rubert

Distillation of BERT model with catalyst framework

MultiLabel classification of cow diseases by text and symptoms recognition (NER)

This repository provides code solution for Data Fusion Contest task 1

Coreference resolution model for Russian fully compatible with AllenNLP package

Cointegrated/rubert-tiny2 model fine-tuned to classify 10 emotions in Russian using a custom dataset

Модель стилометрического анализа на базе ruBERT. Определяет авторство русскоязычных текстов по синтаксическому «почерку», пробивая маскировку и смену темы.

A Telegram bot that filters comments in your channel. It deletes toxic comments

Sentence-transformer model for payments classification. Fine-tuned from rubert-tiny-turbo

Проект с хакатона 14 апреля по определению токсичных комментариев.

This project is about analysis of reviews tonality from KinoPoisk

Synthetic Corpus Generation and Multi-Model Benchmarking (BERT, RoBERTa, QLoRA & Few-Shot LLMs) for Low-Resource Russian Messenger Phishing Detection under Extreme Class Imbalance.

Интерпретация диалектных слов в российских речевых корпусах

Telegram bot that classifies channel posts with a fine-tuned RuBERT model and sorts them into topics of a forum it creates — plus the labeled dataset and notebooks behind the model.

Sentiment analysis of data from the RuSentNE-2023 competition using the RuBERT model, its modifications, and various techniques.

NER for Russian financial documents: rubert-tiny2, FastAPI, Streamlit. F1=1.0 on synthetic data.

RAG база знаний: семантический поиск по документам (PDF/TXT/MD), точные ответы с цитированием источников, ChromaDB, ONNX INT8, FastAPI и React

Классификация обращений клиентов банка на RuBERT и LLM-агент с инструментами: Postgres, RAG на pgvector, эскалация оператору

Русскоязычная QA система обернутая в Telegram-бот

NLP project for classifying marketplace listings using TF-IDF and RuBERT