treat 1.1.2 → 1.2.0

data/lib/treat/workers/formatters/serializers/xml.rb

@@ -33,14 +33,14 @@ class Treat::Workers::Formatters::Serializers::XML
         end
       end
       ############ To be refactored
-      unless entity.dependencies.empty?
-        attributes << "dependencies='"
+      unless entity.edges.empty?
+        attributes << "edges='"
         a = []
-        entity.dependencies.each do |dependency|
-          a << ("{target: #{dependency.target}, "+
-          "type: #{dependency.type}, " +
-          "directed: #{dependency.directed}, " +
-          "direction: #{dependency.direction}}" )
+        entity.edges.each do |edge|
+          a << ("{target: #{edge.target}, "+
+          "type: #{edge.type}, " +
+          "directed: #{edge.directed}, " +
+          "direction: #{edge.direction}}" )
         end
         # Structs.
         attributes << a.join(',') + "'"

data/lib/treat/workers/formatters/unserializers/mongo.rb

@@ -22,7 +22,7 @@ module Treat::Workers::Formatters::Unserializers::Mongo
     entity.type.to_s.capitalize.intern).superclass).downcase
     supertype = entity.type.to_s if supertype == 'entity'
     supertypes = supertype + 's'
-    
+    supertypes = 'documents' if entity.type == :collection
     coll = @@database.collection(supertypes)
     records = coll.find(selector).to_a
     
@@ -30,16 +30,24 @@ module Treat::Workers::Formatters::Unserializers::Mongo
       raise Treat::Exception,
       "Couldn't find any records using " +
       "selector #{selector.inspect}."
-    elsif records.size == 1
-      self.do_unserialize(
-      records.first, options)
-    else
-      matches = []
+    end
+    
+    if entity.type == :document 
+      if records.size == 1
+        self.do_unserialize(
+        records.first, options)
+      else
+        raise Treat::Exception,
+        "More than one document matched" +
+        "your selector #{selector.inspect}."
+      end
+    elsif entity.type == :collection
+      collection = Treat::Entities::Collection.new
       records.each do |record|
-        matches << self.
+        collection << self.
         do_unserialize(record, options)
       end
-      matches
+      collection
     end
     
   end

data/lib/treat/workers/formatters/unserializers/xml.rb

@@ -35,13 +35,13 @@ module Treat::Workers::Formatters::Unserializers::XML
 
       id = nil; value = ''
       attributes = {}
-      dependencies = []
+      edges = []
       
       unless xml_reader.attributes.size == 0
         xml_reader.attributes.each_pair do |k,v|
           if k == 'id'
             id = v.to_i
-          elsif k == 'dependencies'
+          elsif k == 'edges'
             a = v.split('--')
             a.each do |b|
               c = b.split(';')
@@ -54,7 +54,7 @@ module Treat::Workers::Formatters::Unserializers::XML
                 end
 
                 target, type, directed, direction = *vals
-                dependencies << [
+                edges << [
                   target.to_i,
                   type,
                   (directed == 'true' ? true : false),
@@ -87,8 +87,8 @@ module Treat::Workers::Formatters::Unserializers::XML
         end
         current_element.features = attributes
         current_element.features = attributes
-        dependencies.each do |dependency|
-          target, type, directed, direction = *dependency
+        edges.each do |edge|
+          target, type, directed, direction = *edge
           current_element.link(target, type, directed, direction)
         end
       else

data/lib/treat/workers/formatters/visualizers/dot.rb

@@ -106,18 +106,18 @@ class Treat::Workers::Formatters::Visualizers::DOT
         string << "\n#{entity.parent.id} -- #{entity.id};"
       end
     end
-    # Dependencies.
-    if entity.has_dependencies?
-      entity.dependencies.each do |dependency|
+    # edges.
+    if entity.has_edges?
+      entity.edges.each do |edge|
         dir = ''
-        if dependency.directed == true
-          dir = dependency.direction == 1 ? 'forward' : 'back'
+        if edge.directed == true
+          dir = edge.direction == 1 ? 'forward' : 'back'
           dir = ",dir=#{dir}"
         else
           dir = ",dir=both"
         end
-        string << "\n#{entity.id} -- #{dependency.target}"
-        string << "[label=#{dependency.type}#{dir}]"
+        string << "\n#{entity.id} -- #{edge.target}"
+        string << "[label=#{edge.type}#{dir}]"
       end
     end
     # Recurse.

