Ivan Yu. Bondarenko

The python library and service for automatic speech recognition and transcribing in Russian and English

A framework for few-shot evaluation of language models.

This repository contains demos for our Natural Language Processing (NLP) course at Novosibirsk State University

This is a smart chunker for efficient preparing of long document for RAG

RAG benchmark

LLM evaluation for solving "classical" NLP tasks

This project is concerned with my participating in the RuNNE competition https://github.com/dialogue-evaluation/RuNNE

Spoken Language IDentification (LID) using multilingual Whisper model

[NIPS'24] UDA: A Benchmark Suite for Retrieval Augmented Generation in Real-world Document Analysis

LSTM-Seq2Seq on the bases of Keras with the simple ScikitLearn interface

The fine-tuning and evaluation of FRED-T5-based seq2seq models

The script for inverse text normalization of ASR dataset annotations in Russian

Automatic evaluation of the data difficulty for autoregressive LLMs

Named entity recognizer based on ELMo or BERT as feature extractor and CRF as final classifier

The coreference resolution dataset based on RuCoCo

Experiments to build an instruct large language model for Russian language.

Detector of the need to use RAG in addition to LLM: training and analysis

This is a complexity checker of samples in SFT datasets for LLM

This repository is devoted to my experiments with invariant risk minimization (IRM) for deep learning

This repository devoted to a multimodal AI system for the AIJ contest "Strong AI"

Various algorithms for voice activity detection

Experiments with self-adaptive hierarchical learning for text classification

Wav2Vec2-mGPT speech encoder-decoder with multitask learning

The best solution of the Weather Prediction track in the Yandex Shifts challenge

Морфологический анализатор на основе нейронных сетей и pymorphy2

No description provided.

Text classifier, based on the BERT and a Bayesian neural network, which can train on small labeled texts and doubt its decision.

No description provided.

Named entity recognizer based on BERT and CRF

Constrastive learning for Wav2Vec2-based speech-to-text