Table 1
Data quality measures.
| MEASURE | SFUSED ENGLISH |
|---|---|
| Percentage of sound errors | 50.78% |
| Percentage of exchange errors (all types) | 1.2% (cf. 0.57% offline) |
| Percentage phonotactic violations | 5.46% |
| Minutes per error (MPE) | 2.43 |
| Estimated frequency of errors | 48.5 s (or less) |
Figure 1
Longform interface of phonological substitution error.
Table 2
SFUSED Mark-up in longform entries.
| / | Prefixed to error word or phrase |
| # | Suffixed to the end of a multi-word error term |
| ^ | Prefixed to source words |
| $ | Prefixed to trigger word in deletions |
| … | Trailing speech at beginning or end of longform entry |
| = | Suffixed to clipped words (i.e., not completed) |
| A:, B: | Used for speaker turns with more than one speaker |
| xxx | An interruption in the speech stream |
| , | Pause without interrupting speech |