rm: eval_wav

.ipynb_checkpoints/app-checkpoint.py

@@ -0,0 +1,100 @@
+import torchaudio as ta
+import gradio as gr
+import random
+
+
+def login(username, state):
+    state["username"] = username
+        
+    idx_idx = state["order_list"][state["count"]] // 4
+    idx = state["idx_list"][idx_idx]
+    which = state["order_list"][state["count"]] % 4
+    
+    file = state["filepath"][idx][0]
+    file = file.strip().split("/")
+    if which == 0:
+        file[-1] = "vits270/" + file[-1]
+    elif which == 1:
+        file[-1] = "vits002/" + file[-1]
+    elif which == 2:
+        file[-1] = "grad_tts002/" + file[-1]
+    elif which == 3:
+        file[-1] = "GT/" + file[-1]
+    
+    file = "/".join(file)
+    
+    audio, _ = ta.load(file)
+    
+    return (f"{state['count']+1}回目の評価です．", (22050, audio.data.cpu().numpy().squeeze()), state,
+            gr.update(visible=False), gr.update(visible=False), gr.update(visible=True), gr.update(visible=True))
+
+def func(evaluation, state):
+    idx = state["order_list"][state["count"]] // 4
+    which = state["order_list"][state["count"]] % 4
+    
+    with open('./log/{}.log'.format(state['username']), 'a') as log:
+        log.write("{}|{}|{}\n".format(which, idx, evaluation))
+        
+    state["count"] += 1
+    
+    if state['count']==40:
+        return ("以上で音声評価は終了になります．下記のパスワードをランサーズの備考欄に記入することで作業の完了を確認します．パスワード: fhaofkeofu", gr.update(visible=False),
+                state, gr.update(visible=False), gr.update(visible=False))
+        
+    idx = state["order_list"][state["count"]] // 4
+    which = state["order_list"][state["count"]] % 4
+    
+    file = state["filepath"][idx][0]
+    file = file.strip().split("/")
+    if which == 0:
+        file[-1] = "vits270/" + file[-1]
+    elif which == 1:
+        file[-1] = "vits002/" + file[-1]
+    elif which == 2:
+        file[-1] = "grad_tts002/" + file[-1]
+    elif which == 3:
+        file[-1] = "GT/" + file[-1]
+    
+    file = "/".join(file)
+    
+    audio, _ = ta.load(file)
+    
+    return (f"{state['count']+1}回目の評価です．", (22050, audio.data.cpu().numpy().squeeze()), state,
+            gr.update(visible=True), gr.update(visible=True))
+
+with gr.Blocks() as demo:
+    gr.Markdown(
+    """
+    ## 主観的音声評価
+    これから右に音声が表示されますので，その音声の品質を
+    [1:とても悪い, 2:悪い, 3:普通, 4:良い, 5:とても良い]で評価してください．
+    """
+    )
+    with open("./text_expressive.txt", encoding='utf-8') as f:
+        filepath = [line.strip().split("|") for line in f]
+        
+    idx_list = random.sample(range(len(filepath)), k=10)
+    order_list = list(range(0,40))
+    random.shuffle(order_list)
+    
+    state = gr.State({
+        "username": None,
+        "count": 0,
+        "filepath": filepath,
+        "idx_list": idx_list,
+        "order_list": order_list
+    })
+    with gr.Row():
+        with gr.Column():
+            tb_username = gr.Textbox(label="あなたの名前を入力してください")
+            button_login = gr.Button(value="音声評価を始める")
+        with gr.Column():
+            message = gr.Markdown()
+            audioA = gr.Audio(label="評価対象の音声")
+            evaluation = gr.Slider(1, 5, value=3, step=1, label="1:とても悪い, 2:悪い, 3:普通, 4:良い, 5:とても良い", visible=False)
