Navajo (Athabaskan)

Navajo is mentioned in our paper in passing in the discussion of the nature of the learning data. We wanted to look at the language because it has a large inventory of affricates, both strident and non-strident, with laryngeal contrasts. The inevitable frequency differences between these sounds make it a tricky case for our learner, and indeed the results vary depending on the dataset we use. The learner comes the closest to the traditional phonological analyses of Navajo (e.g., Harry Hoijer's) when trained on the most curated dataset is the roots/stems list.

Simulation data at a glance

Click on simulation name to view additional simulation details.

Simulation name	Initial state Learning Data	Initial state features
Stems	LearningData.txt	Features.txt
Web_Words	LearningData.txt	Features.txt
Young_Morgan_Dictionary	LearningData.txt	Features.txt

---

Simulation details for Navajo stems

Summary of inventory changes

Stage	Consonant set
Input	b d ɟ g ɮ p t c k t' k' ʃ' s' ɬ' z s ʃ ʒ h x ɣ ɮ ɬ m n w j ʔ
Output	b d g ɮ t k t' k' z s ʃ ʒ h x ɣ ɮ ɬ m n w j ʔ dʒ tʃ' ts' tɬ' ts tʃ dɮ dz tɬ

---

Simulation details for Navajo web_words

Summary of inventory changes

Stage	Consonant set
Input	b d g ɮ p t k t' k' ʃ' s' ɬ' z s ʃ ʒ h x ɣ ɮ ɬ m n w j ʔ
Output	b d g ɮ p t k t' k' ʃ' ɬ' z s ʃ ʒ h x ɣ ɮ ɬ m n w j ʔ dʒ ts'

---

Simulation details for Navajo young_morgan_dictionary

Summary of inventory changes

Stage	Consonant set
Input	b d ɟ g ɮ p t c k t' k' ʃ' s' ɬ' z s ʃ ʒ h x ɣ ɮ ɬ m n w j ʔ
Output	b d g ɮ t k t' k' z s ʃ ʒ h x ɣ ɮ ɬ m n w j ʔ ɟʒ ts' tɬ' cʃ' ts cʃ

View the source code on GitHub