data/lib/treat/workers/learners/classifiers/id3.rb

@@ -11,8 +11,9 @@ class Treat::Workers::Learners::Classifiers::ID3
     
     if !@@classifiers[cl]
       dec_tree = DecisionTree::ID3Tree.new(
-      cl.labels.map { |l| l.to_s }, set.items, cl.question.default, 
-      cl.question.type)
+      cl.feature_labels.map { |l| l.to_s }, 
+      set.items.map { |i| i[:features]}, 
+      cl.question.default, cl.question.type)
       dec_tree.train
       @@classifiers[cl] = dec_tree
     else
@@ -20,7 +21,7 @@ class Treat::Workers::Learners::Classifiers::ID3
       dec_tree.graph('testingbitch')
     end
     dec_tree.predict(
-      cl.export_item(entity, false)
+      cl.export_features(entity, false)
     )
   end
   

data/lib/treat/workers/learners/classifiers/linear.rb

@@ -0,0 +1,53 @@
+class Treat::Workers::Learners::Classifiers::Linear
+  
+  require 'linear'
+  
+  @@classifiers = {}
+  
+  DefaultOptions = {
+    bias: 1,
+    eps: 0.1,
+    solver_type: MCSVM_CS
+  }
+  
+  def self.classify(entity, options = {})
+    
+    options = DefaultOptions.merge(options)
+    set = options[:training]
+    problem = set.problem
+    
+    if !@@classifiers[problem]
+      labels = problem.question.labels
+      unless labels
+        raise Treat::Exception,
+        "LibLinear requires that you provide the possible " +
+        "labels to assign to classification items when " +
+        "specifying the question."
+      end
+      param = LParameter.new
+      param.solver_type = options[:solver_type]
+      param.eps = options[:eps]
+      bias = options[:bias]
+      data = set.items.map do |item| 
+        self.array_to_hash(item[:features]) 
+      end
+      prob = LProblem.new(labels, data, bias)
+      @@classifiers[problem] = 
+      LModel.new(prob, param)
+    end
+
+    @@classifiers[problem].predict(
+    self.array_to_hash(problem.
+    export_features(entity, false)))
+    
+  end
+  
+  def self.array_to_hash(array)
+    hash = {}
+    0.upto(array.length - 1) do |i|
+      hash[i] = array[i]
+    end
+    hash
+  end
+  
+end

data/lib/treat/workers/learners/classifiers/mlp.rb

@@ -11,11 +11,11 @@ class Treat::Workers::Learners::Classifiers::MLP
     cl = set.problem
       
     if !@@mlps[cl]
-      net = Ai4r::NeuralNetwork::
-      Backpropagation.new([cl.labels.size, 3, 1])
+      net = Ai4r::NeuralNetwork::Backpropagation.new(
+      [cl.feature_labels.size, 3, 1])
       set.items.each do |item|
-        inputs = item[0..-2]
-        outputs = [item[-1]]
+        inputs = item[:features][0..-2]
+        outputs = [item[:features][-1]]
         net.train(inputs, outputs)
       end
       @@mlps[cl] = net
@@ -23,7 +23,7 @@ class Treat::Workers::Learners::Classifiers::MLP
       net = @@mlps[cl]
     end
     
-    net.eval(cl.export_item(entity, false))[0]
+    net.eval(cl.export_features(entity, false))[0]
     
   end
   

data/lib/treat/workers/learners/classifiers/svm.rb

@@ -0,0 +1,31 @@
+class Treat::Workers::Learners::Classifiers::SVM
+  
+  require 'svm'
+  
+  @@classifiers = {}
+  
+  def self.classify(entity, options = {})
+    
+    set = options[:training]
+    problem = set.problem
+    
+    if !@@classifiers[problem]
+      labels = problem.question.labels
+      unless labels
+        raise Treat::Exception,
+        "LibSVM requires that you provide the possible " +
+        "labels to assign to classification items when " +
+        "specifying the question."
+      end
+      data = set.items.map  { |item| item[:features] }
+      prob = Problem.new(labels, data)
+      param = Parameter.new(:kernel_type => LINEAR, :C => 10)
+      @@classifiers[problem] = Model.new(prob, param)
+    end
+
+    @@classifiers[problem].predict_probability(
+    problem.export_features(entity, false))[0]
+    
+  end
+  
+end

data/lib/treat/workers/lexicalizers/taggers/stanford.rb

@@ -53,8 +53,10 @@ class Treat::Workers::Lexicalizers::Taggers::Stanford
   def self.init_tagger(language)
     Treat::Loaders::Stanford.load(language)
     model = StanfordCoreNLP::Config::Models[:pos][language]
-    model = Treat.paths.models + 'stanford/' +
-    StanfordCoreNLP::Config::ModelFolders[:pos] + model
+    model_path = Treat.libraries.stanford.model_path || 
+    Treat.paths.models + 'stanford/'
+    model = model_path + StanfordCoreNLP::
+    Config::ModelFolders[:pos] + model
     @@taggers[language] ||=
     StanfordCoreNLP::MaxentTagger.new(model)
   end

data/lib/treat/workers/processors/parsers/enju.rb

@@ -38,7 +38,7 @@ module Treat::Workers::Processors::Parsers::Enju
     val = entity.to_s
     
