La machine à inventer des mots (avec Code MU)

La machine à inventer des mots (avec Code MU)

The word-inventing machine (with Code MU)

Title history
  • La machine à inventer des mots (avec Code MU) — observed on 20261019-0426 UTC (PSI archive)

by David Louapre · View the channel on PSI

Formal & Physical Sciences AI — Models & Research UYQArtificial intelligence (AI)UYQFGenerative artificial intelligence
🎙 David Louapre 👥 1.5M 📅 October 16, 2015 ⏱ 11 min 👁 452K 📄 science communication 🧭 2026-10-19
Available in: English (current) Français

Keywords

Markov chainprobabilityword generationstatisticslanguage modeling

Summary

This video presents a method for generating new words that sound plausible in a given language, using statistical analysis and Markov chains. The creator, David Louapre, explains how he compiled a list of 300,000 French words from Project Gutenberg and computed the frequencies of letter transitions. He then uses these probabilities to generate new words by sequentially picking the next letter according to the transition probabilities. The results are words that often resemble real French words but are invented, such as ‘angeries’ or ‘gazoïexcombre’. The video demonstrates that such generated words can be used in creative writing, culminating in a poem composed of these invented words mixed with common French terms. The method is computational and reproducible, with code available on GitHub. The video also highlights interesting linguistic oddities, such as exceptions to common rules like ‘q’ always followed by ‘u’. Overall, it is a fun and insightful application of probability to language.

153 words

Critical Evaluation

Value of the Information & Strength of the Argument

The video provides a clear and engaging exploration of how statistical methods can be applied to language. The argumentation is sound: it starts from the observation that random letters do not produce plausible words, then introduces the concept of letter transition probabilities, and demonstrates that using a Markov chain yields words that mimic the statistical structure of French. The method is validated by examples and comparisons with German, and the creator provides a reproducible implementation. The logical progression and concrete illustrations make the case compelling.

Scientific Rigor, Source Quality, Title Accuracy

The scientific rigor is high: the video is based on a large word list (300,000 words) from Project Gutenberg, a well-known digital library. The algorithm is described with enough detail to be replicated, and the code is shared on GitHub. The title accurately reflects the content, and the video credits Code MU for inspiration. The presentation is honest about the limitations (e.g., some generated words coincide with existing ones). No major discrepancies between title and content are detected. The video does not cite external scientific papers, but the underlying concepts (Markov chains, statistical n-grams) are standard and well-established.

198 words

Title / Content Match

The title accurately reflects the content: a machine (algorithm) that invents words, with a collaboration with Code MU.

Quality & Reliability

9/10

The video clearly explains the statistical method and provides reproducible code on GitHub. It correctly references Project Gutenberg and uses Markov chains appropriately. The explanation is accurate and the claims are verifiable.

Key Moments

Markers derived by PSI from the transcript: the creator did not define chapters.

Cited Sources

Contribution & Novelties

The video’s original contribution lies in making Markov chains accessible to a general audience for creative word generation. It demonstrates a practical and entertaining application of probability theory to language, accompanied by reproducible code. It also highlights peculiarities of the French language through the lens of statistical data, such as rare transitions and exceptions to rules.

Pour aller plus loin :

  • Markov chain — Provides the mathematical foundation for the method used in the video.
  • Project Gutenberg — Source of the extensive word list used to compute transition probabilities.
  • Oulipo — A literary movement that uses constrained writing techniques, similar in spirit to the poem generated in the video.

109 words

Radar Profile

The radar profile shows high scores in information quality and quantity, reflecting a well-structured and informative video. The technical level is moderate, suitable for a general audience, while the global reliability is strong due to transparent methodology and open-source code.

Reliability 8/10

💬 Sur les 30 commentaires analysés, le climat est très positif et enthousiaste, avec de nombreux jeux de mots et anecdotes humoristiques sur les mots générés. Les spectateurs expriment leur amusement et leur intérêt pour l'approche statistique.