SESAM workstation or UNIX
workstation with X11 (e.g. SUN)
Description:
The aim of this software is to provide the laboratories
involved
in the SAM project with a common and useful tool to assess semi-automatic labellers. It deals with the comparison
between manual
labelling of speech files (the reference) and semi-automatic
labelling of the same files (both in the PTS format with SAMPA
labels).
Both files must contain the same phonetic strings.
Therefore, the comparison can only be drawn in terms of
boundaries location or centre location shifts.
Included in this package are standard statistic tools, such as
histograms, signed means and the best 10 and worst 10 boundary
locations. It is possible to analyse either fine errors or gross
errors and results can be written in a text file for analysis or
printing. Confidence intervals are computed for a confidence
level of 95%.
Developing lab:
CRIN-CNRS & INRIA laboratories
B.P. 239
54506 Vandoeuvre-les-nancy CEDEX
France
