modified to use mystery + predicted

utils/visualizations.py

@@ -10,7 +10,7 @@ import plotly.graph_objects as go
 from plotly.colors import sample_colorscale
 from gradio import update
 import re
-from utils.interp_space_utils import compute_clusters_style_representation_3, compute_clusters_g2v_representation, compute_precomputed_regions
+from utils.interp_space_utils import compute_clusters_style_representation_3, compute_clusters_g2v_representation, compute_precomputed_regions, compute_task_only_g2v_similarity
 from utils.llm_feat_utils import split_features
 from utils.gram2vec_feat_utils import get_shorthand, get_fullform
 from gram2vec.feature_locator import find_feature_spans
@@ -204,11 +204,12 @@ def load_interp_space(cfg):
 # Function to process G2V features and create display choices
 def format_g2v_features_for_display(g2v_features_with_scores):
     """
-    Convert G2V features with z-scores into display format for Gradio radio buttons.
+    Convert G2V features with a numeric score into display format for Gradio radio buttons.
+    The label uses S= for a generic similarity score (not Z).
     
     Args:
         g2v_features_with_scores: List of tuples like:
-            [('None', None), ('Feature Name', z_score), ...]
+            [('None', None), ('Feature Name', score), ...]
     
     Returns:
         tuple: (display_choices, original_values)
@@ -218,21 +219,21 @@ def format_g2v_features_for_display(g2v_features_with_scores):
     
     for item in g2v_features_with_scores:
         if len(item) == 2:
-            feature_name, z_score = item
+            feature_name, score = item
             
             # Handle None case
-            if feature_name == "None" or z_score is None:
+            if feature_name == "None" or score is None:
                 display_choices.append("None")
                 original_values.append("None")
             else:
                 # Convert numpy float to regular float if needed
-                if hasattr(z_score, 'item'):
-                    z_score = float(z_score.item())
+                if hasattr(score, 'item'):
+                    score = float(score.item())
                 else:
-                    z_score = float(z_score)
+                    score = float(score)
                 
-                # Create display string with z-score
-                display_string = f"{feature_name} | [Z={z_score:.2f}]"
+                # Create display string with similarity score
+                display_string = f"{feature_name}"
                 display_choices.append(display_string)
                 original_values.append(feature_name)
         else:
@@ -275,17 +276,18 @@ def handle_zoom(event_json, bg_proj, bg_lbls, clustered_authors_df, task_authors
     print(f"Task authors: {len(task_authors_df)}, Clustered authors: {len(clustered_authors_df)}")
     merged_authors_df = pd.concat([task_authors_df, clustered_authors_df])
     print(f"Merged authors DataFrame:\n{len(merged_authors_df)}")
-    #style_analysis_response = {'features': [], 'spans': []}
-    style_analysis_response = compute_clusters_style_representation_3(
-        background_corpus_df=merged_authors_df,
-        cluster_ids=visible_authors,
-        cluster_label_clm_name='authorID',
-    )
+    style_analysis_response = {'features': [], 'spans': []}
+    # style_analysis_response = compute_clusters_style_representation_3(
+    #     background_corpus_df=merged_authors_df,
+    #     cluster_ids=visible_authors,
+    #     cluster_label_clm_name='authorID',
+    # )
 
     llm_feats = ['None'] + style_analysis_response['features']
 
 
     merged_authors_df = pd.concat([task_authors_df, clustered_authors_df])
+    # Default: contrastive Gram2Vec features
     g2v_feats = compute_clusters_g2v_representation(
         background_corpus_df=merged_authors_df,
         author_ids=visible_authors,
@@ -293,6 +295,34 @@ def handle_zoom(event_json, bg_proj, bg_lbls, clustered_authors_df, task_authors
         features_clm_name='g2v_vector'
     )
 
+    # If both Mystery and the predicted candidate are inside the zoom, switch to task-only similarity
+    task_author_names = {'Mystery author', 'Candidate Author 1', 'Candidate Author 2', 'Candidate Author 3'}
+    visible_task_names = set(visible_authors).intersection(task_author_names)
+    predicted_in_visible = None
+    if 'predicted' in task_authors_df.columns:
+        preds = task_authors_df[task_authors_df['predicted'] == True]['authorID'].tolist()
+        if preds:
+            predicted_in_visible = preds[0] if preds[0] in visible_task_names else None
+
+    use_task_only = ('Mystery author' in visible_task_names) and (predicted_in_visible is not None)
+    if use_task_only:
+        print("[INFO] Using task-only Gram2Vec similarity (Mystery + Predicted candidate) within zoom")
+        try:
+            g2v_feats = compute_task_only_g2v_similarity(
+                background_corpus_df=merged_authors_df,
+                visible_author_ids=visible_authors,
+                features_clm_name='g2v_vector',
+                top_n=10,
+                require_spans=True
+            )
+            # g2v_feats already enforces spans for both authors; treat as final
+            filtered_g2v_feats = g2v_feats
+        except Exception as e:
+            print(f"[WARN] Task-only similarity failed, falling back to contrastive: {e}")
+            filtered_g2v_feats = None
+    else:
+        filtered_g2v_feats = None
+
     # ── Span-existence filter on task authors in the zoom ───────────────────
     # Keep only features that have at least one detected span in any of the
     # visible task authors' texts
@@ -305,16 +335,17 @@ def handle_zoom(event_json, bg_proj, bg_lbls, clustered_authors_df, task_authors
 
     task_texts = [_to_text(x) for x in visible_task_authors['fullText'].tolist()]
 
-    filtered_g2v_feats = []
-    for feat in g2v_feats:
-        try:
-            # `feat` is shorthand already (e.g., 'pos_bigrams:NOUN PROPN')
-            if any(find_feature_spans(txt, feat[0]) for txt in task_texts):
-                filtered_g2v_feats.append(feat)
-            else:
-                print(f"[INFO] Dropping G2V feature with no spans in task texts: {feat}")
-        except Exception as e:
-            print(f"[WARN] Error while checking spans for {feat}: {e}")
+    if filtered_g2v_feats is None:
+        filtered_g2v_feats = []
+        for feat in g2v_feats:
+            try:
+                # `feat` is shorthand already (e.g., 'pos_bigrams:NOUN PROPN')
+                if any(find_feature_spans(txt, feat[0]) for txt in task_texts):
+                    filtered_g2v_feats.append(feat)
+                else:
+                    print(f"[INFO] Dropping G2V feature with no spans in task texts: {feat}")
+            except Exception as e:
+                print(f"[WARN] Error while checking spans for {feat}: {e}")
 
     # Convert to human readable for display
     HR_g2v_list = []