Adds confidence to suggested pattern and resource

JPrevost · JPrevost · commit 97876521fcdc · 2025-04-30T15:27:59.000-04:00
Why are these changes being introduced: * Storing values in the models rather than in the code better aligns with confidence from detectors Relevant ticket(s): * https://mitlibraries.atlassian.net/browse/TCO-91 How does this address that need: * Adds confidence fields for SuggestedPattern and SuggestedResources Document any side effects to this change: * It feels clunky to assign confidence to each Resource or Pattern. From a data model perspective if feels okay. From a practical standpoint I'm nervous the people managing Resources and Patterns may not know what value is appropriate.
diff --git a/app/dashboards/suggested_pattern_dashboard.rb b/app/dashboards/suggested_pattern_dashboard.rb
@@ -12,6 +12,7 @@ class SuggestedPatternDashboard < Administrate::BaseDashboard
   ATTRIBUTE_TYPES = {
     id: Field::Number,
     category: Field::BelongsTo,
+    confidence: Field::Number.with_options(decimals: 2),
     pattern: Field::String,
     shortcode: Field::String,
     title: Field::String,
@@ -37,6 +38,7 @@ class SuggestedPatternDashboard < Administrate::BaseDashboard
   SHOW_PAGE_ATTRIBUTES = %i[
     id
     category
+    confidence
     pattern
     shortcode
     title
@@ -50,6 +52,7 @@ class SuggestedPatternDashboard < Administrate::BaseDashboard
   # on the model's form (`new` and `edit`) pages.
   FORM_ATTRIBUTES = %i[
     category
+    confidence
     pattern
     shortcode
     title
diff --git a/app/dashboards/suggested_resource_dashboard.rb b/app/dashboards/suggested_resource_dashboard.rb
@@ -12,6 +12,7 @@ class SuggestedResourceDashboard < Administrate::BaseDashboard
   ATTRIBUTE_TYPES = {
     id: Field::Number,
     category: Field::BelongsTo,
+    confidence: Field::Number.with_options(decimals: 2),
     fingerprints: Field::HasMany,
     terms: Field::HasMany,
     title: Field::String,
@@ -30,14 +31,15 @@ class SuggestedResourceDashboard < Administrate::BaseDashboard
     title
     url
     terms
+    category
   ].freeze
 
   # SHOW_PAGE_ATTRIBUTES
   # an array of attributes that will be displayed on the model's show page.
   SHOW_PAGE_ATTRIBUTES = %i[
     id
     category
-    fingerprints
+    confidence
     terms
     title
     url
@@ -50,6 +52,7 @@ class SuggestedResourceDashboard < Administrate::BaseDashboard
   # on the model's form (`new` and `edit`) pages.
   FORM_ATTRIBUTES = %i[
     category
+    confidence
     title
     url
   ].freeze
diff --git a/app/models/detector/suggested_resource.rb b/app/models/detector/suggested_resource.rb
@@ -27,9 +27,10 @@ def self.full_term_match(phrase)
     # @note Multiple detections are irrelevant for this method. If _any_ match is found, a Detection record is created.
     #       The uniqueness contraint on Detection records would make multiple detections irrelevant.
     #
-    # @return Category
+    # @return Hash with keys :category and :confidence (or nil)
     def self.record(term)
       result = full_term_match(term.phrase)
+
       return unless result.any?
 
       Detection.find_or_create_by(
@@ -40,7 +41,13 @@ def self.record(term)
 
       return if result.empty?
 
-      result.first.category
+      # If a category hasn't been set, nil is better than a confidence with no category
+      return if result.first.category.blank?
+
+      {
+        category: result.first.category,
+        confidence: result.first.confidence
+      }
     end
   end
 end
diff --git a/app/models/detector/suggested_resource_pattern.rb b/app/models/detector/suggested_resource_pattern.rb
@@ -28,7 +28,8 @@ def check_patterns(phrase)
           shortcode: sp.shortcode,
           title: sp.title,
           url: sp.url,
-          category: sp.category
+          category: sp.category,
+          confidence: sp.confidence
         }
         @detections = sps
       end
@@ -40,7 +41,7 @@ def check_patterns(phrase)
     # @note There are multiple patterns within SuggestedPattern records. Each check is capable of generating
     #       a separate Detection record.
     #
-    # @return Category
+    # @return Hash with keys :category and :confidence (or nil)
     def self.record(term)
       sp = Detector::SuggestedResourcePattern.new(term.phrase)
 
@@ -53,7 +54,13 @@ def self.record(term)
       end
       return if sp.detections.empty?
 
