text_alignment 0.8.1 → 0.11.1

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: 0c49793b84e9ca5606a8fe9151530f6732337c2b92cfd1af1549a56ea3c2f39e
-  data.tar.gz: 4a0b8328d4c6de43af50bd8c278f83facef311c74516dc3a9a0c9dd5f91fbfc0
+  metadata.gz: ef59c0cd578ed453a67edeb3b29059f4b76c4c541f777fa35a06e76b299e2564
+  data.tar.gz: 5594c0f6eb1d52cc331c210fdf936e7cb09e30277f21933027e55b0c8cfa0e24
 SHA512:
-  metadata.gz: ad70de67b4a7b38290a59d89bed7cd9fa343ce3db62bb67b779e51b499929b9d7c4697871cf25017fa59a34c0525741d5da679559fc2f33d902fcda692f9f1ac
-  data.tar.gz: 6242b35cbb8f53effc477b508d428d9060cba2e93895496134963755cff956f3931411a87a7bc0d33b7b91459f1a27eb67c422897693320e1b2ceb06eff6e22b
+  metadata.gz: ef5bae492d7e7b840c00943ac24e571392c4c992a085c6a63c8cb6db126ed9137ba94f8cd7af42a91e17aa327cb9a5ce24b909944c157f21beb8a88c8ce7528b
+  data.tar.gz: b6d84dcdc35399f91a0f6a5a24d84b11169de9d861a249acca52f0fec86e6f1fbf1bb4dbb47a5d43654fa43735b181aa64034447fb71ca675090818dcf67133a

data/bin/align_annotations

@@ -26,8 +26,8 @@ def read_text(filename)
 	end
 end
 
-def align_denotations(denotations, source_text, target_text, debug = false)
-	alignment = TextAlignment::TextAlignment.new(source_text, target_text, denotations)
+def align_denotations(denotations, source_text, alignment, debug = false)
+	alignment.align(source_text, denotations)
 	new_denotations = alignment.transform_hdenotations(denotations)
 
 	if debug
@@ -50,8 +50,8 @@ def align_denotations(denotations, source_text, target_text, debug = false)
 	new_denotations
 end
 
-def align_mannotations(source_annotations, target_text, debug = false)
-	target_annotations = {text:target_text}
+def align_mannotations(source_annotations, reference_text, alignment, debug = false)
+	target_annotations = {text:reference_text}
 
 	idnum_denotations = 0
 	idnum_relations = 0
@@ -62,7 +62,8 @@ def align_mannotations(source_annotations, target_text, debug = false)
 		if annotations.has_key?(:denotations) && !annotations[:denotations].empty?
 			ididx = {}
 			warn "[#{i}]-=-=-=-=-"
-			denotations = align_denotations(annotations[:denotations], annotations[:text], target_text, debug)
+			denotations = align_denotations(annotations[:denotations], annotations[:text], alignment, debug)
+
 			denotations.each do |d|
 				reid = 'T' + (idnum_denotations += 1).to_s
 				ididx[d[:id]] = reid
@@ -76,7 +77,9 @@ def align_mannotations(source_annotations, target_text, debug = false)
 				annotations[:relations].each do |r|
 					reid = 'R' + (idnum_relations += 1).to_s
 					ididx[r[:id]] = reid
-					target_annotations[:relations] << r.dup.merge({id:reid, subj:ididx[r[:subj]], obj:ididx[r[:obj]]})
+					sid = ididx[r[:subj]]
+					oid = ididx[r[:obj]]
+					target_annotations[:relations] << r.dup.merge({id:reid, subj:sid, obj:oid}) unless sid.nil? || oid.nil?
 				end
 			end
 
@@ -85,7 +88,8 @@ def align_mannotations(source_annotations, target_text, debug = false)
 				annotations[:attributes].each do |a|
 					reid = 'A' + (idnum_attributes += 1).to_s
 					ididx[a[:id]] = reid
-					target_annotations[:attributes] << a.dup.merge({id:reid, subj:ididx[a[:subj]]})
+					sid = ididx[a[:subj]]
+					target_annotations[:attributes] << a.dup.merge({id:reid, subj:sid}) unless sid.nil?
 				end
 			end
 
@@ -94,7 +98,8 @@ def align_mannotations(source_annotations, target_text, debug = false)
 				annotations[:modifications].each do |m|
 					reid = 'M' + (idnum_modifications += 1).to_s
 					ididx[m[:id]] = reid
-					target_annotations[:modifications] << m.dup.merge({id:reid, obj:ididx[m[:obj]]})
+					oid = ididx[m[:obj]]
+					target_annotations[:modifications] << m.dup.merge({id:reid, obj:oid}) unless oid.nil?
 				end
 			end
 		end
@@ -109,13 +114,18 @@ unless ARGV.length == 2
 end
 
 source_annotations = read_annotations(ARGV[0])
-target_text = read_text(ARGV[1])
+reference_text = read_text(ARGV[1])
+
+alignment = TextAlignment::TextAlignment.new(reference_text, true)
 
 target_annotations = if source_annotations.class == Array
-	align_mannotations(source_annotations, target_text, false)
+	# align_mannotations(source_annotations, reference_text, alignment, true)
+	align_mannotations(source_annotations, reference_text, alignment, false)
 else
-	denotations = align_denotations(source_annotations[:denotations], source_annotations[:text], target_text, false)
-	source_annotations.merge({text:target_text, denotations:denotations})
+	# denotations = align_denotations(source_annotations[:denotations], source_annotations[:text], alignment)
+	denotations = align_denotations(source_annotations[:denotations], source_annotations[:text], alignment, false)
+	source_annotations.merge({text:reference_text, denotations:denotations})
 end
 
-# puts target_annotations.to_json
+# pp alignment.block_alignment
+puts target_annotations.to_json

data/lib/text_alignment/anchor_finder.rb

@@ -6,92 +6,141 @@ module TextAlignment; end unless defined? TextAlignment
 
 class TextAlignment::AnchorFinder
 
-	def initialize(source_str, target_str, _size_ngram = nil, _size_window = nil, _text_similiarity_threshold = nil)
-		@size_ngram  = _size_ngram  || TextAlignment::SIZE_NGRAM
-		@size_window = _size_window || TextAlignment::SIZE_WINDOW
-		@sim_threshold = _text_similiarity_threshold || TextAlignment::TEXT_SIMILARITY_THRESHOLD
+	def initialize(source_str, target_str, cultivation_map)
+		@s1 = source_str.downcase
+		@s2 = target_str.downcase
 
-		@reverse = (target_str.length < source_str.length)
+		@cultivation_map = cultivation_map
 
