| 1 | ## -*- Mode: CPerl -*- | ||
| 2 | ## | ||
| 3 | ## File: DTA::CAB::Analyzer::Morph::Helsinki.pm | ||
| 4 | ## Author: Bryan Jurish <moocow@cpan.org> | ||
| 5 | ## Description: morphological analysis via Gfsm automata, for use with Helsinki-style transducers | ||
| 6 | ## + transducers available in HFST format from https://sourceforge.net/projects/hfst/files/resources/morphological-transducers/ | ||
| 7 | |||
| 8 | ##============================================================================== | ||
| 9 | ## Package: Analyzer::Morph | ||
| 10 | ##============================================================================== | ||
| 11 | package DTA::CAB::Analyzer::Morph::Helsinki; | ||
| 12 | use DTA::CAB::Analyzer::Automaton::Gfsm; | ||
| 13 | use Carp; | ||
| 14 | use strict; | ||
| 15 | our @ISA = qw(DTA::CAB::Analyzer::Automaton::Gfsm); | ||
| 16 | |||
| 17 | ## $obj = CLASS_OR_OBJ->new(%args) | ||
| 18 | ## + object structure: see DTA::CAB::Analyzer::Automaton::Gfsm | ||
| 19 | sub new { | ||
| 20 | my $that = shift; | ||
| 21 | my $aut = $that->SUPER::new( | ||
| 22 | ##-- defaults | ||
| 23 | |||
| 24 | ##-- analysis selection | ||
| 25 | label => 'morph', | ||
| 26 | wantAnalysisLo => 0, | ||
| 27 | wantAnalysisLemma => 0, ##-- default=0 | ||
| 28 | tolower => 1, | ||
| 29 | |||
| 30 | ##-- verbosity | ||
| 31 | check_symbols => 0, | ||
| 32 | |||
| 33 | ##-- Helsinki variant (e.g. 'de', 'en', ...) | ||
| 34 | helsinkiLang => '', | ||
| 35 | |||
| 36 | ##-- user args | ||
| 37 | @_ | ||
| 38 | ); | ||
| 39 | return $aut; | ||
| 40 | } | ||
| 41 | |||
| 42 | ##============================================================================== | ||
| 43 | ## Methods: Analysis: v1.x | ||
| 44 | ##============================================================================== | ||
| 45 | |||
| 46 | ## $doc = $anl->analyzeTypes($doc,\%types,\%opts) | ||
| 47 | ## + perform type-wise analysis of all (text) types in %types (= %{$doc->{types}}) | ||
| 48 | sub analyzeTypes { | ||
| 49 | my ($aut,$doc,$types,$opts) = @_; | ||
| 50 | return if (!$aut->SUPER::analyzeTypes($doc,$types,$opts)); | ||
| 51 | |||
| 52 | ##-- post-process: simulate TAGH-notation | ||
| 53 | my $label = $aut->{label}; | ||
| 54 | my $null = []; | ||
| 55 | my ($w,$a,$hi,$tag,$lemma); | ||
| 56 | foreach $w (values %$types) { | ||
| 57 | foreach $a (@{$w->{$label}//$null}) { | ||
| 58 | $hi = $a->{hi}; | ||
| 59 | if ($hi =~ /((?:\\?\[[^\<\>\[\]\/\\]+\\?\])+)$/) { | ||
| 60 | $tag = $1; | ||
| 61 | $lemma = substr($hi, 0, length($hi)-length($tag)); | ||
| 62 | $tag = join('.', ($tag =~ /\[\+?([^\<\>\[\]\/\\]+)\\?\]/g)); | ||
| 63 | } | ||
| 64 | elsif ($hi =~ /(\[\+[^<>\[\]\/\\]+\](?:\[[^\]]*\]|\\.)*)$/) { | ||
| 65 | ##-- de_free: "Haus[<NN>]Mann[<NN>]Kost[+NN][<Fem>][<Akk>][<Sg>]", "laufen[+V][<3>][<Sg>][<Pres>][<Ind>]" | ||
| 66 | $tag = $1; | ||
| 67 | $lemma = substr($hi, 0, length($hi)-length($tag)); | ||
| 68 | $tag = join('.', ($tag =~ /\[\+?<?([^<>\[\]\/\\]+)\\?>?\]/g)); | ||
| 69 | } | ||
| 70 | elsif ($hi =~ /((?:\\?\[\<?[^\<\>\[\]\/\\]+\>?\\?\]))$/) { | ||
| 71 | $tag = $1; | ||
| 72 | $lemma = substr($hi, 0, length($hi)-length($tag)); | ||
| 73 | $tag =~ s/[\\\<\>\[\]\+]//g; | ||
| 74 | } | ||
| 75 | $lemma =~ s/(?:\[[^\+\]]*\]|\\)//g; | ||
| 76 | $lemma =~ s/\[([A-Z]+)\+\]/lc($1)."+"/eg; | ||
| 77 | $lemma =~ s/\[\+([A-Z]+)\]/"~".lc($1)/eg; | ||
| 78 | $a->{hi} = "$lemma\[_$tag]=$hi" if ($lemma || $tag); | ||
| 79 | } | ||
| 80 | } | ||
| 81 | |||
| 82 | return $doc; | ||
| 83 | } | ||
| 84 | |||
| 85 | 1; ##-- be happy | ||
| 86 | |||
| 87 | __END__ |