Sentence examples for medical relation from inspiring English sources

This study proposes a convolutional neural network (CNN) architecture with a multi-pooling operation for medical relation classification on clinical records and explores a loss function with a category-level constraint matrix.

Rule1 says that if TREATMENT event1 and PROBLEM event2 have a semantic medical relation TrAP, there is a prepositional-for dependency from event1 to event2, and where both events are in numbered argument 2 related to predicate started.

In 2010, i2b2 together with VA Salt Lake City Health Care System organized a community-wide shared task (19) that had as one of its subtasks to automatically classify TREATMENT events with PROBLEM events, TEST events with PROBLEM events and event pairs each of type PROBLEM into one of a set of predefined semantic medical relations.

This translates to a relative error reduction of 17 24% in comparison to the baseline result shown in row 1. Regarding the usefulness of each of the seven types of features in this best-performing architecture, we found that adding pairwise features, predicate argument relations, discourse relations and medical semantic relations significantly improves both micro and macro F-scores.

Given the potential usefulness of medical semantic relations for temporal relation classification, as demonstrated earlier through Examples (8)–(12), we create features for temporal relation classification based on these semantic relations.

Given they are not related to each other (via any medical semantic relation), it is unlikely that they occur simultaneously.

First, consider Sentence (9), which shows that the absence of a medical semantic relation between two events could also be useful for temporal relation classification.

As in medical semantic relation classification, we employ a simple extraction tree as a structured feature to represent an instance for temporal relation identification.

Table 6 shows the medical semantic relation classification results.

In this subsection, we describe our corpus-based approach to medical semantic relation classification.

The remaining features come from the 36 feature groups that we used for medical semantic relation classification.