What's new in SPRAAK V1.0

Migrating from HMM75 to SPRAAK V1.0 involves a few steps. For most users the required adaptations will be moderate.

Fileformats

SPRAAK V1.0 uses files with the '.spr' header format, while previous versions of HMM7.x (and pre-releases of SPRAAK till V0.9) used the '.key' header format. The change was made to get rid of a number of inconsistencies that had gotten into the 20yr old header framework. Old HMM7.x files with '.key' headers are still readable in SPRAAK, so they can still be used as long as the task does not require any novel features. However we advise everyone to gradually merge to the new format as there will be no fixes or adapatations for the old format.

The new fileformat looks very similar to the old one and should as such pose no problems. The main difference lies in a change of name of some of the header keys, in order to make their usage more consistent throughout the package. It is possible to simply convert 'old' to 'new' data by:

spr_copy -i <ifname> -o <ofname>

Selecting parts of the data - while converting - is only possible for certain kinds of data such as SAMPLE and TRACK files.

In previous versions only the '.hmm'-files had a version number (HMM_VNUM=7 in the last releases). From now on, file formats and file types will be managed by a single version key in the header 'VNUM' which will be applicable to all files. For SPRAAK V1.0 VNUM is set to '8'.

SPRAAK V1.0 provides better support for other formats including .wav-files, raw(headerless) dat in binary and ASCII, compressed and encrypted datastreams, .. See Data IO in SPRAAK for more details.

Standalone Programs

Using standalone programs will require a few minor adapations from the user, but these are easy to get used to:

• program names have been maintained, except for the added prefix "spr_"
• program arguments have changed somewhat, but are - if anything - more consistent
• all programs share a set of common global options including (see also spr_od_common)
  • online help
  • retrieving of version information
  • common ways for remote executrion
  • debugging

Scripts & Programming

Internally in the code the changes have been more substantial. Hence people writing programs and/or scripts will encounter significant differences. As scripting language we made the shift from CSH to Python. This opens many more possibilities, but may indeed involve some learning for the user. Equally so, the usage of object oriented programming concepts, may pose some challenges.

Manual Pages

SPRAAK V1.0 manual pages are now available online. The manual pages of the most recent release is available through www.spraak.org. Pls. check with your local SPRAAK responsible for the location of the manual pages that fit your installed version. The manual that was built using DOXYGEN. A major effort went into rewriting most of the manual pages. The manual, today, is far more complete than what it has been for years. However, gaps are surely still remaining. Any comments are highly welcomed.

Missing Legacy code

Some utilities, previously available in HMM 7.x, may not be supported currently in SPRAAK because conversion was not trivial. If you need any of these, pls. contact the SPRAAK development team.

Appendix: Key mappings between '.key' and '.spr'

There are two major changes:

1. The data and formatting is now described by 3 keys {FORMAT, TYPE and DATA} in which {FORMAT and TYPE} define 'how' the data is stored and in which 'DATA' specifies the content.
2. The many different names for 'number of items' (NFR, NSEG, NPARAM, NENTRY, ... ) were reduced to 2 parameters, i.e. DIM1 and DIM2 with following generic interpretation.
   • for MATRIX style layouts, DIM1 is 'the number of records' and DIM2 is 'the dimension of each record'
   • for variable length layouts, DIM1 gives 'the total number of records (lines in the STRING datatypes) and DIM2 gives the number of variable length items.

Exact mappings are not possible in general as the usage of certain keys in the old .key format was sometimes erratic.

.key                    .spr
{DATATYPE, PARAMTYPE}   DATA
DATAFORMAT              {TYPE, FORMAT}
NENTRY                  DIM1
NFR                     DIM1
NSEG                    DIM2 
NPARAM                  DIM2 (if DIM1 already specified)
HMM_VNUM                VNUM
...

for more details cfr. File formats and headers