Release 0.16.0

• A distilled Russian BERT
• New Pytorch-based models for classification, question answering and named entity recognition
• Removed some configuration files and components
• Small fixes and updates

Release 0.15.0

• bert_as_summarizer, seq2seq_go_bot, hyperparameter optimization by neural evolution and all deeppavlov.deprecated components were removed
• NER automodel
• Сomponent-based config requirements generation
• Minor edits and fixes

Release 0.14.1

• Fixed installation on python3.6

Release 0.14.0

• Intent Catcher component
• Support of the Hugginface Transformers for classification
• Go-Bot formfilling (tutorial)
• Entity Linking, Wiki Parser and KBQA as separate components
• Minor edits and fixes

Release 0.13.0

• Hugging Face datasets support
• Go-Bot now requires only RASA-based configs to train its components (intent catcher, slot filler, dialogue state tracker also known as Go-Bot itself)
• Prometheus metrics middleware for REST web service
• KBQA models fixes
• Minor edits of the documentation

Release 0.12.1

• Entity Linking for Wikidata (English)
• Bugfixes in KBQA model
• Other minor improvements

Release 0.12.0

• PyTorch Support (w/ Examples)
• Multi-task BERT
• Basic RASA Configs Support for Go-Bot
• Entity Linking
• top_n Answers API Update in ODQA Models
• Hybrid NER Models Trained on OntoNotes: Ru | En
• BoolQ Dataset Reader

Release 0.11.0

• Dataset generation tool with a tutorial for Goal-Oriented Dialog Bot
• Undocumented feature for downloading artifacts from Amazon s3
• New KBQA pipeline for online version of Wikidata
• First version of a KBQA pipeline using Syntax tree for generating SPARQL-queries
• And many small improvements and fixes

Release 0.10.0

• New Knowledge Base Question Answering model for WikiData
• Training interfaces for KBQA models
• Refactored goal-oriented bot architecture
• And many small improvements and fixes

Release 0.9.1

• Fix requirements for ASR and TTS configs