Lexical map of Lolita

Mapping the distinctive vocabulary of Nabokov's novel with TF‑IDF

Darja Stryho ·

Abstract

This project applies TF‑IDF and lexical‑diversity measures to Vladimir Nabokov's Lolita. The 69‑chapter split illustrates the the importance of the aestheticism in the book. A slight difference can be noticed, however, since Part one is saturated with more intimate, domestic terms and Humbert's childhood (diary, key, toilet, frock), while Part Two shifts to more outdoor pursuit because that is when Humbert sets out on a journey across America with Lolita, with the more distinctive words such as (ball, racket, hunter, journey). The most distinctive word in the entire novel is seraph – used only twice, but through which Humbert idealizes his lost childhood experience and his first love Annabel Lee. The frequency of nymphet across chapters (plotted below) confirms that is appears predominantly in Part one, dropping after the Part One/Part Two boundary, which might point to the fatc that Humbert first defines the nymphets, but as the story progresses, Lolita is shown more as a human than a mythological creature.

Result

TF-IDF weighted word clouds for Part One and Part Two — Figure 1. TF‑IDF‑weighted word clouds for Part One (left) and Part Two (right).

Bar chart of top 15 distinctive words — Figure 2. Top 15 overall distinctive words by mean TF‑IDF score.

Frequency of the word 'nymphet' per chapter — Figure 3. Frequency of the word “nymphet” across all 69 chapters.

Mean Segmental TTR per chapter — Figure 4. Mean Segmental TTR (500‑word chunks) per chapter – content words only.

The full chapter‑level data is available as a CSV file.

The numbers

The full data is in chapter_metadata.csv.

Item	Measure	Value
Most distinctive word (Part 1)	TF‑IDF (mean)	sand, key, page, frock, born, sea, enchanted, toilet, box
Most distinctive word (Part 2)	TF‑IDF (mean)	ball, racket, boy, hunter, party, train, girl, understand, journey
Overall most distinctive word	TF‑IDF (max)	seraph

How it was made

The full text of Lolita was split into 69 chapters using Python. Words were lowercased, lemmatised with NLTK's WordNetLemmatizer, and filtered by removing English stopwords as well as a custom list of character names and common narrative words (e.g., “said”, “know”, “like”, “time”). TF‑IDF vectors were computed per chapter with scikit‑learn's TfidfVectorizer, then post‑filtered to exclude words appearing in more than 20% of all chapters. The word clouds were generated from mean TF‑IDF scores per part. The complete pipeline, including all figures and data, can be reproduced by running the Python notebook in this repository. See the README for full details.