VerseVAD is a computational poetry analysis application built to help readers examine patterns in poetic language, sound, structure, and affect.
It brings several established linguistic resources and analytical methods into one workspace, making it possible to study individual poems, larger collections, and relationships among texts.
VerseVAD does not attempt to explain a poem for the reader. It is designed to support closer attention—to help make patterns visible, comparisons more systematic, and interpretive questions easier to pursue
Where the Project Began
VerseVAD began with a literary research question.
Could established affective and linguistic datasets help us describe patterns within a poet’s work—across poems, collections, and periods—without flattening those poems into a single score or automated interpretation?
Answering that question required more than one measurement. Emotional language is only one part of a poem’s texture. Word familiarity, concreteness, rhythm, stress, rhyme, lineation, repetition, and lexical choice may all contribute to how a poem works.
VerseVAD grew from the need to examine those features together rather than through a collection of disconnected tools.
The result is an environment for moving between close reading and computational comparison: from the individual word or line to the poem, the collection, and the reference corpus.
[Optional image placement: a clean screenshot of a corpus or VerseMap analysis.]
What VerseVAD Measures
VerseVAD draws on established lexical datasets, pronunciation dictionaries, and natural-language processing methods to examine features such as:
- Valence, arousal, and dominance
- Emotion words and emotional intensity
- Concreteness
- Word frequency and rarity
- Age of acquisition
- Part of speech and lemmatization
- Syllables, lexical stress, rhythm, and meter
- Rhyme, slant rhyme, and sound patterning
- Line, stanza, and poem-level structure
- Similarity and distance within poetry corpora
These measurements can help identify tendencies that may be difficult to recognize through impression alone. They can also help test an intuition, compare readings, locate unusual poems within a collection, or reveal questions that deserve further attention.
They do not establish what a poem ultimately means.
Measurement is Not Interpretation
Poems are made of measurable language, but they are not reducible to measurements.
A word may carry irony, ambiguity, historical associations, figurative meaning, or emotional force that a general-purpose lexicon cannot fully represent. A dictionary pronunciation may differ from a poet’s intended rhythm or a reader’s performance. Two poems with similar numerical profiles may produce profoundly different experiences.
VerseVAD is built around that distinction.
Its results are best treated as evidence within an interpretation—not as interpretations in themselves. A score can describe a tendency. A visualization can show a relationship. A corpus comparison can identify what is typical or unusual. The reader must still decide why the pattern matters.
Computational analysis is most useful when it returns us to the poem with better questions.
What VerseVAD is Not
VerseVAD is not:
- A measure of poetic quality
- A system for ranking poems from best to worst
- A detector of authorial intention
- An objective account of emotional meaning
- A replacement for close reading or literary history
- A tool that resolves interpretive disagreement
- A substitute for understanding the datasets and methods behind a result
The purpose of VerseVAD is not to remove uncertainty from reading. In many cases, its most valuable result may be the discovery of a complication, exception, or contradiction.
Principles Guiding the Project
Interpretive Restraint
VerseVAD describes patterns without claiming more than the underlying methods can support.
Transparency
Users should be able to understand where a result came from, which words were matched, what settings were applied, and where coverage was incomplete.
Reader Control
Analytical choices matter. VerseVAD provides controls for features such as stopword handling, token and type weighting, reference corpora, thresholds, pronunciation overrides, and other methodological decisions.
Reproducibility
Research results should be connected to identifiable datasets, processing settings, corpus contents, and software versions.
Usability
Computational literary analysis should not require every user to build a programming environment from scratch. VerseVAD is intended to make sophisticated methods available through a practical visual interface.
Respect for Poetic Language
Poetry regularly exceeds the assumptions of general-purpose language tools. VerseVAD is designed with those tensions in mind rather than treating them as inconveniences to hide.
Who VerseVAD is for
VerseVAD is being developed for people who want to examine poetry carefully and systematically, including:
- Scholars conducting computational or corpus-based literary research
- Teachers introducing students to digital approaches to literature
- Students developing research questions or comparing texts
- Poets interested in the language, sound, and structure of their work
- Editors and readers exploring patterns across poems or collections
- Digital-humanities researchers seeking an integrated poetry-analysis environment
No programming experience is required for ordinary use. Some knowledge of literary analysis remains important—not because VerseVAD is inaccessible, but because its results become meaningful through informed reading.
An Evolving Project
VerseVAD remains under active development.
New analytical modules, visualizations, documentation, and workflow improvements will continue to be added. Some methods are more mature than others, and limitations will be documented rather than concealed.
The project’s aim is not to create a finished machine that settles questions about poetry. It is to build a useful, transparent, and adaptable environment in which those questions can be explored.
Explore VerseVAD
Begin with a poem, a collection, or a question you want to investigate.