File size: 7,026 Bytes
c6556f2 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 | import os
import sys
import glob
import numpy as np
import pandas as pd
import matplotlib
matplotlib.use('Agg')
import matplotlib.pyplot as plt
sys.stdout.reconfigure(encoding='utf-8')
BASE_DIR = r"c:\Users\ahmed\.gemini\antigravity\scratch\Allignments\PHONEME_TIMINGS_EACH_RECITER"
def sanitize_filename(name):
replacements = {
':': '_colon_',
'/': '_slash_',
'\\': '_bslash_',
'*': '_star_',
'?': '_qmark_',
'"': '_quote_',
'<': '_lt_',
'>': '_gt_',
'|': '_pipe_'
}
s = str(name)
for char, rep in replacements.items():
s = s.replace(char, rep)
return s
def main():
print("=" * 60)
print("Updating Phase 3 with Custom Rules for Long Vowels (a:, i:, u:)")
print("=" * 60)
pooled_csv_path = os.path.join(BASE_DIR, "all_reciters_phase3_pooled_tokens.csv")
master_df = pd.read_csv(pooled_csv_path)
unique_phones = sorted(master_df['phone'].dropna().unique())
summary_rows = []
plots_dir = os.path.join(BASE_DIR, "Phase3_Unified_Plots")
os.makedirs(plots_dir, exist_ok=True)
for phone in unique_phones:
phone_str = str(phone)
sub = master_df[master_df['phone'] == phone]
t_ms = sub['duration_norm_ms'].values
n_samples = len(t_ms)
if n_samples == 0:
continue
p1_ms = np.percentile(t_ms, 1.0)
p2_5_ms = np.percentile(t_ms, 2.5)
q1_ms = np.percentile(t_ms, 25.0)
med_ms = np.median(t_ms)
q3_ms = np.percentile(t_ms, 75.0)
p97_5_ms = np.percentile(t_ms, 97.5)
p99_ms = np.percentile(t_ms, 99.0)
mean_ms = np.mean(t_ms)
std_ms = np.std(t_ms)
raw_min_ms = np.min(t_ms)
raw_max_ms = np.max(t_ms)
# APPLY CUSTOM USER RULES (Option A):
# 1. For a: no upper outlier filtering (use raw_max_ms)
if phone_str == 'a:':
p99_ms = raw_max_ms
p97_5_ms = raw_max_ms
# 2. For i: stop at P90 (3726.6 ms)
elif phone_str == 'i:':
p99_ms = np.percentile(t_ms, 90.0)
p97_5_ms = np.percentile(t_ms, 88.0)
# 3. For u: stop at P97 (3828.0 ms)
elif phone_str == 'u:':
p99_ms = np.percentile(t_ms, 97.0)
p97_5_ms = np.percentile(t_ms, 95.0)
p1_sec = p1_ms / 1000.0
p2_5_sec = p2_5_ms / 1000.0
q1_sec = q1_ms / 1000.0
med_sec = med_ms / 1000.0
q3_sec = q3_ms / 1000.0
p97_5_sec = p97_5_ms / 1000.0
p99_sec = p99_ms / 1000.0
mean_sec = mean_ms / 1000.0
std_sec = std_ms / 1000.0
raw_min_sec = raw_min_ms / 1000.0
raw_max_sec = raw_max_ms / 1000.0
summary_rows.append({
'phone': phone_str,
'total_tokens': n_samples,
'min_valid_p1_ms': round(p1_ms, 1),
'min_conservative_p2_5_ms': round(p2_5_ms, 1),
'q1_ms': round(q1_ms, 1),
'median_ms': round(med_ms, 1),
'mean_ms': round(mean_ms, 1),
'std_ms': round(std_ms, 1),
'q3_ms': round(q3_ms, 1),
'max_conservative_p97_5_ms': round(p97_5_ms, 1),
'max_valid_p99_ms': round(p99_ms, 1),
'raw_min_ms': round(raw_min_ms, 1),
'raw_max_ms': round(raw_max_ms, 1),
'min_valid_p1_sec': round(p1_sec, 3),
'min_conservative_p2_5_sec': round(p2_5_sec, 3),
