Class: ClassifierReborn::LSI

Inherits:
Object
  • Object
show all
Defined in:
lib/classifier-reborn/lsi.rb

Overview

This class implements a Latent Semantic Indexer, which can search, classify and cluster data based on underlying semantic relations. For more information on the algorithms used, please consult Wikipedia.

Instance Attribute Summary collapse

Instance Method Summary collapse

Constructor Details

#initialize(options = {}) ⇒ LSI

Create a fresh index. If you want to call #build_index manually, use

ClassifierReborn::LSI.new :auto_rebuild => false

If you want to use ContentNodes with cached vector transpositions, use

lsi = ClassifierReborn::LSI.new :cache_node_vectors => true



37
38
39
40
41
42
43
44
45
 
# File 'lib/classifier-reborn/lsi.rb', line 37

def initialize(options = {})
  @auto_rebuild = options[:auto_rebuild] != false
  @word_list, @items = WordList.new, {}
  @version, @built_at_version = 0, -1
  @language = options[:language] || 'en'
  if @cache_node_vectors = options[:cache_node_vectors]
    extend CachedContentNode::InstanceMethods
  end
end

Instance Attribute Details

#auto_rebuild ⇒ Object

Returns the value of attribute auto_rebuild.



29
30
31
 
# File 'lib/classifier-reborn/lsi.rb', line 29

def auto_rebuild
  @auto_rebuild
end

#cache_node_vectors ⇒ Object (readonly)

Returns the value of attribute cache_node_vectors.



28
29
30
 
# File 'lib/classifier-reborn/lsi.rb', line 28

def cache_node_vectors
  @cache_node_vectors
end

#word_list ⇒ Object (readonly)

Returns the value of attribute word_list.



28
29
30
 
# File 'lib/classifier-reborn/lsi.rb', line 28

def word_list
  @word_list
end

Instance Method Details

#<<(item) ⇒ Object

A less flexible shorthand for add_item that assumes you are passing in a string with no categorries. item will be duck typed via to_s .



82
83
84
 
# File 'lib/classifier-reborn/lsi.rb', line 82

def <<( item )
  add_item item
end

#add_item(item, *categories, &block) ⇒ Object

Adds an item to the index. item is assumed to be a string, but any item may be indexed so long as it responds to #to_s or if you provide an optional block explaining how the indexer can fetch fresh string data. This optional block is passed the item, so the item may only be a reference to a URL or file name.

For example:

lsi = ClassifierReborn::LSI.new
lsi.add_item "This is just plain text"
lsi.add_item "/home/me/filename.txt" { |x| File.read x }
ar = ActiveRecordObject.find( :all )
lsi.add_item ar, *ar.categories { |x| ar.content }



67
68
69
70
71
72
73
74
75
76
 
# File 'lib/classifier-reborn/lsi.rb', line 67

def add_item( item, *categories, &block )
  clean_word_hash = Hasher.clean_word_hash((block ? block.call(item) : item.to_s), @language)
  @items[item] = if @cache_node_vectors
    CachedContentNode.new(clean_word_hash, *categories)
  else
    ContentNode.new(clean_word_hash, *categories)
  end
  @version += 1
  build_index if @auto_rebuild
end

#build_index(cutoff = 0.75) ⇒ Object

This function rebuilds the index if needs_rebuild? returns true. For very large document spaces, this indexing operation may take some time to complete, so it may be wise to place the operation in another thread.

As a rule, indexing will be fairly swift on modern machines until you have well over 500 documents indexed, or have an incredibly diverse vocabulary for your documents.

The optional parameter “cutoff” is a tuning parameter. When the index is built, a certain number of s-values are discarded from the system. The cutoff parameter tells the indexer how many of these values to keep. A value of 1 for cutoff means that no semantic analysis will take place, turning the LSI class into a simple vector search engine.



