Python for Linguistic Data Analysis

Setup Python and VS Code

Operating System

Choose based on what you have:

• Windows: PowerShell or Windows Terminal (recommended)
• macOS: Built-in Terminal or iTerm2
• Linux: Native terminal (GNOME, Konsole, etc.)

Terminal/Shell

Where you run Python commands:

• PowerShell (Windows) – Modern, cross-platform, recommended
• Command Prompt (Windows) – Traditional, sufficient
• Bash (macOS/Linux/WSL) – Standard Unix
• Zsh (macOS) – Bash-compatible with nice defaults

Recommendation: Use whatever came with your OS. PowerShell on Windows.

Install VS Code + Extensions

Reference extensions for Python development:

1. Download Visual Studio Code
2. Install from marketplace:
   • Python (Microsoft) – Core language support
   • Jupyter (Microsoft) – Notebook support
   • Pylance (Microsoft) – Type checking & intellisense

Install Python

• Required: Python 3.9+
• Download from python.org
• macOS: brew install python3
• Linux: apt install python3
• Optional: Use pyenv for multiple versions

Verify Installation

Open terminal and run:

python --version    # Should show Python 3.9+
pip --version       # Package manager version

Clone Repository & Git

These slides as well as two datasets are in a public repo called ‘pyws’ on my GitHub Account ‘jobschepens’. You can clone these files using the command line:

git clone https://github.com/jobschepens/pyws.git
cd pyws

Before cloning, make sure git is installed on your system. Open your terminal and navigate to the directory where you want to store the repository (e.g., C:\GitHub on Windows or ~/projects on macOS/Linux). You can also use GitHub Desktop if you prefer a graphical interface.

For today’s workflow, you can work directly in the cloned repository: create new subfolders for your analyses, then commit and push changes to the main branch. For collaborative or more advanced workflows, consider forking the repository on GitHub, cloning your fork, creating a feature branch for your work, committing and pushing your changes, and then opening a Pull Request to merge your work into the original repository. This approach is recommended for larger teams or when contributing to shared projects.

Create Virtual Environment

Isolate project dependencies (venv recommended; conda/poetry/pipenv are alternatives):

# Create
python -m venv .venv

# Activate
# Windows (PowerShell):
.venv\Scripts\Activate.ps1
# Windows (Command Prompt):
.venv\Scripts\activate
# macOS/Linux:
source .venv/bin/activate

Install Packages

pip install -r requirements.txt

Installs: pandas (data), numpy (computing), matplotlib (visualization), jupyter (notebooks). Note that these packages are relatively large (often 100–200 MB total for all dependencies).

Verify in VS Code

1. Open VS Code, open pyws folder

2. Terminal: `Ctrl+`` (backtick) → Activate venv

3. Run verification:

   python --version
   pip list

LLM Configuration & Setup

Choose a Provider

Pick one or more (free tiers available):

What is CoCo AI?

CoCo AI: Code completion service via Chat AI on GWDG’s AcademicCloud

• Integrated LLMs: Llama, Codestral, Qwen models
• No local setup needed: Remote via API
• Access control: SAIA API key (institution login)
• Ideal for: Research institutions with GWDG access

Get API Credentials

• Most providers: Generate API key in account settings
• GitHub Copilot: Use GitHub login (no separate key)
• Store safely: Use environment variables, never commit to git

# Example: .env file (add to .gitignore!)
SAIA_API_KEY=...
OPENAI_API_KEY=sk-...
ANTHROPIC_API_KEY=sk-ant-...

Install VS Code Extension

Popular options for different workflows:

• GitHub Copilot Chat
• Continue
• Cline
• Roo Code

Configure & Test

1. Enter API key or GitHub login
2. Select model (GPT-4, Claude 3.5, Gemini, Llama)
3. Test with: "Write a Python function to read a CSV file"

Optional MCP Servers

Advanced: Model Context Protocol for complex workflows. Not required for basic setup.

Cloud Alternatives: Overview

Popular cloud-based alternatives to local setup:

General CLI vs. Extension Philosophy

Core toolkit (CLI): pip, git, gcloud, docker, poetry, nox

CLI Workflow

# 1. Set up reproducible environment (in version control)
python -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt

# 2. Run batch processing / CI validation
cline-agent --task "refactor test suite" --approve-all

# 3. Final validation & deployment
nox -s test
nox -s lint
gcloud deploy app.yaml