-		@s1, @s2 = if @reverse
-			[target_str.downcase, source_str.downcase]
-		else
-			[source_str.downcase, target_str.downcase]
-		end
+		@size_ngram  = TextAlignment::SIZE_NGRAM
+		@size_window = TextAlignment::SIZE_WINDOW
+		@sim_threshold = TextAlignment::TEXT_SIMILARITY_THRESHOLD
+		@pos_s1_final_possible_begin = @s1.length - @size_ngram - 1
+		@pos_s2_final_possible_end = @s2.length
 
-		# current position in s1
-		@beg_s1 = 0
-		@end_s1_prev = 0
-		@end_s2_prev = 0
+		# positions of last match
+		@pos_s1_last_match = 0
+		@pos_s2_last_match = 0
 	end
 
 	def get_next_anchor
-		# find the position of an anchor ngram in s1 and s2
-		while @beg_s1 < (@s1.length - @size_ngram)
-			if [' ', "\n", "\t"].include? @s1[@beg_s1]
-				@beg_s1 += 1
-				next
-			end
-			anchor = @s1[@beg_s1, @size_ngram]
-
-			# search_position = 0
-			search_position = @end_s2_prev
-			while @beg_s2 = @s2.index(anchor, search_position)
-				# if both the begining points are sufficiantly close to the end points of the last match
-				break if @beg_s1 > 0 && @beg_s2 > 0 && (@beg_s1 - @end_s1_prev < 5) && (@beg_s2 >= @end_s2_prev) && (@beg_s2 - @end_s2_prev < 5)
+		# To find the beginning positions of an anchor ngram in s1 and s2, beginning from the last positions matched
+		beg_s2 = for beg_s1 in @pos_s1_last_match .. @pos_s1_final_possible_begin
 
-				left_window_s1, left_window_s2 = get_left_windows
-				break if left_window_s1 && (text_similarity(left_window_s1, left_window_s2) > @sim_threshold)
+			# To skip whitespace letters
+			next if [' ', "\n", "\t"].include? @s1[beg_s1]
 
-				right_window_s1, right_window_s2 = get_right_windows
-				break if right_window_s2 && (text_similarity(right_window_s1, right_window_s2) > @sim_threshold)
+			_beg_s2 = get_beg_s2(beg_s1)
+			break _beg_s2 unless _beg_s2.nil?
+		end
 
-				search_position = @beg_s2 + 1
-			end
+		# To return nil when it fails to find an anchor
+		return nil if beg_s2.class == Range
 
-			break unless @beg_s2.nil?
+		# To extend the block to the left
+		b1 = beg_s1
+		b2 = beg_s2
+		left_boundary_b2 = [@pos_s2_last_match, (@cultivation_map.last_cultivated_position(b2) || 0)].max
+		while b1 > @pos_s1_last_match && b2 > left_boundary_b2 && @s1[b1 - 1] == @s2[b2 - 1]
+			b1 -= 1; b2 -= 1
+		end
 
-			@beg_s1 += 1
+		# To extend the block to the right
+		e1 = beg_s1 + @size_ngram
+		e2 = beg_s2 + @size_ngram
+		right_boundary_b2 = @cultivation_map.next_cultivated_position(e2) || @pos_s2_final_possible_end
+		while @s1[e1] && e2 < right_boundary_b2 && @s1[e1] == @s2[e2]
+			e1 += 1; e2 += 1
 		end
 
-		return nil if @beg_s1 >= (@s1.length - @size_ngram)
+		@pos_s1_last_match = e1
+		@pos_s2_last_match = e2
 
-		# extend the block
-		b1 = @beg_s1
-		b2 = @beg_s2
-		while b1 >= @end_s1_prev && b2 >= @end_s2_prev && @s1[b1] == @s2[b2]
-			b1 -= 1; b2 -= 1
-		end
+		{source:{begin:b1 , end:e1}, target:{begin:b2, end:e2}}
+	end
 
-		b1 += 1; b2 += 1
+	private
 
-		e1 = @beg_s1 + @size_ngram
-		e2 = @beg_s2 + @size_ngram
-		while @s1[e1] && @s1[e1] == @s2[e2]
-			e1 += 1; e2 += 1
+	def get_beg_s2(beg_s1)
+		# to get the anchor to search for in s2
+		anchor = @s1[beg_s1, @size_ngram]
+
+		# comment out below with the assumption that texts are in the same order
+		# search_position = 0
+		search_position = @pos_s2_last_match
+
+		beg_s2_candidates = find_beg_s2_candidates(anchor, search_position)
+		return nil if beg_s2_candidates.empty?
+
+		find_valid_beg_s2(beg_s1, beg_s2_candidates)
+	end
+
+	# To find beg_s2 which match to the anchor
+	# return nil if the anchor is too much frequent
+	def find_beg_s2_candidates(anchor, search_position)
+		candidates = []
+		while _beg_s2 = @cultivation_map.index(anchor, @s2, search_position)
+			candidates << _beg_s2
+
+			# for speed, skip anchor of high frequency
+			if candidates.length > 5
+				candidates.clear
+				break
+			end
+
+			search_position = _beg_s2 + 1
 		end
+		candidates
+	end
+
+	def find_valid_beg_s2(beg_s1, beg_s2_candidates)
+		valid_beg_s2 = nil
 
-		@end_s1_prev = e1
-		@end_s2_prev = e2
-		@beg_s1 = e1
+		(10 .. 30).step(10).each do |size_window|
+			valid_beg_s2 = nil
 
-		if @reverse
-			{source:{begin:b2 , end:e2}, target:{begin:b1, end:e1}}
-		else
-			{source:{begin:b1 , end:e1}, target:{begin:b2, end:e2}}
+			r = beg_s2_candidates.each do |beg_s2|
+				# if both the begining points are sufficiantly close to the end points of the last match
+				# break if beg_s1 > 0 && beg_s2 > 0 && (beg_s1 - @pos_s1_last_match < 5) && (beg_s2 >= @pos_s2_last_match) && (beg_s2 - @pos_s2_last_match < 5)
+				if beg_s1 > 0 && beg_s2 > 0 && (beg_s1 - @pos_s1_last_match < 5) && (beg_s2 - @pos_s2_last_match < 5)
+					break unless valid_beg_s2.nil?
+					valid_beg_s2 = beg_s2
+					next
+				end
+
+				left_window_s1, left_window_s2 = get_left_windows(beg_s1, beg_s2)
+				if left_window_s1 && (text_similarity(left_window_s1, left_window_s2) > @sim_threshold)
+					break unless valid_beg_s2.nil?
+					valid_beg_s2 = beg_s2
+					next
+				end
+
+				right_window_s1, right_window_s2 = get_right_windows(beg_s1, beg_s2)
+				if right_window_s2 && (text_similarity(right_window_s1, right_window_s2) > @sim_threshold)
+					break unless valid_beg_s2.nil?
+					valid_beg_s2 = beg_s2
+					next
+				end
+			end
+
+			# r == nil means that the inner loop was broken (multiple candidates had passed the tests)
+			# r != nil means that the inner loop was completed (with or w/o a valid beg_s2 found)
+			break unless r.nil?
 		end
+
+		valid_beg_s2
 	end
 