121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
 
# File 'lib/classifier-reborn/lsi.rb', line 121

def build_index( cutoff=0.75 )
  return unless needs_rebuild?
  make_word_list

  doc_list = @items.values
  tda = doc_list.collect { |node| node.raw_vector_with( @word_list ) }

  if $GSL
     tdm = GSL::Matrix.alloc(*tda).trans
     ntdm = build_reduced_matrix(tdm, cutoff)

     ntdm.size[1].times do |col|
       vec = GSL::Vector.alloc( ntdm.column(col) ).row
       doc_list[col].lsi_vector = vec
       doc_list[col].lsi_norm = vec.normalize
     end
  else
     tdm = Matrix.rows(tda).trans
     ntdm = build_reduced_matrix(tdm, cutoff)

     ntdm.row_size.times do |col|
       doc_list[col].lsi_vector = ntdm.column(col) if doc_list[col]
       doc_list[col].lsi_norm = ntdm.column(col).normalize  if doc_list[col]
     end
  end

  @built_at_version = @version
end

#categories_for(item) ⇒ Object

Returns the categories for a given indexed items. You are free to add and remove items from this as you see fit. It does not invalide an index to change its categories.



88
89
90
91
 
# File 'lib/classifier-reborn/lsi.rb', line 88

def categories_for(item)
  return [] unless @items[item]
  return @items[item].categories
end

#classify(doc, cutoff = 0.30, &block) ⇒ Object

Return the most obvious category without the score



252
253
254
 
# File 'lib/classifier-reborn/lsi.rb', line 252

def classify( doc, cutoff=0.30, &block )
  return scored_categories(doc, cutoff, &block).last.first
end

#classify_with_score(doc, cutoff = 0.30, &block) ⇒ Object

Return the most obvious category with the score



247
248
249
 
# File 'lib/classifier-reborn/lsi.rb', line 247

def classify_with_score( doc, cutoff=0.30, &block)
  return scored_categories(doc, cutoff, &block).last
end

#find_related(doc, max_nearest = 3, &block) ⇒ Object

This function takes content and finds other documents that are semantically “close”, returning an array of documents sorted from most to least relavant. max_nearest specifies the number of documents to return. A value of 0 means that it returns all the indexed documents, sorted by relavence.

This is particularly useful for identifing clusters in your document space. For example you may want to identify several “What’s Related” items for weblog articles, or find paragraphs that relate to each other in an essay.



239
240
241
242
243
244
 
# File 'lib/classifier-reborn/lsi.rb', line 239

def find_related( doc, max_nearest=3, &block )
  carry =
    proximity_array_for_content( doc, &block ).reject { |pair| pair[0].eql? doc }
  result = carry.collect { |x| x[0] }
  return result[0..max_nearest-1]
end

#highest_ranked_stems(doc, count = 3) ⇒ Object

Prototype, only works on indexed documents. I have no clue if this is going to work, but in theory it’s supposed to.



282
283
284
285
286
287
 
# File 'lib/classifier-reborn/lsi.rb', line 282

def highest_ranked_stems( doc, count=3 )
  raise "Requested stem ranking on non-indexed content!" unless @items[doc]
  arr = node_for_content(doc).lsi_vector.to_a
  top_n = arr.sort.reverse[0..count-1]
  return top_n.collect { |x| @word_list.word_for_index(arr.index(x))}
end

#highest_relative_content(max_chunks = 10) ⇒ Object

This method returns max_chunks entries, ordered by their average semantic rating. Essentially, the average distance of each entry from all other entries is calculated, the highest are returned.

This can be used to build a summary service, or to provide more information about your dataset’s general content. For example, if you were to use categorize on the results of this data, you could gather information on what your dataset is generally about.



158
159
160
161
162
163
164
165
 
# File 'lib/classifier-reborn/lsi.rb', line 158

def highest_relative_content( max_chunks=10 )
   return [] if needs_rebuild?

   avg_density = Hash.new
   @items.each_key { |item| avg_density[item] = proximity_array_for_content(item).inject(0.0) { |x,y| x + y[1]} }

   avg_density.keys.sort_by { |x| avg_density[x] }.reverse[0..max_chunks-1].map
end

#items ⇒ Object

Returns an array of items that are indexed.