'q1_sec': round(q1_sec, 3),
'median_sec': round(med_sec, 3),
'mean_sec': round(mean_sec, 3),
'std_sec': round(std_sec, 3),
'q3_sec': round(q3_sec, 3),
'max_conservative_p97_5_sec': round(p97_5_sec, 3),
'max_valid_p99_sec': round(p99_sec, 3),
'raw_min_sec': round(raw_min_sec, 3),
'raw_max_sec': round(raw_max_sec, 3)
})
fig, ax = plt.subplots(figsize=(9, 5.5), dpi=120)
plot_upper_limit = p99_ms * 1.2 if p99_ms > 0 else raw_max_ms
t_ms_clipped = t_ms[t_ms <= plot_upper_limit]
counts, bins, patches = ax.hist(
t_ms_clipped,
bins=40,
color='#1d3557',
edgecolor='white',
alpha=0.75,
label=f'Unified Tokens (N={n_samples:,})'
)
ax.axvline(med_ms, color='#e63946', linestyle='-', linewidth=2.2, label=f'Median: {med_ms:.1f} ms')
ax.axvline(mean_ms, color='#9b5de5', linestyle='--', linewidth=1.5, label=f'Mean: {mean_ms:.1f} ms')
ax.axvline(p1_ms, color='#2a9d8f', linestyle='-.', linewidth=1.8, label=f'Min Valid (P1): {p1_ms:.1f} ms')
ax.axvline(p99_ms, color='#e76f51', linestyle='-.', linewidth=1.8, label=f'Max Valid: {p99_ms:.1f} ms')
ax.axvline(q1_ms, color='#457b9d', linestyle=':', linewidth=1.4, label=f'Q1 (P25): {q1_ms:.1f} ms')
ax.axvline(q3_ms, color='#457b9d', linestyle=':', linewidth=1.4, label=f'Q3 (P75): {q3_ms:.1f} ms')
ax.axvspan(q1_ms, q3_ms, color='#a8dadc', alpha=0.25, label='Core Tier [Q1 - Q3]')
ax.set_title(f"Phase 3: Unified Duration Distribution for Phoneme [{phone_str}]\n(Pooled Across All 7 Imams | N = {n_samples:,} samples)",
fontsize=11, fontweight='bold', pad=10)
ax.set_xlabel("Normalized Duration (ms)", fontsize=10, fontweight='bold')
ax.set_ylabel("Token Count", fontsize=10, fontweight='bold')
ax.grid(axis='y', alpha=0.3, linestyle='--')
info_text = (
f"Phoneme: [{phone_str}]\n"
f"Samples: {n_samples:,}\n"
f"Valid Min: {p1_ms:.1f} ms\n"
f"Median: {med_ms:.1f} ms\n"
f"Valid Max: {p99_ms:.1f} ms\n"
f"Mean +/- Std: {mean_ms:.1f} +/- {std_ms:.1f} ms"
)
ax.text(0.97, 0.95, info_text, transform=ax.transAxes, fontsize=8.5,
verticalalignment='top', horizontalalignment='right',
bbox=dict(boxstyle='round,pad=0.5', facecolor='whitesmoke', edgecolor='#adb5bd', alpha=0.9))
ax.legend(loc='upper left', fontsize=8, framealpha=0.9)
fig.tight_layout()
safe_name = sanitize_filename(phone_str)
plot_out = os.path.join(plots_dir, f"{safe_name}_unified_histogram.png")
if os.path.exists(plot_out):
try:
os.remove(plot_out)
except Exception:
pass
try:
fig.savefig(plot_out)
except Exception as e:
import time
time.sleep(0.1)
fig.savefig(plot_out)
plt.close(fig)
summary_df = pd.DataFrame(summary_rows)
master_csv_path = os.path.join(BASE_DIR, "phase3_unified_phoneme_timings.csv")
summary_df.to_csv(master_csv_path, index=False, encoding='utf-8-sig')
print(f"\nPhase 3 Master Timing Table updated: {master_csv_path}")
print("Done!")
if __name__ == '__main__':
main() |