Linguistic Feature Tagging for Automatic Classification of 27 Closely-Related Quechua Varieties

2026-06-25 · Source: Paper Index on ACL Anthology · Field: Technology & Digital — Artificial Intelligence & Machine Learning · Depth: Expert, quick

Summary

Claire Post and Alexis Palmer's 2026 paper introduces a multi-dialect text classifier designed for 27 closely-related Quechua varieties. This system enhances neural models with rule-based linguistic features to tackle challenges inherent in low-resource, morphologically complex languages. The methodology leverages a carefully curated dataset, encompassing various genres like annotated parallel bible corpora, and explicitly encodes manually annotated lexical variation and polypersonal verbal agreement as features within a transformer-based classifier. Experimental results demonstrate that these neural models substantially outperform statistical baselines, achieving high accuracy in multi-class classification across the 27 Quechua dialects. The impact of linguistic augmentation is context-dependent, showing minimal gains in high-resource scenarios but more pronounced benefits in low-resource and cross-domain conditions. This research contributes to developing dialect-sensitive NLP methods for Quechua and other similar languages.

Key takeaway

For NLP Engineers developing solutions for low-resource, morphologically rich languages like Quechua, you should consider augmenting transformer-based models with explicit linguistic features. This approach significantly improves multi-class classification accuracy, especially in low-resource and cross-domain scenarios, outperforming statistical baselines. Prioritize manually annotating lexical variation and verbal agreement to enhance model performance.

Key insights

Augmenting neural models with linguistic features improves low-resource, morphologically complex language classification.

Principles

• Linguistic augmentation benefits low-resource NLP more than high-resource.
• Explicitly encoding morphological features aids dialect classification.

Method

A transformer-based classifier is augmented with rule-based linguistic features, specifically manually annotated lexical variation and polypersonal verbal agreement, trained on a multi-genre dataset including parallel bible corpora.

In practice

• Integrate rule-based linguistic features into neural models for low-resource NLP.
• Curate diverse datasets, including parallel corpora, for dialect classification.

Topics

• Quechua dialects
• Low-resource NLP
• Linguistic feature tagging
• Transformer models
• Dialect classification
• Morphological complexity

Best for: Research Scientist, AI Scientist, NLP Engineer

Related on AIssential

• Low-resource Language Discrimination Towards Chinese Dialects with Transfer learning and Data Augmentation — Transfer learning and data augmentation effectively overcome low-resource challenges in Chinese dialect discrimination.
• An End-to-End Hybrid Framework for Rumour Detection in Low-Resources Algerian Dialect — A hybrid framework combining transformer embeddings with classical classifiers enables effective rumour detection in low-resource Algerian dialect.
• TalkTag: Fine-Grained Morphosyntactic Error Annotation for Transcribed Speech — TalkTag automates fine-grained morphosyntactic error annotation in spoken transcripts using an LLM, even with scarce data.
• Towards interpretable models for language proficiency assessment: Predicting the CEFR level of Estonian learner texts — Careful feature selection yields interpretable, generalizable models for language proficiency assessment with high accuracy.
• Cross-lingual Relation Extraction with Large Language Models: Zero-Shot, Few-Shot, and Fine-Tuned Evaluation on Romanian — QLoRA fine-tuning significantly boosts cross-lingual relation extraction performance for low-resource languages, outperforming prompt-only LLMs.
• Robust Language Identification for Romansh Varieties — The paper titled "Robust Language Identification for Romansh Varieties," authored by Charlotte Model, Sina Ahmadi, and Jannis Vamvas, was presented at the 11th Edition of the Swiss Text Analytics Conf…

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Paper Index on ACL Anthology.