summer internship project @ JetBrains Research

Token Classification task on the Yes We Can dataset

Links to my repositories, where I implement a wide variety of Natural Language Processing models using TensorFlow and Hugging Face.

Scrap, token classification and model deployment for a selective process.

ArabiNizer is a state-of-the-art Arabic named entity recognizer (NER) leveraging the XLMR transformer model with an impressive testing accuracy of 95.00% and a remarkable testing F1-score of 88.00% on the PAN-X.AR subset from XTREME.

A 16M LLM for POS tagging in African languages

This repo provides scripts for fine-tuning HuggingFace Transformers, setting up pipelines and optimizing token classification models for inference. They are based on my experience developing a custom chatbot, I’m sharing these in the hope they will help others to quickly fine-tune and use models in their projects! 😊

Token Classification in essay level, paragraph level and sentence level with BERT, DistillBERT and NER

Keyword extraction to automate the discovery of dataset in publications and public reports

Part-Of-Speech tagging in polish with finetuned RoBERTa model

API for Yoda-NER and Yoda-FITS model. NLP models for Google Feed product optimization

A webapp built using Gradio for demonstrating the capabilities of the Spacy NER pipeline.

RE-Miner Dashboard for Visual Analytics, Review & Market Analysis

Проект в рамках ВКР под названием "Разработка программного модуля для анализа документов, подтверждающих индивидуальные достижения"

End-to-end pipeline for (1) automatic scraping and parsing of NLP research papers, (2) token-level entity annotations in Label Studio, and (3) BERT-based models for span identification and entity recognition

A state-of-the-art Arabic part-of-speech tagger leveraging the XLMR transformer model With an impressive testing accuracy of 97.49% and a remarkable testing F1-score of 96.44% on the Arabic UD Treebank.

Data pipelines for both TensorFlow and PyTorch!