-	private
+	def get_left_windows(beg_s1, beg_s2, size_window = nil)
+		size_window ||= @size_window
 
-	def get_left_windows
-		# commend below with the assumption that the beginning of a document gives a significant locational information
-		# return if @beg_s1 < @size_window || @beg_s2 < @size_window
+		# comment out below with the assumption that the beginning of a document gives a significant locational information
+		# return if @beg_s1 < size_window || @beg_s2 < size_window
 
 		window_s1 = ''
-		loc = @beg_s1 - 1
+		loc = beg_s1 - 1
 		count = 0
-		while count < @size_window && loc >= 0
+		while count < size_window && loc >= 0
 			if @s1[loc] =~ /[0-9a-zA-Z]/
 				window_s1 += @s1[loc]
 				count += 1
@@ -100,9 +149,9 @@ class TextAlignment::AnchorFinder
 		end
 
 		window_s2 = ''
-		loc = @beg_s2 - 1
+		loc = beg_s2 - 1
 		count = 0
-		while count < @size_window && loc >= 0
+		while count < size_window && loc >= 0
 			if @s2[loc] =~ /[0-9a-zA-Z]/
 				window_s2 += @s2[loc]
 				count += 1
@@ -113,15 +162,17 @@ class TextAlignment::AnchorFinder
 		[window_s1, window_s2]
 	end
 
-	def get_right_windows
+	def get_right_windows(beg_s1, beg_s2, size_window = nil)
+		size_window ||= @size_window
+
 		# commend below with the assumption that the end of a document gives a significant locational
-		# return if (@beg_s1 + @size_ngram > (@s1.length - @size_window)) || (@beg_s2 + @size_ngram > (@s2.length - @size_window))
+		# return if (@beg_s1 + @size_ngram > (@s1.length - size_window)) || (@beg_s2 + @size_ngram > (@s2.length - size_window))
 
 		window_s1 = ''
-		loc = @beg_s1 + @size_ngram
+		loc = beg_s1 + @size_ngram
 		len_s1 = @s1.length
 		count = 0
-		while count < @size_window && loc < len_s1
+		while count < size_window && loc < len_s1
 			if @s1[loc] =~ /[0-9a-zA-Z]/
 				window_s1 += @s1[loc]
 				count += 1
@@ -130,10 +181,10 @@ class TextAlignment::AnchorFinder
 		end
 
 		window_s2 = ''
-		loc = @beg_s2 + @size_ngram
+		loc = beg_s2 + @size_ngram
 		len_s2 = @s2.length
 		count = 0
-		while count < @size_window && loc < len_s2
+		while count < size_window && loc < len_s2
 			if @s2[loc] =~ /[0-9a-zA-Z]/
 				window_s2 += @s2[loc]
 				count += 1
@@ -148,5 +199,4 @@ class TextAlignment::AnchorFinder
 		return 0 if str1.nil? || str2.nil?
 		String::Similarity.cosine(str1, str2, ngram:ngram_order)
 	end
-
-end
+end

data/lib/text_alignment/{mappings.rb → char_mapping.rb}

@@ -1,6 +1,6 @@
 module TextAlignment; end unless defined? TextAlignment
 
-TextAlignment::MAPPINGS = [
+TextAlignment::CHAR_MAPPING = [
 	["©", "(c)"],			#U+00A9 (Copyright Sign)
 
 	["α", "alpha"],		#U+03B1 (greek small letter alpha)
@@ -61,9 +61,10 @@ TextAlignment::MAPPINGS = [
 	["•", "*"],				#U+2022 (bullet)
 	[" ", " "],				#U+2009 (thin space)
 	[" ", " "],				#U+200A (hair space)
-	[" ", " "],				#U+00A0 (no-break space)
+	[" ", " "],				#U+00A0 (Non-Breaking space)
 	["　", " "],				#U+3000 (ideographic space)
-	["‑", "-"],				#U+2211 (Non-Breaking Hyphen)
+	["‐", "-"],				#U+2010 (Hyphen)
+	["‑", "-"],				#U+2011 (Non-Breaking Hyphen)
 	["−", "-"],				#U+2212 (minus sign)
 	["–", "-"],				#U+2013 (en dash)
 	["′", "'"],				#U+2032 (prime)
@@ -75,98 +76,114 @@ TextAlignment::MAPPINGS = [
 ]
 
 
-TextAlignment::PADDING_LETTERS = ['@', '^', '|', '#', '$', '%', '&', '_'] unless defined? TextAlignment::PADDING_LETTERS
+class TextAlignment::CharMapping
+	attr_reader :mapped_text
 
+	def initialize(_text, char_mapping = nil)
+		char_mapping ||= TextAlignment::CHAR_MAPPING
+		@mapped_text, offset_mapping = enmap_text(_text, char_mapping)
+		@index_enmap = offset_mapping.to_h
+		@index_demap = offset_mapping.map{|m| m.reverse}.to_h
+	end
 
-class << TextAlignment
-	def single_character_mapping_preprocessing(_str1, _str2, _mappings = nil)
-		_mappings ||= TextAlignment::MAPPINGS
-
-		character_mappings = _mappings.select{|m| m[0].length == 1 && m[1].length == 1}
-		if character_mappings.empty?
-			[_str1, _str2, _mappings]
-		else
-			characters_from = character_mappings.collect{|m| m[0]}.join
-			characters_to   = character_mappings.collect{|m| m[1]}.join
-			characters_to.gsub!(/-/, '\-')
+	def enmap_position(position)
+		@index_enmap[position]
+	end
 
-			str1 = _str1.tr(characters_from, characters_to)
-			str2 = _str2.tr(characters_from, characters_to)
+	def demap_position(position)
+		@index_demap[position]
+	end
 
-			mappings = _mappings.select{|m| m[0].length > 1 || m[1].length > 1}
+	def enmap_denotations(_denotations)
+		return nil if _denotations.nil?
 
