|
|
|
| PHONE1 | PHONE1 files were collected between december 1995 and july 1996 by
means of an automatic system which performed automatic calls to previously
advised speakers. In total 281 speakers were collected, for a total
of 5210 files (including silences). Each speaker has 1 to 64 files
(including silences); 223 speakers have more than 15 files.
Files are divided into higQ (3631 files), medQ (1191) and lowQ (388), according to their content (presence/absence of spontaneous speech phenomena, out of vocabulary words, etc.). This subdivision is indicative, as some files were re-labeled in successive phases. 1652 acoustic phonetically rich sentences
See also a tech report which describes this database, and a paper containing some results. |
| FIELD1 | FIELD1 files were collected on december 1996 in the Exchange of
Acilia, Roma. Files are divided in the following classes, according to the state of the system during which they were collected. Their content may be different. 15 digit isolated digits
See also a page and a paper describing FIELD1,2,3 (and containing detailed information about the real number of usable data) and reporting results. |
| FIELD2 | FIELD2 files were collected on january 1997 in the Exchange of Acilia,
Roma.
Files are divided in the following classes, according to the state of the system during which they were collected. Their content may be different. 24 digit isolated digits
|
| FIELD3 | FIELD3 files were collected from february 13 to february 26, 1997,
in the Exchange of Acilia, Roma.
Files are divided in the following classes, according to the state of the system during which they were collected. Their content may be different. 1499 loopdigit continuous digit sequences
Other files are available (348 digit + 3689 yes-no-a), but not yet transcribed. |
| DEMO1 | DEMO1 files were collected on february 1997 at IRST (internal demo).
Most of the calls were performed by people working at IRST.
Files are divided in the following classes, according to the state of
the demo during which they were collected. Their content may be
133 surnames surnames of people working at ITC/IRST
|
| PHONE2 | PHONE2 files were collected in august 1997 by means of an automatic
system which performed automatic calls to previously advised speakers.
In total 269 speakers were collected, for a total of 4034 files (including
silences). Each speaker has 2 to 37 files (including silences); 121
speakers have more than 15 files.
Files are divided into higQ (3669 files), medQ (122) and lowQ (243), according to their content (presence/absence of spontaneous speech phenomena, out of vocabulary words, etc.). This subdivision is only indicative. 1356 acoustic phonetically rich sentences
|
| FIELD4 | FIELD4 files were collected from october 14 to october 16, 1997, in
the Exchange of Acilia, Roma.
Files are divided in the following classes, according to the state of the system during which they were collected. Their content may be different. 419 loopdigit continuous digit sequences
|
| CARITRO | dec '97 / may '98 field data under collection ... see this page for a description of the application. |
Last update 1/6/1998 - Maintainer Roberto
Gretter