""" README Task 11 System Authors: Jeffrey Smith, Bill Cramer, Evan French """ INTRODUCTION The scope of the Task 11 system is to provide an end to end NLP pipeline for the recognition of named entities in Biomedical text. It utilizes labeled training data to train a Bidirectional LSTM-CRF model. The system will be able to save and load the model as needed for future reproducibility and evaulation. Named Entity Recognition (NER) is a subtask within the Information Extraction (IE) branch of Natural Language Processing (NLP). Given a corpus, an NER system identifies groups of concepts. Historically, NER systems were programmed using hand-crafted rule based algorithms. Now, machine learning techniques have their place thanks to modern computing hardware offering a high degree of computational efficiency. Public, open source general NER systems offer a high degree of precision and accuracy for non-specific recognition tasks. These systems fall short, however, when attempting to analyze domain specific categories. As a result, NER systems have to be custom tailored to the domain and ontology they are meant to represent. One such domain is the biomedical domain, which contains data, relations, and concepts associated with biology and medicine as it relates to human systems. As an example, in the sentence, "Significant for hypertension, hyperlipidemia.", hyperlipidemia and hypertension can be classified under the category symptoms. AUTHORS Jeffrey Smith - Developed the neural network training and evaluation portions along with some of the base classes for sentence structures. Also setup the system to convert words into vector embeddings. Bill Cramer - Developed the preprocessing functionality to parse out date and number structures as well as constructed presentations and obtained sources. Evan French - Developed the annotation parsing system that matches the text from sentences to concept annotation files. USING Run the program using the temporary command: python3 ./run_test.py OR ./runit.sh This command will be replaced in the future with the run.py file that accepts commandline arguments.