-			[str1, str2, mappings]
+		denotations = _denotations.map do |d|
+			d.dup.merge(span:{begin:enmap_position(d[:span][:begin]), end:enmap_position(d[:span][:end])})
 		end
 	end
 
-	def long_to_one_mapping_preprocessing(_str1, _str2, _mappings = nil)
-		_mappings ||= TextAlignment::MAPPINGS
-
-		long_to_one_mappings = _mappings.select{|m| m[0].length == 1 && m[1].length > 1}
-		if long_to_one_mappings.empty?
-			[_str1, _str2, _mappings]
-		else
-			## long to one character mappings
-			pletters = TextAlignment::PADDING_LETTERS
-
-			# find the padding letter for str1
-			@padding_letter1 = begin
-				i = pletters.index{|l| _str2.index(l).nil?}
-				raise RuntimeError, "Could not find a padding letter for str1" if i.nil?
-				TextAlignment::PADDING_LETTERS[i]
-			end
+	private
 
-			# find the padding letter for str2
-			@padding_letter2 = begin
-				i = pletters.index{|l| l != @padding_letter1 && _str1.index(l).nil?}
-				raise RuntimeError, "Could not find a padding letter for str2" if i.nil?
-				TextAlignment::PADDING_LETTERS[i]
-			end
+	def enmap_text(_text, char_mapping)
+		text = _text.dup
 
-			str1 = str2 = nil
-			long_to_one_mappings.each do |f|
-				from = f[1]
-
-				str1 = if _str2.index(f[0])
-					to = f[0] + (@padding_letter1 * (f[1].length - 1))
-					_str1.gsub(from, to)
-				else
-					_str1
-				end
-
-				str2 = if _str1.index(f[0])
-					to = f[0] + (@padding_letter2 * (f[1].length - 1))
-					_str2.gsub(from, to)
-				else
-					_str2
-				end
+		# To execute the single letter mapping
+		char_mapping.each do |one, long|
+			text.gsub!(one, long) if long.length == 1
+		end
+
+		# To get the (location, length) index for replacements
+		loc_len = []
+		char_mapping.each do |one, long|
+			next if long.length == 1
+
+			init_next = 0
+			while loc = text.index(long, init_next)
+				loc_len << [loc, long.length]
+				init_next = loc + long.length
 			end
-			mappings = _mappings.select{|m| m[0].length > 1 || m[1].length == 1}
 
-			[str1, str2, mappings]
+			# a workaround to avoid messing-up due to embedding
+			text.gsub!(long, one * long.length)
 		end
-	end
 
-	def compute_similarity(_s1, _s2, sdiff)
-		return 0 if sdiff.nil?
+		# To get the (location, length) index for consecutive whitespace sequences
+		init_next = 0
+		while loc = text.index(/\s{2,}/, init_next)
+			len = $~[0].length
+			loc_len << [loc, len]
+			init_next = loc + len
+		end
+
+		loc_len.sort!{|a, b| a[0] <=> b[0]}
+
+		# To get the offset_mapping before and after replacement
+		offset_mapping = []
+		init_next = 0
+		j = 0
 
-		# compute the lcs only with non-whitespace letters
-		lcs = sdiff.count{|d| d.action == '=' && d.old_element =~ /\S/ && d.new_element =~ /\S/}
-		return 0 if lcs == 0
+		loc_len.each do |loc, len|
+			offset_mapping += (init_next .. loc).map do |i|
+				j += 1
+				[i, j - 1]
+			end
+			init_next = loc + len
+		end
 
-		s1 = if @padding_letter1
-			_s1.tr(@padding_letter1, ' ')
-		else
-			_s1
+		offset_mapping += (init_next .. text.length).map do |i|
+			j += 1
+			[i, j - 1]
 		end
 
-		s2 = if @padding_letter2
-			_s2.tr(@padding_letter2, ' ')
-		else
-			_s2
+		# To execute the long letter mapping
+		char_mapping.each do |one, long|
+			text.gsub!(one * long.length, one) if long.length > 1
 		end
 
-		similarity = lcs.to_f / [s1.scan(/\S/).count, s2.scan(/\S/).count].min
+		# To replace multi whitespace sequences to a space
+		text.gsub!(/\s{2,}/, ' ')
+
+		[text, offset_mapping]
 	end
+end
+
+if __FILE__ == $0
+	require 'json'
+
+	unless ARGV.length == 1
+		warn "#{$0} an_annotation_json_file.json" 
+		exit
+	end
+	annotations = JSON.parse File.read(ARGV[0]).strip, symbolize_names: true
+	denotations = annotations[:denotations]
+	if denotations.nil? && annotations[:tracks]
+		denotations = annotations[:tracks].first[:denotations]
+	end
+
+	text_mapping = TextAlignment::CharMapping.new(annotations[:text])
+	text_mapped = text_mapping.mapped_text
+	denotations_mapped = text_mapping.enmap_denotations(denotations)
+	new_annotations = {text:text_mapped, denotations:denotations_mapped}
 
+	puts new_annotations.to_json
 end

data/lib/text_alignment/cultivation_map.rb

@@ -0,0 +1,94 @@
+module TextAlignment; end unless defined? TextAlignment
+
+class TextAlignment::CultivationMap
+	attr_reader :map
+
+	def initialize
+		@map = []
+	end
+
+	def cultivate(regions)
+		@map += regions
+		@map.sort!{|a, b| a[0] <=> b[0]}
+		new_map = []
+		@map.each do |region|
+			if new_map.empty?
+				new_map << region
+			elsif new_map.last[1] > region[0]
+				raise "Overlapping regions: #{new_map.last} : #{region}"
+			elsif new_map.last[1] == region[0]
+				new_map.last[1] == region[1]
+			else
+				new_map << region
+			end
+		end
+		@map = new_map
+	end
+
+	def search_again_position(position, end_position = nil)
+		end_position ||= position
+		region = @map.bsearch{|r| end_position < r[1]}
+		if region.nil? || region[0] > position
+			nil
+		else
+			region[1]
+		end
+	end
+
+	def last_cultivated_position(position)
+		ridx = @map.rindex{|r| r[1] <= position}
+		ridx.nil? ? nil : @map[ridx][1]
+	end
+
+	def next_cultivated_position(position)
+		region = @map.bsearch{|r| position < r[0]}
+		region.nil? ? nil : region[0]
+	end
+
+	def in_regions(region)
+		@map.select{|r| (r[1] > region[0] && r[1] <= region[1]) || (r[0] < region[1] && r[0] >= region[0])}
+	end
+
+	def region_state(region)
+		closed_parts = in_regions(region)
+		if closed_parts.empty?
+			[:open, region]
+		else
+			if front_open?(region, closed_parts)
+				if rear_open?(region, closed_parts)
+					[:middle_closed, [closed_parts.first[1], closed_parts.last[0]]]
+				else
+					[:front_open, [region[0], closed_parts.first[0]]]
+				end
+			else
+				if rear_open?(region, closed_parts)
+					[:rear_open, [closed_parts.last[1], region[1]]]
+				else
+					[:closed, nil]
+				end
+			end
+		end
+	end
+
+	def index(target, string, position)
+		length = target.length
+		loop do
+			_begin = string.index(target, position)
+			break if _begin.nil?
+			position = search_again_position(_begin)
+			next unless position.nil?
+			break _begin if region_state([_begin, _begin + length])[0] == :open
+			position = _begin + 1
+		end
+	end
+
+	private
+
+	def front_open?(region, closed_parts)
+		closed_parts.first[0] > region[0]
+	end
+
+	def rear_open?(region, closed_parts)
+		closed_parts.last[1] < region[1]
+	end
+end

