Text Analytics (dv-) HS25

Natural Language Processing and Text Analytics

28 PDFs
5 Notebooks
13 Sessions

Learning Journey

1
Foundations
Weeks 1-5
2
Architectures
Weeks 6-9
3
Applications
Weeks 10-13

Course Overview

This hands-on course takes you from statistical language models to modern transformer architectures through discovery-based learning. You will build practical NLP systems using PyTorch while understanding the mathematical foundations that power today's large language models.

What You'll Learn

  • Build neural language models from scratch using PyTorch
  • Master transformer architectures and attention mechanisms
  • Fine-tune pre-trained models for sentiment analysis and text classification
  • Implement decoding strategies for controllable text generation
  • Understand efficiency techniques: quantization, pruning, and knowledge distillation
  • Apply ethical AI principles to NLP systems
Prerequisites: Programming experience in Python. No prior deep learning knowledge required. Mathematics: basic calculus and linear algebra helpful but not mandatory.

Learning Path

The course follows a progressive structure with three main phases:

Phase 1 Foundations

Weeks 1-5

  • N-Grams & Statistical Models
  • Word Embeddings
  • Neural Networks Primer
  • RNNs & LSTMs
Phase 2 Core Architectures

Weeks 6-9

  • Sequence-to-Sequence
  • Transformers
  • Multi-Agent LLMs
  • Decoding Strategies
Phase 3 Applications

Weeks 10-13

  • Fine-tuning & Transfer
  • Efficiency & Deployment
  • Ethics in NLP
  • Final Projects

Quick Links

Schedule

View 13-week course timeline with mandatory sessions

PDFs

Download 28 lecture slides and exercises

Notebooks

Access 5 interactive Jupyter notebooks

Assignments

Assessment overview and grading weights

Featured Charts

Sample visualizations from course materials:

Word Arithmetic 3D
Word Arithmetic
 LSTM Architecture
LSTM Architecture
 Attention Heatmap
Attention Heatmap
 BERT Architecture
BERT Architecture
 Temperature Effects
Temperature Effects
 Scaling Laws
Scaling Laws
 BPE Tokenization
BPE Tokenization
 LoRA Explanation
LoRA Fine-tuning
 Quantization
Quantization
 Bias Sources
Bias Sources

Frequently Asked Questions

What programming experience do I need? +
Intermediate Python experience is required. You should be comfortable with functions, classes, lists, dictionaries, and file I/O. Experience with NumPy is helpful but not required - we'll cover the basics needed for deep learning.
What software do I need to install? +
Python 3.8+, PyTorch, NumPy, Matplotlib, Jupyter Lab. Alternatively, you can use Google Colab for zero-setup cloud execution. Detailed installation instructions are provided in the first session.
How is the course graded? +
Three assessments: Kurzprasentation (20%, individual 5-min talk), Zwischenprasentation (10%, team mid-term), and Abschlussprasentation (70%, team final project). No written exams. See the Assignments page for details.
Can I access materials after the course ends? +
Yes! All course materials are available on GitHub and will remain accessible indefinitely. Lecture slides, notebooks, and exercises can be downloaded for offline use.