Exploratory Text Analytics of Gothic & Sci-Fi Literature

The Problem

Literary scholars have long debated how Gothic and Science Fiction traditions influence each other, but arguments rely on close reading of individual texts. Can computational methods reveal structural and thematic patterns across entire corpora that human readers might miss?

What I Built

A complete text analytics pipeline processing 12 novels from Project Gutenberg through multiple NLP stages, producing quantitative evidence of cross-genre cultural evolution.

Pipeline

Stage	Method	Purpose
Preprocessing	Tokenization, stopword removal, lemmatization	Normalize text for analysis
Term frequency	TF-IDF vectorization	Identify distinctive vocabulary per text
Dimensionality reduction	PCA	Visualize document similarity in 2D space
Topic modeling	LDA (Latent Dirichlet Allocation)	Discover latent thematic structure
Semantic embeddings	word2vec	Capture contextual word relationships

Key Findings

• PCA clustering: Gothic and Sci-Fi novels form distinct but overlapping clusters in vector space, with transitional works (e.g., Shelley's Frankenstein) positioned between them
• Topic overlap: LDA revealed shared topics around "science and discovery" that bridge both genres, while "supernatural dread" topics are exclusively Gothic
• Temporal evolution: Later Sci-Fi works show decreasing vocabulary overlap with Gothic predecessors, suggesting genre divergence over time
• word2vec: Semantic neighborhoods for key terms ("monster", "machine", "creation") shift meaning between genres in measurable ways

Corpus

12 novels spanning 1764-1898, including works by Shelley, Stoker, Stevenson, Verne, and Wells. All sourced from Project Gutenberg (public domain).

Links

GitHub Repository | Research Report (PDF)