s/confidence/probability

README.md

@@ -37,7 +37,7 @@ pp entities["person"]
 entities["person"].first.text
 # => "Tim Cook"
 
-entities["person"].first.confidence
+entities["person"].first.probability
 # => 92.4
 
 entities["person"].first.offsets
@@ -78,7 +78,7 @@ pp result
 result["sentiment"].label
 # => "negative"
 
-result["sentiment"].confidence
+result["sentiment"].probability
 # => 87.1
 ```
 

SPEC.md

@@ -46,7 +46,7 @@ result = model[text]
 result["sentiment"].label
 # => "negative"
 
-result["sentiment"].confidence
+result["sentiment"].probability
 # => 87.1
 ```
 

lib/gliner.rb

@@ -32,7 +32,7 @@ module Gliner
     :text_len
   )
 
-  Entity = Data.define(:index, :offsets, :text, :name, :confidence) do
+  Entity = Data.define(:index, :offsets, :text, :name, :probability) do
     def to_s = text.to_s
     def to_str = text.to_s
   end
@@ -43,7 +43,7 @@ def list
     end
   end
 
-  Label = Data.define(:label, :confidence) do
+  Label = Data.define(:label, :probability) do
     def to_s = label.to_s
     def to_str = label.to_s
   end

lib/gliner/classifier.rb

@@ -17,12 +17,12 @@ def classification_scores(logits, labels, label_positions, prepared)
       end
     end
 
-    def format_classification(scores, labels:, multi_label:, include_confidence:, cls_threshold:)
+    def format_classification(scores, labels:, multi_label:, include_probability:, cls_threshold:)
       label_scores = sorted_label_scores(scores, labels)
 
-      return format_multi_label(label_scores, cls_threshold, include_confidence) if multi_label
+      return format_multi_label(label_scores, cls_threshold, include_probability) if multi_label
 
-      format_single_label(label_scores.first, include_confidence)
+      format_single_label(label_scores.first, include_probability)
     end
 
     private
@@ -44,27 +44,27 @@ def sorted_label_scores(scores, labels)
         .sort_by { |(_label, score)| -score }
     end
 
-    def format_multi_label(label_scores, cls_threshold, include_confidence)
+    def format_multi_label(label_scores, cls_threshold, include_probability)
       chosen = labels_above_threshold(label_scores, cls_threshold)
 
       chosen
         .sort_by { |(_label, score)| -score }
-        .map { |label, score| format_label(label, score, include_confidence) }
+        .map { |label, score| format_label(label, score, include_probability) }
     end
 
     def labels_above_threshold(label_scores, threshold)
       above = label_scores.select { |_label, score| score >= threshold }
       above.empty? && label_scores.first ? [label_scores.first] : above
     end
 
-    def format_single_label(label_score, include_confidence)
+    def format_single_label(label_score, include_probability)
       label, score = label_score
 
-      format_label(label, score, include_confidence)
+      format_label(label, score, include_probability)
     end
 
-    def format_label(label, score, _include_confidence)
-      Gliner::Label.new(label: label, confidence: score * 100.0)
+    def format_label(label, score, _include_probability)
+      Gliner::Label.new(label: label, probability: score * 100.0)
     end
   end
 end

lib/gliner/model.rb

@@ -101,40 +101,40 @@ def json_task
 
     def extract_entities(text, entity_types, **options)
       threshold = options.fetch(:threshold, Gliner.config.threshold)
-      include_confidence = options.fetch(:include_confidence, false)
+      include_probability = options.fetch(:include_probability, false)
       include_spans = options.fetch(:include_spans, false)
 
       pipeline.execute(
         entity_task,
         text,
         entity_types,
         threshold: threshold,
-        include_confidence: include_confidence,
+        include_probability: include_probability,
         include_spans: include_spans
       )
     end
 
     def classify_text(text, tasks, **options)
-      include_confidence = options.fetch(:include_confidence, false)
+      include_probability = options.fetch(:include_probability, false)
       threshold = options[:threshold]
 
-      task_options = { include_confidence: include_confidence }
+      task_options = { include_probability: include_probability }
       task_options[:threshold] = threshold unless threshold.nil?
 
