text-as-data

What problem does it solve?

Political texts are often too long, noisy, and heterogeneous to analyze by hand, so you need systematic quantitative methods to extract topics, ideology, sentiment, and actor information.

Core Features & Use Cases

• Topic modeling (LDA): Discover thematic structure in legislative speeches and manifestos and quantify topic mixtures per document.
• Ideology & scaling (Wordfish + TF-IDF classifier): Recover latent left-right positions with Wordfish scaling and classify ideology using TF-IDF features and logistic regression.
• Political NLP signals: Compute sentiment with VADER plus a custom political dictionary and extract named entities relevant to political actors and institutions.
• Corpus access for analysis (MANIFESTO API): Fetch annotated manifesto texts needed for reproducible political research.

Quick Start

Use the text-as-data skill to analyze a set of party manifesto paragraphs by extracting tokens, running LDA topic modeling, and returning per-document topic distributions suitable for comparing parties.