hfst
hfst copied to clipboard
hfst
Helsinki Finite-State Technology (library and application suite)
I am building a fst using regexp and multiple composition operations. For the final fst, `hfst-lookup` can find paths yet `hfst-proc` can't find any valid path. The idea of the...
I'm trying to build a toy guesser: ``` $❯ cat guesser.lexc Multichar_Symbols +n +a +sg +nom +ine +ela +all LEXICON Root abcdefghijklmonpqrstuvxyzåäö # ; Guess ; LEXICON Guess < ?+...
Iäm trying to build a toy guesser: ``` $❯ cat guesser.lexc Multichar_Symbols +n +a +sg +nom +ine +ela +all LEXICON Root abcdefghijklmonpqrstuvxyzåäö # ; Guess ; LEXICON Guess < ?+...
While trying to solve #431, I found that when I compile [this twolc file](https://raw.githubusercontent.com/reynoldsnlp/rus_graph2phone/testing-data/scripts/g2p.twolc), (and every other `twolc` file I have tried), the resulting transducer is not equivalent to a...
I was sampling some of my automata with hfst-fst2strings and it seems to always print zero weights in the beginning of the list and in general increasing weights with -r...
Cf this command and its output: ``` $ echo '?* k:j ?*' | hfst-regexp2fst | hfst-fst2txt 0 0 k k 0.000000 0 0 j j 0.000000 0 0 @_IDENTITY_SYMBOL_@ @_IDENTITY_SYMBOL_@...
If the input fst to hfst-reweight is of an unweighed fst type (say, foma), it processes the reweighting and spits out an fst of the same, unweighted type. This does...
``` $ hfst-lexc -V hfst-lexc 0.1 (hfst 3.15.0) ``` The following x.lexc lexicon: ``` Definitions Co = [k|n|p] ; Vo = [a|e|o|u|y] ; VoB = [a|o|u] ; ! Front vowel...
I learned recently that Giellatekno has been moving toward a new tokenization/analysis pipeline as outlined [here](http://giellatekno.uit.no/doc/ling/preprocessor.html#N10014). That page suggests the following approach: ```bash cat testfile.txt | \ hfst-tokenize --giella-cg --weight-classes=1...
It would be nice if HFST were easily installable on MacOS via homebrew.
