Christoforos Kapsalis

ITC6110 – Natural Language Processing

Institution: MSc in Data Science, the American College of Greece

Term: Spring 2024

Full Codebase and Report: GitHub Repository

Overview

This repository presents a structured collection of NLP assignments and deliverables completed during the Fall 2023 semester.

Each folder reflects one core topic, ranging from word embeddings to machine translation and custom NER pipelines.

Project Structure

.
├── 1. embeddings/
├── 2. context_embeddings/
├── 3. text_classification/
├── 4. sentiment_classification/
├── 5. machine_translation/
├── 6. sample_custom_ner/
├── 7. deliverables/
└── README.md

Component Descriptions

Folder Description
1. embeddings/ Static word embeddings via Word2Vec, FastText, GloVe; evaluates cosine similarity and analogy tasks.
2. context_embeddings/ Contextual embeddings (ELMo, BERT); comparisons of token representations across contexts.
3. text_classification/ Classical and neural models for text classification; includes Naive Bayes, SVM, and simple NN.
4. sentiment_classification/ Sentiment analysis on IMDb reviews using transformers (BERT, RoBERTa).
5. machine_translation/ Sequence-to-sequence model for English→German translation using PyTorch. Dataset: custom TSV format.
6. sample_custom_ner/ Custom named entity recognition (NER) using spaCy with a focus on COVID-related categories and JSON-labeled data.
7. deliverables/ Final presentation materials and summary PDFs.

How to Run (General)

  1. Clone the repo.
  2. Install dependencies.
  3. Open .ipynb files from each folder in Jupyter or VSCode and follow the cells.

Requirements

This project relies on the following major packages:

Notes

  1. Code in each folder runs independently.
  2. Folder 6. sample_custom_ner/ includes annotations_dataset/, config/, and my_categories/ used to train and test spaCy NER pipelines.
  3. Folder 5. machine_translation/ includes a small paired English–German dataset.

← Back to projects