     @@id_table = {}
-    @@dependencies_table = {}
+    @@edges_table = {}
     
     stdin, stdout = proc
     text, remove_last = valid_text(val)
@@ -63,7 +63,7 @@ module Treat::Workers::Processors::Parsers::Enju
     end
     
     link_heads(entity)
-    add_dependencies(entity)
+    add_edges(entity)
   end
   
   # Return the process running Enju.
@@ -102,12 +102,12 @@ module Treat::Workers::Processors::Parsers::Enju
         pd = cd
         next
       end
-      # Get and format attributes and dependencies.
+      # Get and format attributes and edges.
       attributes = reader.attributes
       id = attributes.delete('id')
-      new_attr = {}; dependencies = {}
+      new_attr = {}; edges = {}
       unless attributes.size == 0
-        new_attr, dependencies =
+        new_attr, edges =
         cleanup_attributes(reader.name, attributes)
       end
       # Create the appropriate entity for the
@@ -117,17 +117,17 @@ module Treat::Workers::Processors::Parsers::Enju
       when 'sentence'
         entity = Treat::Entities::Sentence.new('')
         @@id_table[id] = entity.id
-        @@dependencies_table[entity.id] = dependencies
+        @@edges_table[entity.id] = edges
         entity.features = new_attr
       when 'cons'
         entity = entity <<
         Treat::Entities::Phrase.new('')
         @@id_table[id] = entity.id
-        @@dependencies_table[entity.id] = dependencies
+        @@edges_table[entity.id] = edges
         entity.features = new_attr
       when 'tok'
         tmp_attributes = new_attr
-        tmp_dependencies = dependencies
+        tmp_edges = edges
       else
         current_value = reader.value.gsub(/\s+/, "")
         unless current_value.size == 0
@@ -136,7 +136,7 @@ module Treat::Workers::Processors::Parsers::Enju
           if entity.is_a?(Treat::Entities::Word)
             entity.features = tmp_attributes
             @@id_table[id] = entity.id
-            @@dependencies_table[entity.id] = tmp_dependencies
+            @@edges_table[entity.id] = tmp_edges
           else
             # Do something useful here
             entity.set :tag, 'SYM'
@@ -179,15 +179,15 @@ module Treat::Workers::Processors::Parsers::Enju
     end
   end
   
-  # Add dependencies a posteriori to a parsed entity.
-  def self.add_dependencies(entity2)
+  # Add edges a posteriori to a parsed entity.
+  def self.add_edges(entity2)
     
     entity2.each_entity(:word, :phrase) do |entity|
-      @@dependencies_table.each_pair do |id, dependencies|
-        next if dependencies.nil?
+      @@edges_table.each_pair do |id, edges|
+        next if edges.nil?
         entity = entity2.root.find(id)
         next if entity.nil?
-        dependencies.each_pair do |argument, type|
+        edges.each_pair do |argument, type|
           # Skip this argument if we 
           # don't know the target node.
           next if argument == 'unk'
@@ -205,7 +205,7 @@ module Treat::Workers::Processors::Parsers::Enju
   def self.cleanup_attributes(name, attributes)
     
     new_attr = {}
-    dependencies = {}
+    edges = {}
     pred = attributes.delete('pred')
     
     attributes.each_pair do |attribute2, value|
@@ -214,7 +214,7 @@ module Treat::Workers::Processors::Parsers::Enju
       
       if attribute == 'arg1' || 
         attribute == 'arg2'
-        dependencies[value] = pred
+        edges[value] = pred
         next
       end
       
@@ -256,7 +256,7 @@ module Treat::Workers::Processors::Parsers::Enju
       new_attr.delete :base
     end
     
-    return new_attr, dependencies
+    return new_attr, edges
   
   end
   

data/lib/treat/workers/processors/segmenters/punkt.rb

@@ -63,7 +63,9 @@ module Treat::Workers::Processors::Segmenters::Punkt
     if options[:model]
       model = options[:model]
     else
-      model = "#{Treat.paths.models}punkt/#{lang}.yaml"
+      model_path = Treat.libraries.punkt.model_path || 
+      Treat.paths.models + 'punkt/'
+      model = model_path + "#{lang}.yaml"
       unless File.readable?(model)
         raise Treat::Exception,
         "Could not get the language model " +

data/spec/collection.rb

@@ -1,4 +1,4 @@
-require_relative '../lib/treat'
+require_relative 'helper'
 
 describe Treat::Entities::Collection do
 
@@ -29,7 +29,7 @@ describe Treat::Entities::Collection do
         f = Treat.paths.spec + 'samples/test'
         c = Treat::Entities::Collection.build(f)
         c << Treat::Entities::Document.new
-        c.size.should eql 2
+        c.size.should eql 1
         FileUtils.rm_rf(f)
       end
     end
@@ -45,7 +45,7 @@ describe Treat::Entities::Collection do
         it "recursively searches the folder for " +
         "files and opens them into a collection of documents" do
           collection = Treat::Entities::Collection.build(@file)
-          collection.size.should eql 6
+          collection.size.should eql 5
         end
 
       end