Release 0.3.0

• Full spectrum of BERT-based models
• Multilingual BERT based NER for 100 languages and 19 classes
• Pre-trained BERT-base model for Russian Language
• Updated rusentiment model trained on twitter
• New AIML wrapper component
• New SMN, DAM, DAM-USE-T ranking models
• Minor fixes

Release 0.2.0

• Multilingual and English BERT-based models for classification and SQuAD
• Knowledge base based Question Answering model
• Refactor train script to class to simplify its modification and extension
• Super-convergence Keras model class
• +3 F1 for speller correction with new custom vocab_penalty property for searcher_component
• Refactored similarity matching skill
• Add universal method destroy in Component class
• --no-default-skill optional flag if your component implements an interface of DeepPavlov skill
• --stateful and --multiinstance flags for Amazon Alexa integration
• Move utils module from project root into deeppavlov module
• Many minor improvements and bug fixes, including logging, ELMo save/load and many other

Release 0.1.6

• ODQA with scores: LogitRanker now returns a batch of best answers and their scores
• Experimental similarity matching skill with updated tutorial
• Amazon EC2 deployment guide
• Links to RuSentiment dataset were removed due to legal issues
• Minor bug fixes

Release 0.1.5.1

• Fix go_bot compatibility with the previous release
• Extended support for env variables in configs

Release 0.1.5

• Out-of-the-box Amazon Alexa integration for any skill/model
• New popularity ranking for ODQA
• New fine-tuned models for paraphrase identification
• Support for learning rate & momentum schedulers

Release 0.1.1

• Bug fixes and minor improvements
• New ELMO configs
• Manage some settings with env variables

Release 0.1.0

• Variables in configs and improved python API
• Trainable ELMo model with TF Hub wrapper and support for large datasets
• "Few shot" NER model
• New ODQA model with support for "no answer" response
• Model for sentiment analysis in Russian with ELMo embeddings
• Pretrained model for classification Informational vs. Conversational questions on Yahoo dataset
• eCommerce bot with tfidf retrieve
• Improved logging, including dialog logging for Agents, docs and many other improvements and bug fixes
• Also check out our docker images for popular models on hub.docker.com

0.1.0b

• Human readable logs

0.1.0a

• Dialog logging
• API refactoring
• Settings files refactoring
• Other minor improvements and bug fixes

Release 0.0.9

• Improved MorphoTagger now supports 10 languages
• Yandex.Alice integration for agents with many skills
• New SQUAD model that supports "no answer" token
• Universal wrapper for sklearn estimators and vectorizers
• New universal interface for metrics calculation
• Many other improvements and bug fixes