data/lib/text_alignment/glcs_alignment_fast.rb

@@ -5,7 +5,7 @@ require 'text_alignment/find_divisions'
 require 'text_alignment/lcs_comparison'
 require 'text_alignment/lcs_alignment'
 require 'text_alignment/glcs_alignment'
-require 'text_alignment/mappings'
+require 'text_alignment/char_mapping'
 
 module TextAlignment; end unless defined? TextAlignment
 
@@ -106,7 +106,7 @@ if __FILE__ == $0
 
 	dictionary = [["β", "beta"]]
 	# align = TextAlignment::TextAlignment.new(str1, str2)
-	align = TextAlignment::TextAlignment.new(str1, str2, TextAlignment::MAPPINGS)
+	align = TextAlignment::TextAlignment.new(str1, str2, TextAlignment::CHAR_MAPPING)
 	p align.common_elements
 	p align.mapped_elements
 end

data/lib/text_alignment/mixed_alignment.rb

@@ -6,7 +6,7 @@ require 'text_alignment/lcs_comparison'
 require 'text_alignment/lcs_alignment'
 require 'text_alignment/lcs_cdiff'
 require 'text_alignment/glcs_alignment'
-require 'text_alignment/mappings'
+require 'text_alignment/char_mapping'
 
 module TextAlignment; end unless defined? TextAlignment
 
@@ -20,7 +20,9 @@ class TextAlignment::MixedAlignment
 	def initialize(_str1, _str2, _mappings = nil)
 		raise ArgumentError, "nil string" if _str1.nil? || _str2.nil?
 
-		str1, str2, mappings = TextAlignment::long_to_one_mapping_preprocessing(_str1, _str2, _mappings)
+		mappings ||= TextAlignment::CHAR_MAPPING
+		str1 = _str1.dup
+		str2 = _str2.dup
 
 		_compute_mixed_alignment(str1, str2, mappings)
 	end
@@ -63,7 +65,7 @@ class TextAlignment::MixedAlignment
 		end
 
 		cmp = TextAlignment::LCSComparison.new(str1, str2, lcs, @sdiff)
-		@similarity         = TextAlignment::compute_similarity(str1, str2, @sdiff)
+		@similarity         = compute_similarity(str1, str2, @sdiff)
 		@str1_match_initial = cmp.str1_match_initial
 		@str1_match_final   = cmp.str1_match_final
 		@str2_match_initial = cmp.str2_match_initial
@@ -139,4 +141,20 @@ class TextAlignment::MixedAlignment
 		@position_map_end = posmap_end.sort.to_h
 	end
 
+	def compute_similarity(s1, s2, sdiff)
+		return 0 if sdiff.nil?
+
+		# recoverbility
+		count_nws =	sdiff.count{|d| d.old_element =~ /\S/}
+		count_nws_match =	sdiff.count{|d| d.action == '=' && d.old_element =~ /\S/}
+
+		coverage = count_nws_match.to_f / count_nws
+
+		# fragmentation rate
+		count_ofrag = sdiff.count{|d| d.old_element =~ /\s/}
+		count_frag = sdiff.collect{|d| (d.action == '=') && (d.old_element =~/\s/) ? ' ' : d.action}.join.scan(/=+/).count
+		rate_frag = count_ofrag.to_f / count_frag
+
+		similarity = coverage * rate_frag
+	end
 end

data/lib/text_alignment/text_alignment.rb

@@ -2,6 +2,7 @@
 require 'text_alignment/constants'
 require 'text_alignment/anchor_finder'
 require 'text_alignment/mixed_alignment'
+require 'text_alignment/cultivation_map'
 
 module TextAlignment; end unless defined? TextAlignment
 
@@ -10,253 +11,71 @@ class TextAlignment::TextAlignment
 	attr_reader :similarity
 	attr_reader :lost_annotations
 
-	def initialize(_str1, _str2, denotations = nil, _size_ngram = nil, _size_window = nil, _text_similiarity_threshold = nil)
-		raise ArgumentError, "nil string" if _str1.nil? || _str2.nil?
+	# Initialize with a reference text, again which texts will be aligned
+	def initialize(reference_text, to_prevent_overlap = false)
+		raise ArgumentError, "nil text" if reference_text.nil?
 
-		@block_alignment = {source_text:_str1, target_text:_str2}
-		@original_str1 = _str1
-		@original_str2 = _str2
+		@original_reference_text = reference_text
+		@rtext_mapping = TextAlignment::CharMapping.new(reference_text)
+		@mapped_reference_text = @rtext_mapping.mapped_text
+		@to_prevent_overlap = to_prevent_overlap
 
-		str1, str2, @mappings = TextAlignment::single_character_mapping_preprocessing(_str1, _str2)
-
-		if r = whole_block_alignment(str1, str2)
-			@block_alignment[:blocks] = r
-			return
-		end
+		@original_text = nil
+		@block_alignment = nil
+		@cultivation_map = TextAlignment::CultivationMap.new
+	end
 
-		## to find block alignments
-		anchor_finder = TextAlignment::AnchorFinder.new(str1, str2, _size_ngram, _size_window, _text_similiarity_threshold)
+	def align(text, denotations = nil)
+		# To maintain the cultivation map
+		update_cultivation_map if @to_prevent_overlap
 
-		blocks = []
-		while block = anchor_finder.get_next_anchor
-			last = blocks.last
-			if last && (block[:source][:begin] == last[:source][:end] + 1) && (block[:target][:begin] == last[:target][:end] + 1)
-				last[:source][:end] = block[:source][:end]
-				last[:target][:end] = block[:target][:end]
-			else
-				blocks << block.merge(alignment: :block, delta: block[:target][:begin] - block[:source][:begin])
-			end
+		# In case the input text is the same as the previous one, reuse the previous text mapping
+		unless @original_text && @original_text == text
+			@original_text = text
+			@text_mapping = TextAlignment::CharMapping.new(text)
 		end
 
-		# pp blocks
-		# puts "-----"
-		# puts
-		# exit
-		# blocks.each do |b|
-		# 	p [b[:source], b[:target]]
-		# 	puts "---"
-		# 	puts str1[b[:source][:begin] ... b[:source][:end]]
-		# 	puts "---"
-		# 	puts str2[b[:target][:begin] ... b[:target][:end]]
-		# 	puts "====="
-		# 	puts
-		# end
-		# puts "-=-=-=-=-"
-		# puts
-
-		## to fill the gaps
-		last_block = nil
-		blocks2 = blocks.inject([]) do |sum, block|
-			b1 = last_block ? last_block[:source][:end] : 0
-			e1 = block[:source][:begin]
-
-			sum += if b1 == e1
-				[block]
-			else
-				b2 = last_block ? last_block[:target][:end] : 0
-				e2 = block[:target][:begin]
-
-				if b2 == e2
-					[
-						{source:{begin:b1, end:e1}, target:{begin:b2, end:e2}, alignment: :empty},
-						block
-					]
-				else
-					if b1 == 0 && b2 == 0
-						len_buffer = (e1 * (1 + TextAlignment::BUFFER_RATE)).to_i + TextAlignment::BUFFER_MIN
-						b2 = e2 - len_buffer if e2 > len_buffer
-					end
-
-					_str1 = str1[b1 ... e1]
-					_str2 = str2[b2 ... e2]
-
-					if _str1.strip.empty? || _str2.strip.empty?
-						[
-							{source:{begin:b1, end:e1}, target:{begin:b2, end:e2}, alignment: :empty},
-							block
-						]
-					else
-						local_alignment_blocks(str1, b1, e1, str2, b2, e2, denotations) << block
-					end
-				end
-			end
+		@mapped_text = @text_mapping.mapped_text
+		denotations_mapped = @text_mapping.enmap_denotations(denotations)
 
