Automatic systems that identify temporal relations from clinical narratives suffer from the complexity of the task, which mixes both explicit and implicit temporal relations in a single set despite the fact that the types of evidence that should be used for the identification of explicit and implicit temporal relations are different.
We propose to enhance the performance of the state-of-the-art clinical temporal relation identification systems by focusing on explicit temporal relations. We construct a corpus of explicit temporal relations, or direct temporal relations.
ANNOTATION GUIDELINE:
Download the annotation guideline (PDF) here at https://sbmi.uth.edu/ccb/tlink/.
PUBLICATION:
+ Hee-Jin Lee, Yaoyun Zhang, Min Jiang, Jun Xu, Cui Tao, Hua Xu, "Towards practical temporal relation extraction from clinical notes: an analysis of direct temporal relations", the 2nd International Workshop on Semantics-Powered Data Analytics (SEPDA 2017) in conjunction with 2017 IEEE International Conference on Bioinformatics and Biomedicine (BIBM 2017).