Hybrid Pipeline - CLAMP
The CLAMP system provides high-performance NLP modules and pipelines customized for clinical text. Its user-friendly interfaces quickly build customized NLP pipelines.
-
Annotate local data and retrain machine learning models with a few clicks
-
Specify rules on top of ML models to improve performance
-
Export NLP pipelines as web services for easy integration
-
Provide Java APIs for individual components for system integration
Clamp FEATURES
Move to structured, adaptable and actionable data
NLP BUILT FOR CLINICAL TEXT
-
Built on proven and accurate methods and models trained using clinical corpora
-
Award-Winning algorithms from many clinical NLP challenges such as i2b2 clinical NER (2009/2010-#2), SHARE/CLEF (2013-#1), and SemEval2014 UMLS encoding (#1)
ANNOTATION AND CORPORA MANAGEMENT
-
Built-in, adaptable annotation tool that supports different IO methods (e.g., files or databases) and collaborative annotation by a team
-
Compatible/flexible between different annotation tools (e.g., Brat, eHost, MAE etc.)
MACHINE LEARNING AND HYBRID APPROACHES
-
With one simple click, gain unparalleled access to unstructured textual data to train your own models using machine and deep learning algorithms.
-
Compare different models with provided evaluation programs and visualization interfaces
CUSTOMIZED PIPELINES
-
Build your own NLP pipelines using both machine learning and rules using our adaptable software.
-
Leverage existing pipelines to avoid NLP development from scratch
INTEROPERABILITY AND SCALABILITY
-
Build on the UIMA framework, thus compatible with other systems such as cTAKES
-
Thread safe and validated on parallel computing platforms such as SPARK and Hadoop