Complex Segment Learner

Complex Segment Learner

Khalkha Mongolian (Mongolic)

Mongolian is not in the paper. The language is described by Svantesson et al. (2005, Phonology of Mongolian) as having the affricates [ts], [tsʰ], [tʃ], [tʃʰ]. It furthermore has a palatalization and an aspiration contrast for most of its obstruents, neither of which we not tackle here.
The learner was trained on an An Crúbadán corpus, transcribed from Cyrillic according to Svantesson et al. (See here). The learner does pretty well on this corpus, even though it is not the best data source: it identifies all affricates but the plain [ts].

Simulation data at a glance

Click on simulation name to view additional simulation details.

Simulation nameInitial state Learning DataInitial state features
only version LearningData.txt Features.txt

Simulation details for Mongolian

Input:

LearningData.txt | Features.txt

Summary of iterations:

IterationLearning Data producedFeatures producedInseparabilityNew Segments addedSegments removed
1 LearningData.txt Features.txt [download] [view] tʃʰ, tsʰ, tʃ None
2 No new learning data No new features [download] [view] None None

Summary of inventory changes

StageConsonant set
Inputj w wʲ r rʲ n nʲ m mʲ ɮ ɮʲ ŋ ɴ tʰ tʰʲ t tʲ ʃʰ sʰ x xʲ s ʃ p pʰ pʲ pʰʲ χ g ɢ gʲ
Outputj w wʲ r rʲ n nʲ m mʲ ɮ ɮʲ ŋ ɴ tʰ tʰʲ t tʲ ʃʰ sʰ x xʲ s ʃ p pʰ pʲ pʰʲ χ g ɢ gʲ tʃʰ tsʰ tʃ

Simulation Plots

/media/mongolian/simulation/insep_plots.png