cos.rb

# encoding: utf-8

require 'natto'
require 'pp'
require 'matrix'

@natto = Natto::MeCab.new

def simcos(v1, v2)
  v1.inner_product(v2).to_f / (v1.norm * v2.norm)
end

def to_bigrams(sent)
  words = []
  @natto.parse(sent) { |n| words << n.surface }
  bigrams = []
  words[0..-3].each_with_index do |word, i|
    bigrams << [words[i], words[i+1]].join('')
  end
  bigrams
end

all_bigrams =[] 
File::open("all_bigram.csv") {|f|　# 全ての2-gramファイル
  f.each {|line| all_bigrams << line.chomp}
}

vecs =[]
File::open("vecs.csv") {|f| # ベクトル
  f.each {|line| vecs << line.chomp.split(',')}
}

sents = []
File::open("amami_n.csv") {|f| # 元コーパス
  f.each {|line| sents << line.chomp.split('|||')[1]}
}

test_case = []
loop do
  input = gets
  break if input.nil?
  test_case << input.chomp
end
test_case.delete_if{|v|v == ''}

test_case.each do |sent|
  print sent

  dims = []
  bag = to_bigrams(sent)

  bag.each do |bigram|
    all_bigrams.each_with_index do |v, i|
      if bigram == v
        dims << i

      end
    end
  end

  vecsent = Array.new(40355, 0)
  dims.each do |d|
    vecsent[d] = 1
  end
  vecsent = Vector.elements(vecsent)

  nvecs = []
  vecs.each do |vec|
    vec.each_with_index do |v, i|
      if i % 2 == 1 and dims.include? v.to_i
        nvecs << vec
        break
      end
    end
  end

  ret = nvecs.map do |vec|
    nvec = Array.new(40355, 0)
    vec.each_with_index do |v, i|
      nvec[v.to_i] = vec[i-1].to_f if i % 2 == 1
    end
    [simcos(Vector.elements(nvec), vecsent), sents[vecs.index(vec)]]
  end
  print " - "
  puts ret.sort_by{|v|v[0]}.map{|v|v[1]}[0]


end