+            button = gr.Button(value="評価を送信する", visible=False)
+
+    button_login.click(login, [tb_username, state], [message, audioA, state, tb_username, button_login, evaluation, button])
+    button.click(func, [evaluation, state], [message, audioA, state, evaluation, button])
+
+demo.launch()

app.py

@@ -0,0 +1,100 @@
+import torchaudio as ta
+import gradio as gr
+import random
+
+
+def login(username, state):
+    state["username"] = username
+        
+    idx_idx = state["order_list"][state["count"]] // 4
+    idx = state["idx_list"][idx_idx]
+    which = state["order_list"][state["count"]] % 4
+    
+    file = state["filepath"][idx][0]
+    file = file.strip().split("/")
+    if which == 0:
+        file[-1] = "vits270/" + file[-1]
+    elif which == 1:
+        file[-1] = "vits002/" + file[-1]
+    elif which == 2:
+        file[-1] = "grad_tts002/" + file[-1]
+    elif which == 3:
+        file[-1] = "GT/" + file[-1]
+    
+    file = "/".join(file)
+    
+    audio, _ = ta.load(file)
+    
+    return (f"{state['count']+1}回目の評価です．", (22050, audio.data.cpu().numpy().squeeze()), state,
+            gr.update(visible=False), gr.update(visible=False), gr.update(visible=True), gr.update(visible=True))
+
+def func(evaluation, state):
+    idx = state["order_list"][state["count"]] // 4
+    which = state["order_list"][state["count"]] % 4
+    
+    with open('./log/{}.log'.format(state['username']), 'a') as log:
+        log.write("{}|{}|{}\n".format(which, idx, evaluation))
+        
+    state["count"] += 1
+    
+    if state['count']==40:
+        return ("以上で音声評価は終了になります．下記のパスワードをランサーズの備考欄に記入することで作業の完了を確認します．パスワード: fhaofkeofu", gr.update(visible=False),
+                state, gr.update(visible=False), gr.update(visible=False))
+        
+    idx = state["order_list"][state["count"]] // 4
+    which = state["order_list"][state["count"]] % 4
+    
+    file = state["filepath"][idx][0]
+    file = file.strip().split("/")
+    if which == 0:
+        file[-1] = "vits270/" + file[-1]
+    elif which == 1:
+        file[-1] = "vits002/" + file[-1]
+    elif which == 2:
+        file[-1] = "grad_tts002/" + file[-1]
+    elif which == 3:
+        file[-1] = "GT/" + file[-1]
+    
+    file = "/".join(file)
+    
+    audio, _ = ta.load(file)
+    
+    return (f"{state['count']+1}回目の評価です．", (22050, audio.data.cpu().numpy().squeeze()), state,
+            gr.update(visible=True), gr.update(visible=True))
+
+with gr.Blocks() as demo:
+    gr.Markdown(
+    """
+    ## 主観的音声評価
+    これから右に音声が表示されますので，その音声の品質を
+    [1:とても悪い, 2:悪い, 3:普通, 4:良い, 5:とても良い]で評価してください．
+    """
+    )
+    with open("./text_expressive.txt", encoding='utf-8') as f:
+        filepath = [line.strip().split("|") for line in f]
+        
+    idx_list = random.sample(range(len(filepath)), k=10)
+    order_list = list(range(0,40))
+    random.shuffle(order_list)
+    
+    state = gr.State({
+        "username": None,
+        "count": 0,
+        "filepath": filepath,
+        "idx_list": idx_list,
+        "order_list": order_list
+    })
+    with gr.Row():
+        with gr.Column():
+            tb_username = gr.Textbox(label="あなたの名前を入力してください")
