This repository contains Jupyter notebooks detailing the experiments conducted in our research paper on Ukrainian news classification. We introduce a framework for simple classification dataset creation with minimal labeling effort, and further compare several pretrained models for the Ukrainian language.

NVIDIA Speech Artificial Intelligence. Speech AI Summit 2022.

このリポジトリは、xLSTMを用いた言語モデルを学習するコードを実装したものです。 論文「xLSTM: Extended Long Short-Term Memory」に基づいて実装しています。

Comparative fake news detection framework using TF-IDF + classical machine learning models and BERT fine-tuning, achieving up to 99.26% accuracy.

A curated list of resources dedicated to Natural Language Processing.

This project implements a Text Simplification system using a hybrid BERT–GPT-2 architecture. It leverages BERT for semantic understanding and GPT-2 for fluent generation, aiming to transform complex text into simpler, more readable versions while preserving meaning. The system is trained and evaluated on benchmark simplification datasets and design

This repository contains the implementation of the coherence assessment models discussed in the Morad Litime's paper "Improving Dialogue Management for DarijaGenie through Neural Coherence Modeling"

Public web client for Inline, a Socratic AI tutor designed to build critical thinking skills using a Human-in-the-Loop data engine.

CurupiraIA: A Brazilian Portuguese hate speech detection model using BERT fine-tuning, inspired by folklore guardianship principles to protect digital communities from toxic content.

Context sensitivity benchmark across 14 LLMs (GPT-4o/5.2, Claude, Gemini Flash, Llama 4, DeepSeek V3.1, Kimi K2, Qwen3, Mistral). 112,500 responses across 25 model-domain runs using ΔRCI metric. Medical + philosophy domains. Validates information hierarchy in 96% of runs. Reproducible cross-domain AI evaluation framework.