Table 1.
Variables chosen for dialect prediction (MHG = Middle High German) according to [29], examples, numbers, and types of Swiss German variables; adapted from [17].
Fig 1.
Variants of schneien ‘to snow’ (left), Bett ‘bed’ (center), and the four-quadrant intersection of the two maps (right; maps adapted from [28]).
Fig 2.
Pronunciation variants for fragen ‘to ask’ (left), dialect prediction displayed as a list of five best hits (center) and as pins on a map (right).
Fig 3.
Evaluation of prediction: if satisfied, users are prompted to participate in research (left) by indicating their age and gender (center left).
If dissatisfied, users select their locality from a drop-down menu (center right) and then indicate age and gender (right).
Fig 4.
Swiss German-speaking population per canton (left), number of users per canton (center), and user percentage per canton (right).
Fig 5.
Number of respondents by locality.
Each Thiessen polygon represents one locality. The larger the black dot, the more respondents per locality. Polygons are based on Swiss commune centroids derived from generalized commune boundaries available from Swiss Federal Statistics Office (SFSO), Swiss Statistics Website.
Fig 6.
Agreement with Atlas by variable type (left) and for each phonetic variable (right).
Fig 7.
Distribution of variants for Apfelüberrest (Atlas).
Fig 8.
Distribution of variants for Apfelüberrest (DÄ).
Fig 9.
Distribution of variants for heben (Atlas).
Fig 10.
Distribution of variants for heben (DÄ).
Fig 11.
Distribution of vocalized variants in Kelle.
Light green denotes areas where the Atlas had documented vocalization and DÄ shows the same result. Colors other than light green show regions where nowadays DÄ shows vocalization, but the Atlas did not.
Fig 12.
Atlas agreement scores by variable. The darker the purple, the higher the agreement with the Atlas.
Fig 13.
Scatterplot matrix of number of levels per variable as a function of Atlas agreement score.
Fig 14.
Distribution of number of speakers per age group (left); age group by agreement scores (right).
Groups 1–10 as well as 91–100 and 101–110 are not displayed.
Fig 15.
Scatterplot matrix of the degree of vocalization as captured by the traditional method as a function of the degree of vocalization as measured by crowdsourcing.