How I built a rhyme quality scoring algorithm for the French language

2026-04-16 · Source: Naturallanguageprocessing on Medium · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Natural Language Processing · Depth: Advanced, medium

Summary

A new algorithm has been developed to score rhyme quality in French, addressing the limitations of existing tools that only provide simple rhyme lists. This system, integrated into the SemiRimeS platform, quantifies rhyme richness by analyzing phonemic proximity rather than letter-based matching. It converts words into International Phonetic Alphabet (IPA) sequences using the Lexique database, then aligns them based on the last stressed vowel and subsequent consonants. The core innovation involves two phonemic proximity matrices (one for 16 vowels, one for 20 consonants) that score similarity from 1 to 9 based on articulatory features. A final score, ranging up to 999.9, is computed by assigning a base score of 900 for identified rhymes and adding weighted scores for preceding phonemes, allowing for nuanced classification into "poor," "sufficient," "rich," and "very rich" categories. This unified scoring model applies to rhymes, assonances, and consonantal rhymes, providing a single comparable quality indicator.

Key takeaway

For NLP engineers or computational linguists developing language-specific tools, you should consider adopting a phoneme-centric approach for tasks like rhyme analysis. Your systems will achieve greater accuracy and nuance by moving beyond letter-based comparisons and incorporating articulatory feature-based phonemic proximity matrices, especially for languages with complex orthography-to-pronunciation mappings like French. This method allows for a more granular and linguistically sound assessment of phonetic similarity.

Key insights

Quantifying rhyme quality requires phonemic analysis and articulatory proximity matrices, not just letter matching.

Principles

• French linguistic tools must operate at the phoneme level.
• Articulatory features define phoneme similarity.
• Positional weighting enhances rhyme scoring accuracy.

Method

The method involves converting words to IPA phonemes, aligning them by the last vowel, and then using articulatory-feature-based proximity matrices for vowels and consonants to compute a weighted score for rhyme quality.

In practice

• Use Lexique database for French phonemic transcriptions.
• Develop separate proximity matrices for vowels and consonants.
• Apply decreasing positional weights to phonemes preceding the rhyme.

Topics

• Rhyme Quality Scoring
• French Phonology
• Phonemic Proximity Matrices
• SemiRimeS Platform
• Computational Poetry

Best for: NLP Engineer, AI Scientist, Research Scientist

Related on AIssential

• Training Data Size Sensitivity in Unsupervised Rhyme Recognition — RhymeTagger excels at unsupervised rhyme recognition, outperforming humans and LLMs when adequately trained.
• Phun-Bench: Evaluating LLMs on Phonological Understanding in Chinese — LLMs struggle with flexible phonological understanding despite recalling pronunciations, indicating a research gap.
• Automatic Metrical Scansion of Galician Poetry: First Results — A new system provides automatic metrical scansion for Galician poetry with 88% accuracy.
• Token-Level Pun Location Using Multi-Layer BERT with Mixture of Experts — A two-stage BERT-MoE model effectively locates puns in Portuguese by specializing expert components.
• AfriSUD: A Dependency Treebank Collection for Evaluating Models on African Languages — Models exhibit a significant "syntax gap" in processing African languages, struggling with their structural diversity.
• NLP part 2 — Edit distance quantifies string differences for spelling correction, with advanced variations improving accuracy.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Naturallanguageprocessing on Medium.