Fig 2. Model performance based on the overlap measure.

We tested the online syllable recognition accuracy of the model based on simulation results on 220 sentences (giving a total of about 3000 syllables). Accuracy was evaluated based on the overlap of the recognized syllable sequence and durations with those of the input sentence. Data for each model variant is represented by the color of the outlines on the left panel. The figure shows the mean performance and 95% confidence interval for each frequency value of precision units. Diagrams on the left indicate the main functional groups of the model: θ corresponds to the theta-module, γ and ⍵ to syllable and gamma units respectively. Arrows indicate connections between functional groups (θ → γ represents rate and onset information from theta module to gamma units, whereas γ → ⍵ indicates the reset of accumulated evidence by the last gamma unit). π represents precision units, and the arrows originating from it indicate which functional groups they control.