-			last_block = block
-			sum
-		end
+		## To generate the block_alignment of the input text against the reference text
+		# Initialization
+		# @block_alignment = {text: @original_text, reference_text: @original_reference_text, denotations: denotations}
+		@block_alignment = {text: @mapped_text, reference_text: @mapped_reference_text, denotations: denotations}
 
-		# the last step
-		blocks2 += if last_block.nil?
-			local_alignment_blocks(str1, 0, str1.length, str2, 0, str2.length, denotations)
+		# Generation
+		@block_alignment[:blocks] = if r = whole_block_alignment(@mapped_text, @mapped_reference_text, @cultivation_map)
+			r
 		else
-			b1 = last_block[:source][:end]
-			if b1 < str1.length
-				e1 = str1.length
-
-				b2 = last_block[:target][:end]
-				if b2 < str2.length
-					len_buffer = ((e1 - b1) * (1 + TextAlignment::BUFFER_RATE)).to_i + TextAlignment::BUFFER_MIN
-					e2 = (str2.length - b2) > len_buffer ? b2 + len_buffer : str2.length
-					local_alignment_blocks(str1, b1, e1, str2, b2, e2, denotations)
-				else
-					[{source:{begin:last_block[:source][:end], end:str1.length}, alignment: :empty}]
-				end
-			else
-				[]
-			end
+			find_block_alignment(@mapped_text, @mapped_reference_text, denotations_mapped, @cultivation_map)
 		end
-
-		@block_alignment[:blocks] = blocks2
 	end
 
-	def whole_block_alignment(str1, str2)
-		## Block exact match
-		block_begin = str2.index(str1)
-		unless block_begin.nil?
-			return [{source:{begin:0, end:str1.length}, target:{begin:block_begin, end:block_begin + str1.length}, delta:block_begin, alignment: :block}]
-		end
+	def update_cultivation_map
+		return if @block_alignment.nil? || @block_alignment[:blocks].nil?
 
-		block_begin = str2.downcase.index(str1.downcase)
-		unless block_begin.nil?
-			return [{source:{begin:0, end:str1.length}, target:{begin:block_begin, end:block_begin + str1.length}, delta:block_begin, alignment: :block}]
-		end
-
-		nil
-	end
-
-	def local_alignment_blocks(str1, b1, e1, str2, b2, e2, denotations = nil)
-		block2 = str2[b2 ... e2]
-
-		## term-based alignment
-		tblocks = if denotations
-			ds_in_scope = denotations.select{|d| d[:span][:begin] >= b1 && d[:span][:end] <= e1}.
-							sort{|d1, d2| d1[:span][:begin] <=> d2[:span][:begin] || d2[:span][:end] <=> d1[:span][:end] }.
-							map{|d| d.merge(lex:str1[d[:span][:begin] ... d[:span][:end]])}
-
-			position = 0
-			tblocks = ds_in_scope.map do |term|
-				lex = term[:lex]
-				r = block2.index(lex, position)
-				if r.nil?
-					position = nil
-					break
-				end
-				position = r + lex.length
-				{source:term[:span], target:{begin:r + b2, end:r + b2 + lex.length}, alignment: :term, delta: r - term[:span][:begin]}
-			end
-
-			# missing term found
-			tblocks = [] if position.nil?
-
-			# redundant matching found
-			unless position.nil?
-				ds_in_scope.each do |term|
-					lex = term[:lex]
-					look_forward = block2.index(lex, position)
-					unless look_forward.nil?
-						tblocks = []
-						break
-					end
-				end
-			end
-
-			tblocks
-		else
-			[]
-		end
-
-		if tblocks.empty?
-			if b1 == 0 && e1 == str1.length
-				if (e1 > 2000) || (e2 > 2000)
-					[{source:{begin:b1, end:e1}, target:{begin:b2, end:e2}, alignment: :empty}]
-				else
-					block1 = str1[b1 ... e1]
-					block2 = str2[b2 ... e2]
-
-					## character-based alignment
-					alignment = TextAlignment::MixedAlignment.new(block1.downcase, block2.downcase, @mappings)
-					if alignment.sdiff.nil?
-						[{source:{begin:b1, end:e1}, target:{begin:b2, end:e2}, alignment: :empty}]
-					else
-						[{source:{begin:b1, end:e1}, target:{begin:b2, end:e2}, alignment: alignment, similarity: alignment.similarity}]
-					end
-				end
+		## To update the cultivation map
+		newly_cultivated_regions = @block_alignment[:blocks].collect do |b|
+			if b[:alignment] == :block || b[:alignment] == :term
+				[b[:target][:begin], b[:target][:end]]
 			else
-				block1 = str1[b1 ... e1]
-				block2 = str2[b2 ... e2]
-
-				## character-based alignment
-				alignment = TextAlignment::MixedAlignment.new(block1.downcase, block2.downcase, @mappings)
-				if alignment.sdiff.nil?
-					[{source:{begin:b1, end:e1}, target:{begin:b2, end:e2}, alignment: :empty}]
-				else
-					[{source:{begin:b1, end:e1}, target:{begin:b2, end:e2}, alignment: alignment, similarity: alignment.similarity}]
-				end
-			end
-		else
-			last_tblock = nil
-			lblocks = tblocks.inject([]) do |sum, tblock|
-				tb1 = last_tblock ? last_tblock[:source][:end] : b1
-				te1 = tblock[:source][:begin]
-
-				sum += if te1 == tb1
-					[tblock]
-				else
-					tb2 = last_tblock ? last_tblock[:target][:end] : b2
-					te2 = tblock[:target][:begin]
-
-					if b2 == e2
-						[
-							{source:{begin:tb1, end:te1}, alignment: :empty},
-							tblock
-						]
-					else
-						[
-							{source:{begin:tb1, end:te1}, target:{begin:tb2, end:te2}, alignment: :empty},
-							tblock
-						]
-					end
-				end
-
-				last_tblock = tblock
-				sum
+				nil
 			end
-
-			if last_tblock[:source][:end] < e1
-				if last_tblock[:target][:end] < e2
-					lblocks << {source:{begin:last_tblock[:source][:end], end:e1}, target:{begin:last_tblock[:target][:end], end:e2}, alignment: :empty}
-				else
-					lblocks << {source:{begin:last_tblock[:source][:end], end:e1}, alignment: :empty}
-				end
+		end.compact.inject([]) do |condensed, region|
+			if condensed.empty? || (condensed.last.last + 1 < region.first)
+				condensed.push region
+			else
+				condensed.last[1] = region.last
 			end
-
-			lblocks
+			condensed
 		end
-	end
-
 