-      sp.detections.first[:category]
+      # If a category hasn't been set, nil is better than a confidence with no category
+      return if sp.detections.first[:category].blank?
+
+      {
+        category: sp.detections.first[:category],
+        confidence: sp.detections.first[:confidence]
+      }
     end
   end
 end
diff --git a/app/models/suggested_pattern.rb b/app/models/suggested_pattern.rb
@@ -12,6 +12,7 @@
 #  created_at  :datetime         not null
 #  updated_at  :datetime         not null
 #  category_id :integer
+#  confidence  :float            default(0.9)
 #
 class SuggestedPattern < ApplicationRecord
   validates :title, presence: true
diff --git a/app/models/suggested_resource.rb b/app/models/suggested_resource.rb
@@ -10,6 +10,7 @@
 #  created_at  :datetime         not null
 #  updated_at  :datetime         not null
 #  category_id :integer
+#  confidence  :float            default(0.9)
 #
 class SuggestedResource < ApplicationRecord
   has_many :terms, dependent: :nullify
diff --git a/app/models/term.rb b/app/models/term.rb
@@ -129,9 +129,21 @@ def retrieve_detection_scores
     # The detections.scores method returns data like [{3=>0.91}, {1=>0.1}] and [{3=>0.95}]
     raw = detections.current.flat_map(&:scores)
     # raw looks like [{3=>0.91}, {1=>0.1}, {3=>0.95}]
-    raw << { @suggested_pattern_category.id => 0.9 } if @suggested_pattern_category.present?
-    raw << { @suggested_resource_category.id => 0.9 } if @suggested_resource_category.present?
+
+    raw << suggested_pattern_score if @suggested_pattern_category.present?
+
+    raw << suggested_resource_score if @suggested_resource_category.present?
 
     raw.group_by { |h| h.keys.first }.map { |k, v| { k => v.map { |h| h.values.first } } }
   end
+
+  # Retrieve category and score for suggested patterns
+  def suggested_pattern_score
+    { @suggested_pattern_category[:category].id => @suggested_pattern_category[:confidence] }
+  end
+
+  # Retrieve category and score for suggested resources
+  def suggested_resource_score
+    { @suggested_resource_category[:category].id => @suggested_resource_category[:confidence] }
+  end
 end
diff --git a/db/migrate/20250429202816_add_confidence_to_suggested_patterns_and_resources.rb b/db/migrate/20250429202816_add_confidence_to_suggested_patterns_and_resources.rb
@@ -0,0 +1,6 @@
+class AddConfidenceToSuggestedPatternsAndResources < ActiveRecord::Migration[7.2]
+  def change
+    add_column :suggested_patterns, :confidence, :float, default: 0.9
+    add_column :suggested_resources, :confidence, :float, default: 0.9
+  end
+end
diff --git a/db/schema.rb b/db/schema.rb
diff --git a/test/fixtures/fingerprints.yml b/test/fixtures/fingerprints.yml
@@ -57,3 +57,6 @@ nobel_laureate:
 
 astm:
   value: 'astm 1'
+
+suggested_resource_with_category:
+  value: 'are categories cool'
diff --git a/test/fixtures/suggested_patterns.yml b/test/fixtures/suggested_patterns.yml
@@ -10,6 +10,7 @@
 #  created_at  :datetime         not null
 #  updated_at  :datetime         not null
 #  category_id :integer
+#  confidence  :float            default(0.9)
 #
 
 astm:
diff --git a/test/fixtures/suggested_resources.yml b/test/fixtures/suggested_resources.yml
@@ -8,6 +8,7 @@
 #  created_at  :datetime         not null
 #  updated_at  :datetime         not null
 #  category_id :integer
+#  confidence  :float            default(0.9)
 #
 