+            button_login = gr.Button(value="音声評価を始める")
+        with gr.Column():
+            message = gr.Markdown()
+            audioA = gr.Audio(label="評価対象の音声")
+            evaluation = gr.Slider(1, 5, value=3, step=1, label="1:とても悪い, 2:悪い, 3:普通, 4:良い, 5:とても良い", visible=False)
+            button = gr.Button(value="評価を送信する", visible=False)
+
+    button_login.click(login, [tb_username, state], [message, audioA, state, tb_username, button_login, evaluation, button])
+    button.click(func, [evaluation, state], [message, audioA, state, evaluation, button])
+
+demo.launch()

text_cloud.py

@@ -0,0 +1,162 @@
+import os
+import pyopenjtalk
+import re
+
+out_path = './text_cloud.txt'
+output = []
+j=0
+in_path = '/mnt/vol21/i21_mkazuki/text/transcripts_utf8.txt'
+for i in range(1,101):
+    if i in [1,3,5,6,9,11,12,13,20,21,22,23,28,31,32,33,34,37,41,42,44,45,46,47,48,49,50,52,54,68,70,71,73,74,75,76,77,78,79,80,81,86,87,88,89,97,98,99,100]:
+        continue
+    with open(in_path) as f:
+        lines = f.readlines()
+        for line in lines:
+            strs = line.split(':')
+            if not (os.path.isfile('/mnt/nfs-mnj-archive-03/group/creative/i21_mkazuki/wav/jvs'+format(i, '03')+'/parallel100/'+strs[0]+'.wav')):
+                continue
+            strs[1] = pyopenjtalk.g2p(strs[1])
+            strs[1] = strs[1].replace('pau',',')
+            
+    nonpara_path = '/mnt/nfs-mnj-archive-03/group/creative/i21_mkazuki/wav/jvs_ver1/jvs'+format(i, '03')+'/nonpara30/transcripts_utf8.txt'
+    with open(nonpara_path) as f:
+        lines = f.readlines()
+        for line in lines:
+            strs = line.split(':')
+            if not (os.path.isfile('/mnt/nfs-mnj-archive-03/group/creative/i21_mkazuki/wav/jvs'+format(i, '03')+'/nonpara30/'+strs[0]+'.wav')):
+                continue
+            strs[1] = pyopenjtalk.g2p(strs[1])
+            strs[1] = strs[1].replace('pau',',')
+            
+            if j==0:
+                output.append('/home/mkazuki/eval_wav/'+strs[0]+'.wav|'+strs[1]+'|'+str(j)+'\n')
+            
+    whisper_path = '/mnt/nfs-mnj-archive-03/group/creative/i21_mkazuki/wav/jvs_ver1/jvs'+format(i, '03')+'/whisper10/transcripts_utf8.txt'
+    with open(whisper_path) as f:
+        lines = f.readlines()
+        for line in lines:
+            strs = line.split(':')
+            if not (os.path.isfile('/mnt/nfs-mnj-archive-03/group/creative/i21_mkazuki/wav/jvs'+format(i, '03')+'/whisper10/'+strs[0]+'.wav')):
+                continue
+            strs[1] = pyopenjtalk.g2p(strs[1])
+            strs[1] = strs[1].replace('pau',',')
+            
+            if j==0:
+                output.append('/home/mkazuki/eval_wav/whisper_'+strs[0]+'.wav|'+strs[1]+'|'+str(j)+'\n')
+    j+=1
+    
+k=1
+
+with open(in_path) as f:
+    lines = f.readlines()
+    for line in lines:
+        strs = line.split(':')
+        strs[1] = pyopenjtalk.g2p(strs[1])
+        strs[1] = strs[1].replace('pau',',')
+        
+        if k>95:
+            output.append('/home/mkazuki/eval_wav/tsuchiya_normal_'+format(k, '03')+'.wav|'+strs[1]+'|'+str(j)+'\n')
+        k+=1
+        
