questions on temporal ordering, and found that RoBERTa-large achieved 51% exact-match accuracy, 30% below human performance.
A critical part of reading is being able to understand the temporal relationships between events described in a passage of text, even when those relationships are not explicitly stated. However, current machine reading comprehension benchmarks have practically no questions that test temporal phenomena, so systems trained on these benchmarks have no capacity to answer questions such as "what happened before/after [some event]?" We introduce TORQUE, a new English reading comprehension benchmark built on 3.2k news snippets with 21k human-generated questions querying temporal relationships. Results show that RoBERTa-large achieves an exact-match score of 51% on the test set of TORQUE, about 30% behind human performance.