TACRED is one of the largest and most widely used sentence-level relation extraction datasets. Proposed models that are evaluated using this dataset consistently set new state-of-the-art performance. However, they still exhibit large error rates despite leveraging external knowledge unsupervised pretraining on text corpora. A recent study suggested may be due to poor quality. The observed over ...

Mathematics is highly structured and underpins most of science and engineering. For this reason it has proved a very suitable test domain for intelligent tutoring system (ITS) research with the result that probably more tutoring systems have been structured for the domain than for any othw. However, there still exists no consensus on any approach for the design of such systems. Consequently, ex...