+k=1        
+with open(in_path) as f:
+    lines = f.readlines()
+    for line in lines:
+        strs = line.split(':')
+        strs[1] = pyopenjtalk.g2p(strs[1])
+        strs[1] = strs[1].replace('pau',',')
+        
+        if k>95:
+            output.append('/home/mkazuki/eval_wav/tsuchiya_happy_'+format(k, '03')+'.wav|'+strs[1]+'|'+str(j)+'|2'+'\n')
+
+        k+=1
+        
+k=1        
+with open(in_path) as f:
+    lines = f.readlines()
+    for line in lines:
+        strs = line.split(':')
+        strs[1] = pyopenjtalk.g2p(strs[1])
+        strs[1] = strs[1].replace('pau',',')
+        
+        if k>95:
+            output.append('/home/mkazuki/eval_wav/tsuchiya_angry_'+format(k, '03')+'.wav|'+strs[1]+'|'+str(j)+'|3'+'\n')
+        k+=1    
+j+=1
+k=1
+with open(in_path) as f:
+    lines = f.readlines()
+    for line in lines:
+        strs = line.split(':')
+        strs[1] = pyopenjtalk.g2p(strs[1])
+        strs[1] = strs[1].replace('pau',',')
+        
+        if k>95:
+            output.append('/home/mkazuki/eval_wav/uemura_normal_'+format(k, '03')+'.wav|'+strs[1]+'|'+str(j)+'\n')
+        k+=1
+        
+k=1        
+with open(in_path) as f:
+    lines = f.readlines()
+    for line in lines:
+        strs = line.split(':')
+        strs[1] = pyopenjtalk.g2p(strs[1])
+        strs[1] = strs[1].replace('pau',',')
+        
+        if k>95:
+            output.append('/home/mkazuki/eval_wav/uemura_happy_'+format(k, '03')+'.wav|'+strs[1]+'|'+str(j)+'|2'+'\n')
+
+        k+=1
+        
+k=1        
+with open(in_path) as f:
+    lines = f.readlines()
+    for line in lines:
+        strs = line.split(':')
+        strs[1] = pyopenjtalk.g2p(strs[1])
+        strs[1] = strs[1].replace('pau',',')
+        
+        if k>95:
+            output.append('/home/mkazuki/eval_wav/uemura_angry_'+format(k, '03')+'.wav|'+strs[1]+'|'+str(j)+'|3'+'\n')
+        k+=1    
+
+j+=1
+k=1
+with open(in_path) as f:
+    lines = f.readlines()
+    for line in lines:
+        strs = line.split(':')
+        strs[1] = pyopenjtalk.g2p(strs[1])
+        strs[1] = strs[1].replace('pau',',')
+        
+        if k>95:
+            output.append('/home/mkazuki/eval_wav/fujitou_normal_'+format(k, '03')+'.wav|'+strs[1]+'|'+str(j)+'\n')
+        k+=1
+        
+k=1        
+with open(in_path) as f:
+    lines = f.readlines()
+    for line in lines:
+        strs = line.split(':')
+        strs[1] = pyopenjtalk.g2p(strs[1])
+        strs[1] = strs[1].replace('pau',',')
+        
+        if k>95:
+            output.append('/home/mkazuki/eval_wav/fujitou_happy_'+format(k, '03')+'.wav|'+strs[1]+'|'+str(j)+'|2'+'\n')
+        k+=1
+        
+k=1        
+with open(in_path) as f:
+    lines = f.readlines()
+    for line in lines:
+        strs = line.split(':')
+        strs[1] = pyopenjtalk.g2p(strs[1])
+        strs[1] = strs[1].replace('pau',',')
+        
+        if k>95:
+            output.append('/home/mkazuki/eval_wav/fujitou_angry_'+format(k, '03')+'.wav|'+strs[1]+'|'+str(j)+'|3'+'\n')
+        k+=1    
+        
+        
+with open(out_path, 'w') as f:
+    f.writelines(output)

text_expressive.py

@@ -0,0 +1,151 @@
