Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Classifying several dialectal Nawatl varieties

Juan-José Guzmán-Landa, Juan-Manuel Torres-Moreno, Miguel Figueroa-Saavedra, Carlos-Emiliano González-Gallardo, Graham Ranger, Martha Lorena-Avendaño-Garrido

TL;DR

This study tackles automatic classification of Nawatl dialects, a low-resource language with extensive regional variation, by evaluating multiple classifiers on the π-yalli corpus. It compares a character $n$-gram baseline (TextCat) with neural models—SVM, CNN, LSTM, and a Siamese C-LSTM—across 18 dialectal varieties labeled in the corpus. The results show that the LSTM model achieves the best overall F1 score about $0.952$, significantly outperforming the baseline and matching the complex, highly variable orthography and morpho-syntax of Nawatl varieties. The work demonstrates the feasibility of dialect identification in Nahuatl and points to ensemble strategies and linguistic-rule augmentation as promising future directions to improve reliability and practical impact for NLP tools and language preservation.

Abstract

Mexico is a country with a large number of indigenous languages, among which the most widely spoken is Nawatl, with more than two million people currently speaking it (mainly in North and Central America). Despite its rich cultural heritage, which dates back to the 15th century, Nawatl is a language with few computer resources. The problem is compounded when it comes to its dialectal varieties, with approximately 30 varieties recognised, not counting the different spellings in the written forms of the language. In this research work, we addressed the problem of classifying Nawatl varieties using Machine Learning and Neural Networks.

Classifying several dialectal Nawatl varieties

TL;DR

This study tackles automatic classification of Nawatl dialects, a low-resource language with extensive regional variation, by evaluating multiple classifiers on the π-yalli corpus. It compares a character -gram baseline (TextCat) with neural models—SVM, CNN, LSTM, and a Siamese C-LSTM—across 18 dialectal varieties labeled in the corpus. The results show that the LSTM model achieves the best overall F1 score about , significantly outperforming the baseline and matching the complex, highly variable orthography and morpho-syntax of Nawatl varieties. The work demonstrates the feasibility of dialect identification in Nahuatl and points to ensemble strategies and linguistic-rule augmentation as promising future directions to improve reliability and practical impact for NLP tools and language preservation.

Abstract

Mexico is a country with a large number of indigenous languages, among which the most widely spoken is Nawatl, with more than two million people currently speaking it (mainly in North and Central America). Despite its rich cultural heritage, which dates back to the 15th century, Nawatl is a language with few computer resources. The problem is compounded when it comes to its dialectal varieties, with approximately 30 varieties recognised, not counting the different spellings in the written forms of the language. In this research work, we addressed the problem of classifying Nawatl varieties using Machine Learning and Neural Networks.
Paper Structure (14 sections, 5 figures, 4 tables)

This paper contains 14 sections, 5 figures, 4 tables.

Table of Contents

  1. Introduction
  2. Dialectal varieties of Nawatl
  3. Nawatl corpora
  4. Corpus characterization
  5. Pre-processing
  6. Classification algorithms
  7. Baseline: $n$-grams of characters
  8. Word embeddings
  9. Support Vector Machines
  10. Convolutional Neural Networks
  11. Long Short-Term Memory
  12. Siamese Networks
  13. Results
  14. Conclusions and Future Works

Figures (5)

  • Figure 1: Number of documents concerning the 25 Nawatl varieties from $\pi$-yalli corpus. The varieties H, IST, OAX, CEO SOP, ORP, TLA, SNSP, CEB, ANP, SNP were excluded from the graph because they have fewer than three documents.
  • Figure 2: Tokens' number concerning the $k$=18 most representative varieties Nawatl from $\pi$-yalli corpus.
  • Figure 3: Average ratio between the number of characters and tokens in the Nawatl varieties of the $\pi$-yalli corpus. A token is an element divided between two spaces, but in Nawatl, the token can be a sentence thanks to agglutination.
  • Figure 4: Token length in the $\pi$-yalli corpus.
  • Figure 5: LSTM: Confusion matrix.