103
104
105
 
# File 'lib/classifier-reborn/lsi.rb', line 103

def items
  @items.keys
end

#needs_rebuild? ⇒ Boolean

Returns true if the index needs to be rebuilt. The index needs to be built after all informaton is added, but before you start using it for search, classification and cluster detection.

Returns:

  • (Boolean) 


50
51
52
 
# File 'lib/classifier-reborn/lsi.rb', line 50

def needs_rebuild?
  (@items.size > 1) && (@version != @built_at_version)
end

#proximity_array_for_content(doc, &block) ⇒ Object

This function is the primitive that find_related and classify build upon. It returns an array of 2-element arrays. The first element of this array is a document, and the second is its “score”, defining how “close” it is to other indexed items.

These values are somewhat arbitrary, having to do with the vector space created by your content, so the magnitude is interpretable but not always meaningful between indexes.

The parameter doc is the content to compare. If that content is not indexed, you can pass an optional block to define how to create the text data. See add_item for examples of how this works.



179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
 
# File 'lib/classifier-reborn/lsi.rb', line 179

def proximity_array_for_content( doc, &block )
  return [] if needs_rebuild?

  content_node = node_for_content( doc, &block )
  result =
    @items.keys.collect do |item|
      if $GSL
        val = content_node.search_vector * @items[item].transposed_search_vector
      else
        val = (Matrix[content_node.search_vector] * @items[item].search_vector)[0]
      end
      [item, val]
    end
  result.sort_by { |x| x[1] }.reverse
end

#proximity_norms_for_content(doc, &block) ⇒ Object

Similar to proximity_array_for_content, this function takes similar arguments and returns a similar array. However, it uses the normalized calculated vectors instead of their full versions. This is useful when you’re trying to perform operations on content that is much smaller than the text you’re working with. search uses this primitive.



200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
 
# File 'lib/classifier-reborn/lsi.rb', line 200

def proximity_norms_for_content( doc, &block )
  return [] if needs_rebuild?

  content_node = node_for_content( doc, &block )
  result =
    @items.keys.collect do |item|
      if $GSL
        val = content_node.search_norm * @items[item].search_norm.col
      else
        val = (Matrix[content_node.search_norm] * @items[item].search_norm)[0]
      end
      [item, val]
    end
  result.sort_by { |x| x[1] }.reverse
end

#remove_item(item) ⇒ Object

Removes an item from the database, if it is indexed.



95
96
97
98
99
100
 
# File 'lib/classifier-reborn/lsi.rb', line 95

def remove_item( item )
  if @items.key? item
    @items.delete item
    @version += 1
  end
end

#scored_categories(doc, cutoff = 0.30, &block) ⇒ Object

This function uses a voting system to categorize documents, based on the categories of other documents. It uses the same logic as the find_related function to find related documents, then returns the list of sorted categories.

cutoff signifies the number of documents to consider when clasifying text. A cutoff of 1 means that every document in the index votes on what category the document is in. This may not always make sense.



265
266
267
268
269
270
271
272
273
274
275
276
277
 
# File 'lib/classifier-reborn/lsi.rb', line 265

def scored_categories( doc, cutoff=0.30, &block )
  icutoff = (@items.size * cutoff).round
  carry = proximity_array_for_content( doc, &block )
  carry = carry[0..icutoff-1]
  votes = Hash.new(0.0)
  carry.each do |pair|
    @items[pair[0]].categories.each do |category|
      votes[category] += pair[1]
    end
  end

  return votes.sort_by { |_, score| score }
end

#search(string, max_nearest = 3) ⇒ Object

This function allows for text-based search of your index. Unlike other functions like find_related and classify, search only takes short strings. It will also ignore factors like repeated words. It is best for short, google-like search terms. A search will first priortize lexical relationships, then semantic ones.

While this may seem backwards compared to the other functions that LSI supports, it is actually the same algorithm, just applied on a smaller document.



223
224
225
226
227
228
 
# File 'lib/classifier-reborn/lsi.rb', line 223

def search( string, max_nearest=3 )
  return [] if needs_rebuild?
  carry = proximity_norms_for_content( string )
  result = carry.collect { |x| x[0] }
  return result[0..max_nearest-1]
end