 
@@ -27,3 +28,8 @@ web_of_knowledge:
 nobel_laureate:
   title: Professor Moungi Bawendi
   url: https://news.mit.edu/2023/mit-chemist-moungi-bawendi-shares-nobel-prize-chemistry-1004
+
+suggested_resource_with_category:
+  title: Categories are cool
+  url: https://example.com/categories-are-cool
+  category: transactional
diff --git a/test/fixtures/terms.yml b/test/fixtures/terms.yml
@@ -86,3 +86,8 @@ nobel_laureate:
 astm:
   phrase: 'astm 1'
   fingerprint: astm
+
+suggested_resource_with_category:
+  phrase: 'categories are cool'
+  fingerprint: suggested_resource_with_category
+  suggested_resource: suggested_resource_with_category
diff --git a/test/models/suggested_pattern_test.rb b/test/models/suggested_pattern_test.rb
@@ -12,6 +12,7 @@
 #  created_at  :datetime         not null
 #  updated_at  :datetime         not null
 #  category_id :integer
+#  confidence  :float            default(0.9)
 #
 require 'test_helper'
 
diff --git a/test/models/suggested_resource_test.rb b/test/models/suggested_resource_test.rb
@@ -10,6 +10,7 @@
 #  created_at  :datetime         not null
 #  updated_at  :datetime         not null
 #  category_id :integer
+#  confidence  :float            default(0.9)
 #
 require 'test_helper'
 
diff --git a/test/models/term_test.rb b/test/models/term_test.rb
@@ -300,6 +300,70 @@ class TermTest < ActiveSupport::TestCase
     end
   end
 
+  test 'calculate_categorization includes SuggestedResources with categories set' do
+    sr = suggested_resources('suggested_resource_with_category')
+    t = terms('suggested_resource_with_category')
+
+    # confirm a category was set
+    assert_equal(categories('transactional'), sr.category)
+
+    # Before categorization
+    categorization_count = Categorization.count
+
+    t.calculate_categorizations
+
+    # Expect categorization count to increase
+    assert_equal categorization_count + 1, Categorization.count
+  end
+
+  test 'calculate_categorization skips SuggestedResources with categories not set' do
+    sr = suggested_resources('nobel_laureate')
+    t = terms('nobel_laureate')
+
+    # confirm no category was set
+    assert_nil(sr.category)
+
+    # Before categorization
+    categorization_count = Categorization.count
+
+    t.calculate_categorizations
+
+    # Expect no change in categorization count
+    assert_equal categorization_count, Categorization.count
+  end
+
+  test 'calculate_categorization includes SuggestedPatterns with categories set' do
+    sp = suggested_patterns('iso')
+    t = Term.find_or_create_by(phrase: 'iso 1234')
+
+    # confirm a category was set
+    assert_equal(categories('transactional'), sp.category)
+
+    # Before categorization
+    categorization_count = Categorization.count
+
+    t.calculate_categorizations
+
+    # Expect categorization count to increase
+    assert_equal categorization_count + 1, Categorization.count
+  end
+
+  test 'calculate_categorization skips SuggestedPatterns with categories not set' do
+    sp = suggested_patterns('astm')
+    t = terms('astm')
+
+    # confirm no category was set
+    assert_nil(sp.category)
+
+    # Before categorization
+    categorization_count = Categorization.count
+
+    t.calculate_categorizations
+
+    # Expect no change in categorization count
+    assert_equal categorization_count, Categorization.count
+  end
+
   test 'categorized scope returns an active record relation' do
     assert_kind_of ActiveRecord::Relation, Term.categorized
   end

Original file line number	Diff line number	Diff line change
`@@ -12,6 +12,7 @@`
`12`	`12`	`# created_at :datetime not null`
`13`	`13`	`# updated_at :datetime not null`
`14`	`14`	`# category_id :integer`
	`15`	`+# confidence :float default(0.9)`
`15`	`16`	`#`
`16`	`17`	`class SuggestedPattern < ApplicationRecord`
`17`	`18`	`validates :title, presence: true`
Original file line number	Diff line number	Diff line change
`@@ -10,6 +10,7 @@`
`10`	`10`	`# created_at :datetime not null`
`11`	`11`	`# updated_at :datetime not null`
`12`	`12`	`# category_id :integer`
	`13`	`+# confidence :float default(0.9)`
`13`	`14`	`#`
`14`	`15`	`class SuggestedResource < ApplicationRecord`
`15`	`16`	`has_many :terms, dependent: :nullify`