techwitz/teer

A tool to detect temporal expression from free text

TEXer

A tool to detect temporal expression from free text

About

Automatic translation of clinical researcher data requests to executable database queries is instrumental to an effective interface between clinical researchers and “Big Clinical Data”. A necessary step towards this goal is to parse ample temporal expressions in free-text researcher requests. This paper reports a novel algorithm called TEXer. It uses heuristic rule and pattern learning for extracting and normalizing temporal expressions in researcher requests. Based on 400 real clinical queries with human annotations, we compared our method with four baseline methods. TEXer achieved a precision of 0.945 and a recall of 0.858, outperforming all the baseline methods. We conclude that TEXer is an effective method for temporal expression extraction from free-text clinical data requests.

Usage

TEXer_English: the TEXer tool for English free medical text

TEXer_Chinese: the TEXer tool for English free medical text

Data usage

training: some training instances

training_gold: the gold standard (human annotation) of the training data

training_pattern: the trained patterns

testing: some testing instances

testing_gold: the gold standard (human annotation) of the testing data

Citation

Tianyong Hao, Alex Rusanov, Chunhua Weng. Extracting and Normalizing Temporal Expressions in Clinical Data Requests from Researchers. Lecture Notes in Computer Science, Volume 8040, pp 41-51, Springer, 2013.

Contributors

Tianyong Hao

Zhiying Gu (Improve English version and develop Chinese version for EMR text)

Xiaoyi Pan (Improve Chinese version)