Skip to content
Lizely

Rhyme Match

Match five real rhyming pairs by stressed sound tails rather than misleading spelling patterns.

Privacy: your files never leave your device. All processing happens locally in your browser.

How to use

  1. 1.Read the target word and choose the candidate you believe shares its complete ending sound from the last stressed vowel onward.
  2. 2.Use touch, pointer, keys 1 through 4, or arrows and Space to select, then choose Submit match or press Enter.
  3. 3.Clear five verified rounds for 200 points each and an exact score of 1,000; review the ARPAbet tail only after a correct answer.

About Rhyme Match

Rhyme Match is a five-round pronunciation game that asks one clear question at a time: which candidate shares the target word's ending sound? The game uses an explicit, testable definition rather than guessing from spelling. For each cited CMUdict pronunciation, it finds the last vowel carrying primary or secondary stress and compares every phoneme from that vowel through the end of the word. A valid match must have exactly the same stressed tail and a different onset. Each correct round awards 200 points, and solving all five produces exactly 1,000 points.

The verified pairs are light and night, cake and lake, moon and spoon, train and rain, and blue and glue. Their CMUdict tails are AY1 T, EY1 K, UW1 N, EY1 N, and UW1 respectively. Correct choices are distributed across option positions 1, 3, 0, 2, and 1 using zero-based indexing, so repeatedly choosing one location cannot clear the game. Every round contains exactly one candidate satisfying the production rule. Distractors are selected to fail for a specific phonetic reason, such as a different stressed vowel or a different final consonant.

Spelling is not used as a rhyme oracle. The boundary pair rough and cough shares the conspicuous written ending “ough,” yet CMUdict records R AH1 F and K AA1 F, so the stressed tails differ. Light and late both finish with the phoneme T, but AY1 T does not equal EY1 T. Moon and moan share final N while UW1 N differs from OW1 N. The game also excludes an identical pronunciation paired with itself because the rule requires a different onset; this prevents simple word duplication from counting as a match.

The source data is auditable. The maintained CMU Pronouncing Dictionary repository and its raw dictionary file are the primary ARPAbet sources. Cambridge Dictionary pages independently cross-check the ten played words plus rough and cough. The repository, raw data URL, and all twelve Cambridge pages returned HTTP 200 during implementation. A focused test contains its own literal pronunciation fixture and a separately written tail oracle. It does not import runtime answer tables to decide whether a pair rhymes.

Before grading, the accessible page contains only the target word, four neutral candidate words, cursor state, and selection state. It does not expose the current ARPAbet strings, the shared tail, a correct-answer data attribute, a correct CSS class, or an answer-bearing ARIA description. After a correct submission, a review note may show the previous pair and its verified CMUdict tail. Wrong submissions do not reveal the answer. This keeps the phonetic explanation useful without turning the current question into a DOM inspection exercise.

Touch and pointer users receive four answer controls and a full-width submit button, all at least 44 pixels high. Options use one column on narrow screens and two columns when space permits, with breakable text and no forced page overflow. Keyboard users can move through the two-by-two option layout with arrow keys, select using keys 1 through 4 or Space, and submit the highlighted choice with Enter. GameShell provides Restart, optional local best score, and the double-Escape boss screen.

Wrong attempts use a signature made from round and option. Repeating the identical wrong selection is an atomic no-op, while a second different wrong signature closes the run until Restart. Invalid option indexes, submitting without a selection, cursor moves beyond the option grid, and every action after completion or deadlock preserve state. The game runs entirely in the browser and loads no pronunciation audio, external asset, API, account flow, or new dependency.

Pronunciation varies across speakers, regions, and accents. This tool therefore makes a narrow claim: the five answers match under the cited CMUdict entries and the disclosed stressed-tail rule. It does not claim that every English accent realizes every vowel identically, that spelling predicts speech, or that the five-round set is a universal rhyming dictionary. The interface states that limitation in plain language beside the rule.

Methodology & sources

CMUdict ARPAbet arrays are scanned backward for the last phoneme ending in stress 1 or 2. The tail from that vowel through the final phoneme must match exactly, while the pre-tail onset must differ. Five literal rounds each contain one match and distributed correct indices [1,3,0,2,1]. An independent test fixture and private oracle verify five positive pairs plus spelling and vowel boundary negatives. Correct rounds award 200; repeated round-option error signatures dedupe and the second distinct error deadlocks. Invalid and terminal actions are atomic.

Frequently asked questions

Does matching spelling guarantee a rhyme?
No. The game compares CMUdict phonemes from the last stressed vowel through the word ending; rough and cough demonstrate why spelling alone can mislead.
Does this cover every English accent?
No. Pronunciation varies. The answers are limited to the cited CMUdict entries and are independently cross-checked with Cambridge word pages.
Can the page reveal the answer before submission?
No. Current options have neutral selection markup only; phonemes, shared tail, and correct markers appear only in a previous-pair note after success.

Related tools

Mini Games guides

View all