v2021.2.0

Major changes

• Upgrade spacy to v3.0
• Add native HuggingFace support (#191), re-writting BertClassifier using transformers
• Disables HDBscan from the possible clustering techniques due to a conflict with the new numpy version (#197)

Bug fixes

• Resolves issues #195 and #198 with thew pip reference resolver, introduced in pip>20.3

v2021.1.0

Merge pull request #183 from wellcometrust/feature/nsorros/upgrade-sp…

v2020.11.1

Merge pull request #180 from wellcometrust/fix-dataclasses-version

Pin dataclasses to 0.6 as required by spacy-transformers

v2020.11.0

Merge pull request #175 from wellcometrust/feature/bert_vectorizer_pr…

v2020.9.0

Models

• Add param l2, dense_size, attention_heads, metrics, callbacks, feature_approach to cnn and bilstm classifiers
• Faster predictions in BertClassifier through the use of spacy's pipe
• Adds TextClustering

v2020.7.1

Bugs

• Fixes pypi conflict by pinning down dependencies.

v2020.7.0

Models

• Adds Doc2VecVectorizer
• Adds WellcomeVotingClassifier
• Adds Sent2VecVectorizer
• Adds SemanticEquivalenceMetaClassifier
• Adds CategoricalMetrics and MetricMiniBatchHistory

Datasets

• Adds CONLL dataset
• Adds Winer dataset

Features

• Automatically load models like en_core_web_sm and en_trf_bertbaseuncased_lg but also download packages like sent2vec, only when needed
• Adds docs based on sphinx and read the docs

Repo

• Adds pep8 / flake8 checks and address violations
• Adds badges for build, codecov and license
• Adds pull request template that forces link to issue or trello

Bugs

• Fix dependency on non pypi packages for tests
• Pin spacy transformers to 0.5.1
• Fix codecov running in separate travis venv

v2020.5.1

ML

• Add l2 and validation_split to BertClassifier
• Add kwargs and load model to semantic similarity

Extras

• Add command line interface to download models
• Removed models from deployment
• Moved package to pypi

Pre-release v2020.5.0

ML

• Add CNNClassifier
• Add BiLSTMClassifier
• Add attention layers
• Add Semantic equivalence classifier
• Add embedding based entity linker

Datasets

• Add Hoc dataset

Pre-release v2020.4.0

ML

• Add partial_fit to BERTClassifier
• Add mean_last_four embedding to BertVectorizer
• Use nlp.pipe for prediction as its quicker
• Add generator to transform data on demand for spacy to reduce memory usage
• Add multilabel and architecture parameter in SpacyClassifier
• Modify SpacyClassifier to accept sparse Y for multilabel classification
• Add pretrain_vectors_path parameters to SpacyClassifier
• Add speed metric to SpacyClassifier and BertClassifier
• Fix tests in BertClassifier to check for loss reduction after 5 iterations