-	def indices(str, target)
-	  position = 0
-	  len = target.len
-	  Enumerator.new do |yielder|
-	    while idx = str.index(target, position)
-	      yielder << idx
-	      position = idx + len
-	    end
-	  end
+		@cultivation_map.cultivate(newly_cultivated_regions)
 	end
 
-	def transform_begin_position(begin_position)
+	def transform_begin_position(_begin_position)
+		begin_position = @text_mapping.enmap_position(_begin_position)
+
 		i = @block_alignment[:blocks].index{|b| b[:source][:end] > begin_position}
 		block = @block_alignment[:blocks][i]
 
@@ -272,9 +91,13 @@ class TextAlignment::TextAlignment
 			r = block[:alignment].transform_begin_position(begin_position - block[:source][:begin])
 			r.nil? ? nil : r + block[:target][:begin]
 		end
+
+		@rtext_mapping.demap_position(b)
 	end
 
-	def transform_end_position(end_position)
+	def transform_end_position(_end_position)
+		end_position = @text_mapping.enmap_position(_end_position)
+
 		i = @block_alignment[:blocks].index{|b| b[:source][:end] >= end_position}
 		block = @block_alignment[:blocks][i]
 
@@ -290,6 +113,8 @@ class TextAlignment::TextAlignment
 			r = block[:alignment].transform_end_position(end_position - block[:source][:begin])
 			r.nil? ? nil : r + block[:target][:begin]
 		end
+
+		@rtext_mapping.demap_position(e)
 	end
 
 	def transform_a_span(span)
@@ -308,7 +133,7 @@ class TextAlignment::TextAlignment
 			source = {begin:d.begin, end:d.end}
 			d.begin = transform_begin_position(d.begin);
 			d.end = transform_end_position(d.end);
-			raise "invalid transform" unless !d.begin.nil? && !d.end.nil? && d.begin >= 0 && d.end > d.begin && d.end <= @original_str2.length
+			raise "invalid transform" unless !d.begin.nil? && !d.end.nil? && d.begin >= 0 && d.end > d.begin && d.end <= @original_reference_text.length
 		rescue
 			@lost_annotations << {source: source, target:{begin:d.begin, end:d.end}}
 			d.begin = nil
@@ -324,7 +149,7 @@ class TextAlignment::TextAlignment
 
 		r = hdenotations.collect do |d|
 			t = transform_a_span(d[:span])
-			raise "invalid transform" unless !t[:begin].nil? && !t[:end].nil? && t[:begin] >= 0 && t[:end] > t[:begin] && t[:end] <= @original_str2.length
+			raise "invalid transform" unless !t[:begin].nil? && !t[:end].nil? && t[:begin] >= 0 && t[:end] > t[:begin] && t[:end] <= @original_reference_text.length
 			new_d = d.dup.merge({span:t})
 		rescue
 			@lost_annotations << {source: d[:span], target:t}
@@ -335,8 +160,8 @@ class TextAlignment::TextAlignment
 	end
 
 	def alignment_show
-		stext = @block_alignment[:source_text]
-		ttext = @block_alignment[:target_text]
+		stext = @mapped_text
+		ttext = @mapped_reference_text
 
 		show = ''
 		@block_alignment[:blocks].each do |a|
@@ -392,9 +217,192 @@ class TextAlignment::TextAlignment
 
