# Russian NLP

A Russian spam detection model based on the Deberta architecture, capable of classifying text as spam or non-spam.

A fine-tuned Russian multi-label emotion classification model based on RuBERT-tiny2, supporting 28 emotion categories recognition

A Russian text summarization model fine-tuned based on google/mt5-base, trained on approximately 50,000 samples, specializing in generating high-quality Russian summaries.

Vikhr is a Russian language model based on the Mistral architecture, undergoing three rounds of pre-training with 400 million tokens. It outperforms the original Mistral in Russian tasks but may have shortcomings in code processing.

This model is an open-source model supporting the Russian language, released under the Apache-2.0 license.

A Russian pre-trained language model based on the T5 architecture, employing a mixed training strategy with 7 denoisers similar to UL2, supporting various text generation tasks.

DeBERTa-base is a pre-trained bidirectional encoder for Russian, mainly used for processing Russian text tasks.

A bidirectional encoder model pretrained specifically for Russian language, trained on large-scale text corpora, supporting feature extraction tasks.

Large Russian NLP model provided by spaCy, optimized for CPU, featuring a complete NLP processing pipeline

This is a model for sentiment classification of Russian text, built on the ruBERT conversational embedder, capable of identifying neutral, positive, and negative sentiments.

Russian text sentiment classification model based on BERT architecture

A Russian large language model pre-trained by SberDevices team, based on Transformer architecture with 427 million parameters and 30GB training data

Stanza is a precise and efficient multilingual analysis toolkit that provides text analysis capabilities for Russian.

CPU-optimized Russian processing pipeline including token classification, dependency parsing, named entity recognition and other NLP tasks

spaCy's CPU-optimized Russian processing pipeline, including tokenization, part-of-speech tagging, dependency parsing, named entity recognition, and other features

RuBERT is a Russian BERT model trained on Russian Wikipedia and news data, with 180 million parameters, supporting masked language modeling and next sentence prediction tasks.

This is a compact version of the multilingual distilled BERT base model (case-sensitive), specifically optimized for Russian, capable of generating semantic representations identical to the original model while maintaining its accuracy.

Russian abstractive summarization model based on rugpt3medium_based_on_gpt2, specifically trained on the Gazeta dataset