Named Entity Recognition (NER) is the key ingredient of any NLP system and it is regarded as the first building block of question answering, topic modelling, information retrieval, etc. In the medical domain, NER plays the most crucial role by giving out the first meaningful chunks of a clinical note and then feeding them as an input to the subsequent downstream tasks such as clinical assertion status, clinical entity resolvers, and deidentification.

Reimplementing a well known BiLSTM-CNN-Char NER architecture into Spark environment, Veysel will present a single trainable NER model that would be used with nearly all the NER data sets with no configuration change and exceeding the latest state-of-the-art biomedical and clinical NER benchmarks on publicly available datasets in less than 10 epochs.

This production-ready NER implementation can also be extended to other spoken languages with zero code changes and can scale up in Spark clusters.