SPRAAK
|
Extract information from a corpus file. More...
Functions | |
void | spr_cor_info_free (SprSspInfo *Info) |
int | spr_cor_info_setup (SprSspInfo *Info, const char **descript, void *aux_info) |
int | spr_cor_info_process (SprSspInfo *Info, const void *frame_in, void *frame_out) |
void | spr_cor_info_reset (SprSspInfo *Info, SprSspStatus *action) |
Extract information from a corpus file.
This program inserts information read from a corpus file into the parameter stream. I.e. it finds the corpus entry that corresponds to the current data frame (usualy the same as for the previous data frame) and outputs one or more (extra) fields from that corpus entry (in numerical format). This way off-line collected information, e.g. silence/speech detection, speaker identity and so on, can be used in the on-line preprocessing. if a piece of data is covered by multiple corpus entries, a random entry is used.
[cor_info] | |
---|---|
corpus <fname> | |
Read information from the given corpus file. | |
field <nr> <hashF/hashI/float/int> [def_val](-1) | |
Get info from field <nr> (base 0). Numerical info can be converted to a float or int, other info is hashed to a unique number (starting from 0) and optionally converted to a float. | |
hash <fname> | |
Read a (string,value) file to use for the hashing in the preceding 'field ... hash? ...' line (otherwise, numbers are assigned according to when the string first occured). | |
report <new_file/default> [MSG/ERROR/WARNING/INFO/OFF](MSG) [lvl](20) | |
Report when a file is not covered by the corpus or when falling back to the default values (uncovered region). |