MedaCy 1.0.0 - BERT Implementation, Improved CLI, Package Overhaul

• BERT learner now available
• More robust command-line interface with both built-in and online documentation
• Overhaul of the internal structure of the package; this will only affect those importing internal medaCy modules and not CLI users

MedaCy 0.1.1 - SpaCy Model, Scispacy Pipeline, and CLI

• Added a spaCy model for comparing spaCy only results
• Added a Scispacy pipeline
• Added a CLI tool for train, predict, and evaluate functions
• Added groundtruth directory saving
• Improved logging

MedaCy 0.1.0 - Relation extraction, Dataset Analysis

• Segmented NER to allow for relation extraction functionality
• Implemented improved data analysis functionality.

Included Dataset analysis functionality, Travis builds

• New Dataset analysis functionality, see API.

Vagrant Integration, Documentation improvements, Restructuring of file management

• Integration provided for launching a Vagrant virtual machine
• Restructuring to allow for validation at the document level as opposed to only at the sequence level.
• Improved documentation and many documentation error fixes

New Dataset interface, New features on Annotation Analysis

New Dataset interface, New features on Annotation Analysis

• Dataset.load_external() now returns a tuple containing an training_dataset, evaluation_data, and meta_data. Meta_data is a dictionary that can include entities, relations, and any other information about the dataset.
• Annotations object now support analysis between two annotations. This functionality includes .compute_confusion_matrix(), compute_ambiguity(), intersection(), etc. These methods can be used to analyze the output of NER systems with gold annotated data. See https://medacy.readthedocs.io/en/latest/medacy.tools.annotations.html for details.

Improvements to Annotation class, Documentation and Examples

Annotations

• An annotation is now view-able with displaCy.

Improved documentation

• Documentation on contributing has been improved.
• Typos have been fixed in examples

Live API Documentation, Improved Annotation analysis, Detailed Examples

• The API docs are hosted here
• New features are available to analyze and compare annotations.
  1. Compare functionality to see a difference between gold annotations and system predictions
• Detailed example and guides provided

Models, Dataset's, Tests and Documentation

Models, Datasets, and Tests

Data:

• Added Dataset class, removing DataLoader, to facilitate loading data.
• Added capabilities to integrate medaCy compatible data packages into Dataset.
• MedaCy compatible data packages can be versioned, pre-metamapped, and easily distributed (or not!).

Models:

• Added support for a clinical notes NER model that is installed from a medaCy compatible model python package.
• MedaCy compatible model packages can be versioned, benchmarked, and easily distributed (or not!).

Tests

• Extensive unit tests for all major medaCy that integrate with outside medacy_data_end python package.

Documentation

• Updated README.
• Added framework for examples
• Added Sphinx and auto-doc.

Compatible with v1.0.0 of medacy_clinical_note_model

Compatible with v1.0.0 of medacy_clinical_note_model for prediction of raw strings. Anyone can install this release to try out prediction functionality on their local machine.