SamanthaStorm commited on
Commit
bd4971f
·
verified ·
1 Parent(s): 28fc37c

Update app.py

Browse files
Files changed (1) hide show
  1. app.py +4 -2
app.py CHANGED
@@ -71,7 +71,7 @@ def calculate_abuse_level(scores, thresholds, motif_hits=None):
71
  # Boost score if high-risk motifs were detected
72
  motif_hits = motif_hits or []
73
  if any(label in motif_hits for label in {"physical_threat", "suicidal_threat", "extreme_control"}):
74
- base_score = max(base_score, 60.0) # Push to "Severe / Harmful Pattern Present"
75
 
76
  return base_score
77
 
@@ -116,7 +116,9 @@ def analyze_messages(input_text, risk_flags):
116
  high_risk_context = any(flag in contextual_flags for flag in critical_flags)
117
 
118
  non_abusive_score = scores[LABELS.index('non_abusive')]
119
- if non_abusive_score > adjusted_thresholds['non_abusive']:
 
 
120
  return "This message is classified as non-abusive."
121
 
122
  abuse_level = calculate_abuse_level(scores, adjusted_thresholds, motif_hits=[label for label, _ in matched_phrases])
 
71
  # Boost score if high-risk motifs were detected
72
  motif_hits = motif_hits or []
73
  if any(label in motif_hits for label in {"physical_threat", "suicidal_threat", "extreme_control"}):
74
+ base_score = max(base_score, 75.0) # Push to "Severe / Harmful Pattern Present"
75
 
76
  return base_score
77
 
 
116
  high_risk_context = any(flag in contextual_flags for flag in critical_flags)
117
 
118
  non_abusive_score = scores[LABELS.index('non_abusive')]
119
+ non_abusive_confident = non_abusive_score > adjusted_thresholds['non_abusive']
120
+
121
+ if non_abusive_confident and danger_flag_count == 0 and not matched_phrases:
122
  return "This message is classified as non-abusive."
123
 
124
  abuse_level = calculate_abuse_level(scores, adjusted_thresholds, motif_hits=[label for label, _ in matched_phrases])