On the Multi-Property Extraction and Beyond
This work addresses information extraction and machine reading comprehension challenges for researchers and practitioners in natural language processing, though it appears incremental with a focus on dataset refinement and model adaptation.
The paper tackles the problem of multiple property extraction from text by proposing a Dual-source Transformer architecture that achieves state-of-the-art performance on the WikiReading dataset, with a significant performance improvement over previous methods. It also introduces WikiReading Recycled, a new public dataset designed to overcome limitations of the original while supporting multi-property extraction tasks.
In this paper, we investigate the Dual-source Transformer architecture on the WikiReading information extraction and machine reading comprehension dataset. The proposed model outperforms the current state-of-the-art by a large margin. Next, we introduce WikiReading Recycled - a newly developed public dataset, supporting the task of multiple property extraction. It keeps the spirit of the original WikiReading but does not inherit the identified disadvantages of its predecessor.