Lingua::LTS

compiler/interpreter for festival-style letter-to-sound rules

LTS Rulesets

A number of festival LTS rule-sets and associated pre- and post-processing filters are available, mostly derived from the ims-german rule-set by Gregor Möhler. These rule-sets currently reside in the SVN directory svn+ssh://odo.dwds.de/dev/grimm/trunk/lts/.

By convention, rule-sets for the Lingua::LTS module carry the suffix ".lts". Filters are implemented as finite-state transducers which are composed with the "raw" rule-set transducer in the course of compilation. GNU Makefiles are provided for compiling the rulesets (with and/or without associated filters) to finite state automata. Use of these makefiles requires checking out the entire grimm/trunk/lts/ sub-tree however.

  • ims-german/
    includes the original unmodified ims-german rule-set (in festival SCHEME and in .lts format), the BOMP phonetic lexicon, and various filters derived from the ims-german phonetic post-processing SCHEME code.

  • grimm/
    variant of ims-german ruleset used by the Taxi::Mysql::Grimm* index classes. Includes some generally useful pre- and post-processing filters for mapping "strange" UTF-8 characters to latin-1 approximations (but see grimm/trunk/entities/grimm.entity-table.PL and the code in that directory for more).

  • ims-german-noh/
    a variant of ims-german with some rules for Middle- and Old High German, modified by Bryan Jurish and Wolfgang Seeker.

  • dta/
    simplified variant of the ims-german-noh ruleset for use by the DTA::CAB module (although that module can use any arbitrary LTS transducer). Optimized for speedy canonicalization of extinct graphematic forms (assumedly at the expense of phonetization accuracy).

  • ahd-nhd/
    actually a gfsmxl rewrite cascade (and not an LTS ruleset itself), but makes use of the orthographic filters in ims-german/ and grimm/.

  • lev+tagh/
    (abandoned) gfsmxl rewrite cascade composing a Levenshtein editor with the TAGH morphology.

  • phonident/
    gfsmxl phonetic identity cascade using the grimm ruleset and various filters, by Wolfgang Seeker.

  • syllable/
    rule-based syllable boundary detector by Wolfgang Seeker. requires fsm2.