CLAMP Version 1.5.0
New Features
• CLAMP GUI 1.5.0 version support pre-annotation
The model in corpus project gets updated automatically during your annotation and provides pre-annotated results on new documents for faster annotation.
• CLAMP GUI 1.5.0 version support new flexible evaluation function
Users can select any two datasets (e.g., gold standard vs. predicted results) to evaluate system’s performance.
Improvements
• Java is embedded into the package so that you don’t have to install/modify your own Java environment.
• New editor to modify the ‘typedef.xml’;.
Bug fixing
• UI freeze on some Mac OS versions has been fixed now.
CLAMP Version 1.4.0
New Features
• CLAMP GUI 1.4.0 version support reading/writing database via JDBC driver
Currently tested on MySQL, PostgreSQL, and Oracle. The setup screenshot as below for the reference.
Database Configration
• Improved concept mapping function
Besides UMLS CUIs, you can now map different types of concepts to selected terminologies (ICD10, SNOMED-CT, LOINC, RxNorm) .
Mapping Function
• Pre-annotated machine learning based test-value extraction component
Now CLAMP can Map to selected terminologies (ICD10, SNOMEDCT, LOINC, RxNorm) based on concept semantic types. Add ConceptMapping to your pipeline components cluster, then can check the output of the terminologies mapping after running the pipeline.
Test-value Extraction
• Online live demo available now
By clicking this link (https://clamp.uth.edu/clampdemo.php), you can start surfing with CLAMP as you want.
Online Demo
Improvements
• Reduce the size of the default CRF NER model, increase memory efficiency.
• Implement a Java-based CRF predictor, improved running speed on large corpus.
• Improve the accuracy of NER-attribute / test-value / medication-signature pipeline.
Bug fixing
• Fix one resource leak.
• Fix a bug when selecting the file path from the file system.
• Use bright colors to highlight entities in the online demo.