       classification_task.execute_all(pipeline, text, tasks, **task_options)
     end
 
     def extract_json(text, structures, **options)
       threshold = options.fetch(:threshold, Gliner.config.threshold)
-      include_confidence = options.fetch(:include_confidence, false)
+      include_probability = options.fetch(:include_probability, false)
       include_spans = options.fetch(:include_spans, false)
 
       json_task.execute_all(
         pipeline,
         text,
         structures,
         threshold: threshold,
-        include_confidence: include_confidence,
+        include_probability: include_probability,
         include_spans: include_spans
       )
     end

lib/gliner/span_extractor.rb

@@ -109,7 +109,7 @@ def format_span(span, _opts, label:, index:)
         offsets: [span.start, span.end],
         text: span.text,
         name: label&.to_s,
-        confidence: span.score * 100.0
+        probability: span.score * 100.0
       )
     end
 

lib/gliner/tasks/classification.rb

@@ -38,7 +38,7 @@ def needs_cls_logits?
       end
 
       def process_output(logits, parsed, prepared, options)
-        include_confidence = options.fetch(:include_confidence, false)
+        include_probability = options.fetch(:include_probability, false)
         threshold_override = options[:threshold]
         cls_threshold = threshold_override.nil? ? parsed[:cls_threshold] : threshold_override
 
@@ -47,7 +47,7 @@ def process_output(logits, parsed, prepared, options)
           scores,
           labels: parsed[:labels],
           multi_label: parsed[:multi_label],
-          include_confidence: include_confidence,
+          include_probability: include_probability,
           cls_threshold: cls_threshold
         )
       end

spec/gliner/model_spec.rb

@@ -14,15 +14,15 @@
         Gliner::Span.new(text: 'Cook', score: 0.7, start: 4, end: 8)
       ]
 
-      out = span_extractor.format_spans(spans, include_confidence: false, include_spans: false, label: 'person')
+      out = span_extractor.format_spans(spans, include_probability: false, include_spans: false, label: 'person')
       expect(out.map(&:text)).to eq(['Tim Cook'])
       expect(out.first).to eq(
         Gliner::Entity.new(
           index: 0,
           offsets: [0, 8],
           text: 'Tim Cook',
           name: 'person',
-          confidence: 90.0
+          probability: 90.0
         )
       )
     end
@@ -99,7 +99,7 @@
       }
 
       instances = structured_extractor
-        .build_structure_instances(parsed_fields, spans_by_label, include_confidence: false, include_spans: false)
+        .build_structure_instances(parsed_fields, spans_by_label, include_probability: false, include_spans: false)
 
       expect(instances.length).to eq(2)
       expect(instances[0]).to eq(
@@ -109,21 +109,21 @@
             offsets: [0, 5],
             text: 'Jan 5',
             name: 'date',
-            confidence: 90.0
+            probability: 90.0
           ),
           'merchant' => Gliner::Entity.new(
             index: 0,
             offsets: [7, 16],
             text: 'Starbucks',
             name: 'merchant',
-            confidence: 90.0
+            probability: 90.0
           ),
           'amount' => Gliner::Entity.new(
             index: 0,
             offsets: [17, 22],
             text: '$5.50',
             name: 'amount',
-            confidence: 90.0
+            probability: 90.0
           )
         })
       )
@@ -134,21 +134,21 @@
             offsets: [40, 45],
             text: 'Jan 6',
             name: 'date',
-            confidence: 80.0
+            probability: 80.0
           ),
           'merchant' => Gliner::Entity.new(
             index: 0,
             offsets: [47, 53],
             text: 'Amazon',
             name: 'merchant',
-            confidence: 70.0
+            probability: 70.0
           ),
           'amount' => Gliner::Entity.new(
             index: 0,
             offsets: [54, 61],
             text: '$156.99',
             name: 'amount',
-            confidence: 80.0
+            probability: 80.0
           )
         })
       )

spec/integration_spec.rb

@@ -47,7 +47,7 @@
 
         expect(sentiment).to be_a(Gliner::Label)
         expect(sentiment.label).to eq('negative')
-        expect(sentiment.confidence).to be_a(Float)
+        expect(sentiment.probability).to be_a(Float)
       end
     end