B

BERT

Generátor novinářských fotbalových reportů v češtině pomocí LLM (Claude, Ollama). Obsahuje evaluaci výstupu přes ROUGE, BERTScore a faktickou přesnost. Kontroly pomocí pomocí modelu mBERT (bert-base-multilingual-cased).

A modular Clinical NLP Pipeline built to process and analyze unstructured medical text using both traditional machine learning and transformer-based approaches.

The project combines multiple components including OCR, text preprocessing, feature engineering, classification, named entity recognition, and visualization into a single end-to-end pipeline. It supports extracting clinical insights from raw documents and predicting medical categories using both TF-IDF + SVM and BERT-based models.

The system was designed and implemented as a structured Python project, with each stage separated into independent modules for scalability and maintainability.

Key Highlights

Built an end-to-end NLP pipeline for clinical text processing. Implemented SVM (≈51% accuracy) and BERT (≈77% accuracy) models. Integrated OCR for extracting text from medical documents. Performed Named Entity Recognition (NER) on clinical data. Designed modular architecture (src/) for clean code organization. Exported outputs for visualization and dashboard integration.

Command Line Interface for developers of qary -- the open source, teachable AI assistant that truly assists, rather than manipulating you.

Using Natural Language Processing (NLP) on job ads for applications in Econometrics.

Demo of Semantic Similarity using spaCy vector and transformer models

In MT-BERT we reproduce a neural language understanding model which implements a Multi-Task Deep Neural Network (MT-DNN) for learning representations across multiple NLU tasks.

Sentiment Analysis on IMDB movie review dataset using BERT.