Class: Brill::Tagger

Inherits:
Object
  • Object
show all
Defined in:
lib/brill/tagger.rb

Instance Method Summary collapse

Constructor Details

#initialize(lexicon = nil, lexical_rules = nil, contextual_rules = nil) ⇒ Tagger

will use the brown corpus as the default



10
11
12
13
14
15
16
17
18
19
 
# File 'lib/brill/tagger.rb', line 10

def initialize( lexicon = nil, lexical_rules = nil, contextual_rules = nil)
  @tagger = ::Tagger::BrillTagger.new
  lexicon ||= File.join(File.dirname(__FILE__),"brown","LEXICON")
  lexical_rules ||= File.join(File.dirname(__FILE__),"brown","LEXICALRULEFILE")
  contextual_rules ||= File.join(File.dirname(__FILE__),"brown","CONTEXTUALRULEFILE")

  Brill::Tagger.load_lexicon(@tagger, lexicon )
  Brill::Tagger.load_lexical_rules(@tagger, lexical_rules )
  Brill::Tagger.load_contextual_rules(@tagger, contextual_rules )
end

Instance Method Details

#adjectives(text) ⇒ Object

given a body of text return a list of adjectives



22
23
24
 
# File 'lib/brill/tagger.rb', line 22

def adjectives( text )
  tag(text).select{|t| t.last == 'JJ' }
end

#noun_phrases(text) ⇒ Object

see: cpansearch.perl.org/src/ACOBURN/Lingua-EN-Tagger-0.15/Tagger.pm



32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
 
# File 'lib/brill/tagger.rb', line 32

def noun_phrases(text)
  # ?:$PREP|$DET|$NUM)
  #
  tags = tag(text.gsub(/[^\w]/,' '))
  phrases = []
  phrase = []
  mark = -1

  tags.each_with_index do|tag,i|
    if phrase.empty?
      mark = i if tag.last.match(/PRP\$|DT/)
      if tag.last == 'NNP' and mark != -1
        phrase = tags[mark..i]
        #mark = -1
      end
      mark = -1 if i - mark > 8
    elsif tag.last.match(/NN/)
      phrase += tag
    else
      phrases << phrase.flatten
      phrase = []
      mark = -1
    end
  end
  phrases
end

#nouns(text) ⇒ Object

given a body of text return a list of nouns



27
28
29
 
# File 'lib/brill/tagger.rb', line 27

def nouns( text )
  tag(text).select{|t| t.last.match(/NN/) }
end

#suggest(text, max = 10) ⇒ Object

returns similar results as tag, but further reduced by only selecting nouns



60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
 
# File 'lib/brill/tagger.rb', line 60

def suggest( text, max = 10 )
  tags = tag(text)
  #puts tags.inspect
  ptag = [nil,nil]
  # join NNP's together for names
  reduced_tags = []
  mappings = {} # keep a mapping of the joined words to expand
  tags.each{|tag|
    if ptag.last == 'NNP' and tag.last == 'NNP' and !ptag.first.match(/\.$/)
      ptag[0] += " " + tag.first
      # before combining these two create a mapping for each word to each word
      words = ptag.first.split(/\s/)
      i = 0
      #puts words.inspect
      until (i + 1) == words.size
        mappings[words[i]] = ptag.first
        mappings[words[i+1]] = ptag.first
        i += 1
      end
      #puts mappings.inspect
    elsif tag.last == 'NNP'
      ptag = tag
    elsif tag.last != 'NNP' and ptag.first != nil
      reduced_tags << ptag
      reduced_tags << tag if tag.last.match( /^\w+$/ ) and tag.first.match(/^\w+$/)
      ptag = [nil,nil]
    elsif tag.last.match( /^\w+$/ ) and tag.first.match(/^\w+$/)
      reduced_tags << tag
    end
  }
  # now expand any NNP that appear
  tags = reduced_tags.map{|tag|
    if tag.last == 'NNP'
      #puts "#{tag.first} => #{mappings[tag.first]}"
      tag[0] = mappings[tag.first] if mappings.key?(tag.first)
    end
    tag
  }
  results = tags.select{|tag| tag.last.match(/NN/) and tag.first.size > 3 }
  if results.size > max
    counts = {}
    tags = []
    results.each {|tag| counts[tag.first] = 0 }
    results.each do |tag|
      tags << tag if counts[tag.first] == 0
      counts[tag.first] += tag.last == 'NNP' ? 3 : (tag.last == 'NNS' ? 2 : 1)
    end
    tags.map!{|tag| [tag.first, tag.last,counts[tag.first]]}
    t = 1
    until tags.size <= max
      tags = tags.sort_by{|tag| tag.last}.select{|tag| tag.last > t }
      t += 1
      if t == 5
        tags = tags.reverse[0..max]
        break
      end
    end
    tags
  else
    results
  end
end

#tag(text) ⇒ Object

Tag a body of text returns an array like [[token,tag],…[token,tag]]



126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
 
# File 'lib/brill/tagger.rb', line 126

def tag( text )
  # XXX: the list of contractions is much larger then this... find'em
  text = text.gsub(/dont/,"don't").gsub(/Dont/,"Don't")
  text = text.gsub(/youre/,"you're")
  tokens = Brill::Tagger.tokenize( text )
  tags = Brill::Tagger.tag_start( tokens )

  @tagger.apply_lexical_rules( tokens, tags, [], 0 )
  @tagger.default_tag_finish( tokens, tags )

  # Brill uses these fake "STAART" tags to delimit the start & end of sentence.
  tokens << "STAART"
  tokens << "STAART"
  tokens.unshift "STAART"
  tokens.unshift "STAART"
  tags << "STAART"
  tags << "STAART"
  tags.unshift "STAART"
  tags.unshift "STAART"

  @tagger.apply_contextual_rules( tokens, tags, 1 )

  tags.shift
  tags.shift
  tokens.shift
  tokens.shift
  tags.pop
  tags.pop
  tokens.pop
  tokens.pop

  pairs = []
  tokens.each_with_index do|t,i|
    pairs << [t,tags[i]]
  end
  pairs
end