+import os
+import pyopenjtalk
+import re
+
+out_path = './text_expressive.txt'
+output = []
+j=0
+in_path = '/mnt/vol21/i21_mkazuki/text/transcripts_utf8.txt'
+for i in range(1,101):
+    if i in [1,3,5,6,9,11,12,13,20,21,22,23,28,31,32,33,34,37,41,42,44,45,46,47,48,49,50,52,54,68,70,71,73,74,75,76,77,78,79,80,81,86,87,88,89,97,98,99,100]:
+        continue
+    with open(in_path) as f:
+        lines = f.readlines()
+        for line in lines:
+            strs = line.split(':')
+            if not (os.path.isfile('/mnt/nfs-mnj-archive-03/group/creative/i21_mkazuki/wav/jvs'+format(i, '03')+'/parallel100/'+strs[0]+'.wav')):
+                continue
+            strs[1] = pyopenjtalk.g2p(strs[1])
+            strs[1] = strs[1].replace('pau',',')
+            
+    nonpara_path = '/mnt/nfs-mnj-archive-03/group/creative/i21_mkazuki/wav/jvs_ver1/jvs'+format(i, '03')+'/nonpara30/transcripts_utf8.txt'
+    with open(nonpara_path) as f:
+        lines = f.readlines()
+        for line in lines:
+            strs = line.split(':')
+            if not (os.path.isfile('/mnt/nfs-mnj-archive-03/group/creative/i21_mkazuki/wav/jvs'+format(i, '03')+'/nonpara30/'+strs[0]+'.wav')):
+                continue
+            strs[1] = pyopenjtalk.g2p(strs[1])
+            strs[1] = strs[1].replace('pau',',')
+            
+            
+    whisper_path = '/mnt/nfs-mnj-archive-03/group/creative/i21_mkazuki/wav/jvs_ver1/jvs'+format(i, '03')+'/whisper10/transcripts_utf8.txt'
+    with open(whisper_path) as f:
+        lines = f.readlines()
+        for line in lines:
+            strs = line.split(':')
+            if not (os.path.isfile('/mnt/nfs-mnj-archive-03/group/creative/i21_mkazuki/wav/jvs'+format(i, '03')+'/whisper10/'+strs[0]+'.wav')):
+                continue
+            strs[1] = pyopenjtalk.g2p(strs[1])
+            strs[1] = strs[1].replace('pau',',')
+            
+            if j==0:
+                output.append('./eval_wav/whisper_'+strs[0]+'.wav|'+strs[1]+'|'+str(j)+'\n')
+    j+=1
+    
+k=1
+
+with open(in_path) as f:
+    lines = f.readlines()
+    for line in lines:
+        strs = line.split(':')
+        strs[1] = pyopenjtalk.g2p(strs[1])
+        strs[1] = strs[1].replace('pau',',')
+        
+        
+k=1        
+with open(in_path) as f:
+    lines = f.readlines()
+    for line in lines:
+        strs = line.split(':')
+        strs[1] = pyopenjtalk.g2p(strs[1])
+        strs[1] = strs[1].replace('pau',',')
+        
+        if k>95:
+            output.append('./eval_wav/tsuchiya_happy_'+format(k, '03')+'.wav|'+strs[1]+'|'+str(j)+'|2'+'\n')
+
+        k+=1
+        
+k=1        
+with open(in_path) as f:
+    lines = f.readlines()
+    for line in lines:
+        strs = line.split(':')
+        strs[1] = pyopenjtalk.g2p(strs[1])
+        strs[1] = strs[1].replace('pau',',')
+        
+        if k>95:
+            output.append('./eval_wav/tsuchiya_angry_'+format(k, '03')+'.wav|'+strs[1]+'|'+str(j)+'|3'+'\n')
+        k+=1    
+j+=1
+k=1
+with open(in_path) as f:
+    lines = f.readlines()
+    for line in lines:
+        strs = line.split(':')
+        strs[1] = pyopenjtalk.g2p(strs[1])