 				"***** local mismatch [#{a[:source][:begin]} - #{a[:source][:end]}] [#{a[:target][:begin]} - #{a[:target][:end]}] (similarity: #{a[:similarity]})\n" +
 				"[#{astr1}]\n" +
-				"[#{astr2}]\n\n"
+				"[#{astr2.gsub("\n", " ")}]\n\n"
 			end
 		end
 		show
 	end
+
+	private
+
+	def find_block_alignment(str1, str2, denotations, cultivation_map)
+		## to find block alignments
+		anchor_finder = TextAlignment::AnchorFinder.new(str1, str2, cultivation_map)
+
+		blocks = []
+		while block = anchor_finder.get_next_anchor
+			last = blocks.last
+			if last && (block[:source][:begin] == last[:source][:end] + 1) && (block[:target][:begin] == last[:target][:end] + 1)
+				last[:source][:end] = block[:source][:end]
+				last[:target][:end] = block[:target][:end]
+			else
+				blocks << block.merge(alignment: :block, delta: block[:target][:begin] - block[:source][:begin])
+			end
+		end
+
+		# pp blocks
+		# puts "-----"
+		# puts
+		# exit
+		# blocks.each do |b|
+		# 	p [b[:source], b[:target]]
+		# 	puts "---"
+		# 	puts str1[b[:source][:begin] ... b[:source][:end]]
+		# 	puts "---"
+		# 	puts str2[b[:target][:begin] ... b[:target][:end]]
+		# 	puts "====="
+		# 	puts
+		# end
+		# puts "-=-=-=-=-"
+		# puts
+
+		## To fill the gaps
+		## lblock: last block, cblock: current block
+		lblock = nil
+		blocks2 = (blocks + [nil]).inject([]) do |sum, cblock|
+			b1 = lblock.nil? ? 0 : lblock[:source][:end]
+			e1 = cblock.nil? ? str1.length : cblock[:source][:begin]
+
+			if b1 < e1
+				b2 = lblock.nil? ? 0 : lblock[:target][:end]
+				e2 = cblock.nil? ? str2.length : cblock[:target][:begin]
+				_str1 = str1[b1 ... e1]
+				_str2 = str2[b2 ... e2]
+
+				sum += if _str1.strip.empty? || _str2.strip.empty?
+					[{source:{begin:b1, end:e1}, target:{begin:b2, end:e2}, alignment: :empty}]
+				else
+					len_buffer = ((e1 - b1) * (1 + TextAlignment::BUFFER_RATE)).to_i + TextAlignment::BUFFER_MIN
+					region_state, state_region = cultivation_map.region_state([b2, e2])
+					case region_state
+					when :closed
+						[]
+					when :front_open
+						oe2 = state_region[1]
+						me2 = (oe2 - b2) > len_buffer ? b2 + len_buffer : oe2
+						local_alignment(str1, b1, e1, str2, b2, me2, denotations, cultivation_map)
+					when :rear_open
+						ob2 = state_region[0]
+						mb2 = (e2 - ob2) > len_buffer ? e2 - len_buffer : ob2
+						local_alignment(str1, b1, e1, str2, mb2, e2, denotations, cultivation_map)
+					when :middle_closed
+						oe2 = state_region[0]
+						me2 = (oe2 - b2) > len_buffer ? b2 + len_buffer : oe2
+						attempt1 = local_alignment(str1, b1, e1, str2, b2, me2, denotations, cultivation_map)
+						if attempt1.empty?
+							ob2 = state_region[1]
+							mb2 = (e2 - ob2) > len_buffer ? e2 - len_buffer : ob2
+							local_alignment(str1, b1, e1, str2, mb2, e2, denotations, cultivation_map)
+						else
+							attempt1
+						end
+					else # :open
+						if (e2 - b2) > len_buffer
+							attempt1 = local_alignment(str1, b1, e1, str2, b2, b2 + len_buffer, denotations, cultivation_map)
+							if attempt1.empty?
+								local_alignment(str1, b1, e1, str2, e2 - len_buffer, e2, denotations, cultivation_map)
+							else
+								attempt1
+							end
+						else
+							local_alignment(str1, b1, e1, str2, b2, e2, denotations, cultivation_map)
+						end
+					end
+				end
+			end
+
+			lblock = cblock
+			cblock.nil? ? sum : sum << cblock
+		end
+
+	end
+
+	def whole_block_alignment(str1, str2, cultivation_map)
+		block_begin = cultivation_map.index(str1, str2, 0)
+		return [{source:{begin:0, end:str1.length}, target:{begin:block_begin, end:block_begin + str1.length}, delta:block_begin, alignment: :block}] unless block_begin.nil?
+
+		block_begin = cultivation_map.index(str1.downcase, str2.downcase, 0)
+		return [{source:{begin:0, end:str1.length}, target:{begin:block_begin, end:block_begin + str1.length}, delta:block_begin, alignment: :block}] unless block_begin.nil?
+
+		nil
+	end
+
+	def local_alignment(str1, b1, e1, str2, b2, e2, denotations = nil, cultivation_map)
+		tblocks = term_based_alignment(str1, b1, e1, str2, b2, e2, denotations, cultivation_map)
+		if tblocks.empty?
+			lcs_alignment(str1, b1, e1, str2, b2, e2, cultivation_map)
+		else
+			tblocks
+		end
+	end
+
+	def term_based_alignment(str1, b1, e1, str2, b2, e2, denotations = nil, cultivation_map)
+		str2_block = str2[0 ... e2]
+
+		## term-based alignment
+		tblocks = if denotations
+			denotations_in_scope = denotations.select{|d| d[:span][:begin] >= b1 && d[:span][:end] <= e1}.
+							sort{|d1, d2| d1[:span][:begin] <=> d2[:span][:begin] || d2[:span][:end] <=> d1[:span][:end] }.
+							map{|d| d.merge(lex:str1[d[:span][:begin] ... d[:span][:end]])}
+
+			search_position = b2
+			_tblocks = denotations_in_scope.map do |denotation|
+				lex = denotation[:lex]
+				term_begin = cultivation_map.index(lex, str2_block, search_position)
+				break [] if term_begin.nil? # break the loop if a missing term is found
+				search_position = term_begin + lex.length
+				{source:denotation[:span], target:{begin:term_begin, end:term_begin + lex.length}, alignment: :term, similarity: 0.9, delta: term_begin - denotation[:span][:begin]}
+			end
+
+			# redundant matching found
+			unless _tblocks.empty?
+				search_position = _tblocks.last[:target][:end]
+				denotations_in_scope.each do |term|
+					look_forward = cultivation_map.index(term[:lex], str2_block, search_position)
+					unless look_forward.nil?
+						_tblocks = []
+						break
+					end
+				end
+			end
+
+			_tblocks
+		else
+			[]
+		end
+
+		ltblock = nil
+		tblocks2 = (tblocks + [nil]).inject([]) do |sum, ctblock|
+			tb1 = ltblock.nil? ? b1 : ltblock[:source][:end]
+			te1 = ctblock.nil? ? e1 : ctblock[:source][:begin]
+
+			if te1 > tb1
+				tb2 = ltblock.nil? ? b2 : ltblock[:target][:end]
+				te2 = ctblock.nil? ? e2 : ctblock[:target][:begin]
+				sum << {source:{begin:tb1, end:te1}, target:{begin:tb2, end:te2}, alignment: :empty}
+			end
+
+			ltblock = ctblock
+			ctblock.nil? ? sum : sum << ctblock
+		end
+
+		tblocks2
+	end
+
+	def lcs_alignment(str1, b1, e1, str2, b2, e2, cultivation_map)
+		source = {begin:b1, end:e1}
+		target = {begin:b2, end:e2}
+
+		if (e1 - b1) > 2000
+			[{source:source, target:target, alignment: :empty}]
+		else
+			alignment = TextAlignment::MixedAlignment.new(str1[b1 ... e1].downcase, str2[b2 ... e2].downcase)
+			if alignment.similarity < 0.5
+				[{source:source, target:target, alignment: :empty}]
+			else
+				[{source:source, target:target, alignment: alignment, similarity: alignment.similarity}]
+			end
+		end
+	end
+
 end

data/lib/text_alignment/version.rb

@@ -1,3 +1,3 @@
 class TextAlignment
-	VERSION = '0.8.1'
+	VERSION = '0.11.1'
 end

metadata

@@ -1,14 +1,14 @@
 --- !ruby/object:Gem::Specification
 name: text_alignment
 version: !ruby/object:Gem::Version
-  version: 0.8.1
+  version: 0.11.1
 platform: ruby
 authors:
 - Jin-Dong Kim
 autorequire: 
 bindir: bin
 cert_chain: []
-date: 2020-10-26 00:00:00.000000000 Z
+date: 2021-03-08 00:00:00.000000000 Z
 dependencies:
 - !ruby/object:Gem::Dependency
   name: ruby-dictionary
@@ -77,7 +77,9 @@ files:
 - lib/text_alignment.rb
 - lib/text_alignment/anchor_finder.rb
 - lib/text_alignment/approximate_fit.rb
+- lib/text_alignment/char_mapping.rb
 - lib/text_alignment/constants.rb
+- lib/text_alignment/cultivation_map.rb
 - lib/text_alignment/find_divisions.rb
 - lib/text_alignment/glcs_alignment.rb
 - lib/text_alignment/glcs_alignment_fast.rb
@@ -86,7 +88,6 @@ files:
 - lib/text_alignment/lcs_cdiff.rb
 - lib/text_alignment/lcs_comparison.rb
 - lib/text_alignment/lcs_min.rb
-- lib/text_alignment/mappings.rb
 - lib/text_alignment/mixed_alignment.rb
 - lib/text_alignment/text_alignment.rb
 - lib/text_alignment/version.rb