+        strs[1] = strs[1].replace('pau',',')
+        
+        
+k=1        
+with open(in_path) as f:
+    lines = f.readlines()
+    for line in lines:
+        strs = line.split(':')
+        strs[1] = pyopenjtalk.g2p(strs[1])
+        strs[1] = strs[1].replace('pau',',')
+        
+        if k>95:
+            output.append('./eval_wav/uemura_happy_'+format(k, '03')+'.wav|'+strs[1]+'|'+str(j)+'|2'+'\n')
+
+        k+=1
+        
+k=1        
+with open(in_path) as f:
+    lines = f.readlines()
+    for line in lines:
+        strs = line.split(':')
+        strs[1] = pyopenjtalk.g2p(strs[1])
+        strs[1] = strs[1].replace('pau',',')
+        
+        if k>95:
+            output.append('./eval_wav/uemura_angry_'+format(k, '03')+'.wav|'+strs[1]+'|'+str(j)+'|3'+'\n')
+        k+=1    
+
+j+=1
+k=1
+with open(in_path) as f:
+    lines = f.readlines()
+    for line in lines:
+        strs = line.split(':')
+        strs[1] = pyopenjtalk.g2p(strs[1])
+        strs[1] = strs[1].replace('pau',',')
+        
+        
+k=1        
+with open(in_path) as f:
+    lines = f.readlines()
+    for line in lines:
+        strs = line.split(':')
+        strs[1] = pyopenjtalk.g2p(strs[1])
+        strs[1] = strs[1].replace('pau',',')
+        
+        if k>95:
+            output.append('./eval_wav/fujitou_happy_'+format(k, '03')+'.wav|'+strs[1]+'|'+str(j)+'|2'+'\n')
+        k+=1
+        
+k=1        
+with open(in_path) as f:
+    lines = f.readlines()
+    for line in lines:
+        strs = line.split(':')
+        strs[1] = pyopenjtalk.g2p(strs[1])
+        strs[1] = strs[1].replace('pau',',')
+        
+        if k>95:
+            output.append('./eval_wav/fujitou_angry_'+format(k, '03')+'.wav|'+strs[1]+'|'+str(j)+'|3'+'\n')
+        k+=1    
+        
+        
+with open(out_path, 'w') as f:
+    f.writelines(output)

text_expressive.txt

@@ -0,0 +1,40 @@
+./eval_wav/whisper_VOICEACTRESS100_001.wav|m a t a , t o o j i n o y o o n i , g o d a i my o o o o t o y o b a r e r u , sh u y o o n a my o o o o n o ch u u o o n i h a i s a r e r u k o t o m o o o i|0
+./eval_wav/whisper_VOICEACTRESS100_002.wav|ny u u i N g u r a N d o f u u w a , gy u u ny u u o b e e s U t o sh I t a , sh i r o i k u r i i m u s u u p u d e a r i , b o s U t o N k u r a m u ch a u d a a t o m o y o b a r e r u|0
+./eval_wav/whisper_VOICEACTRESS100_003.wav|k o N py u u t a g e e m u n o m e e k a a y a , gy o o k a i d a N t a i n a d o n i k a N r e N s u r u j i N b u ts u n o k a t e g o r i|0
+./eval_wav/whisper_VOICEACTRESS100_004.wav|s a a b i s u m a n e e j a a d o o ny u u e k i n o t a m e , o o i m a ch i e k I k a r a , e N k a k U k a N r i sh I t e i r u|0
+./eval_wav/whisper_VOICEACTRESS100_005.wav|sh i r u b a a s a a f a a sh u u g e k i j i k e N m a d e n i , r i ch a a z u w a , ch i i m u m e e t o t o m o n i , k o k U s a i t e k i n i s u u p a a h i i r o o , o y o b i , y u u m e e j i N t o sh I t e , n i N ch I s a r e t e i r u|0
+./eval_wav/whisper_TRAVEL1000_0772.wav|d o ch i r a k a t e g a a i t e o r a r e m a s U k a|0
+./eval_wav/whisper_BASIC5000_0114.wav|s a N n e N t o i u w a , n a g a i j i k a N d a t o w a t a sh i w a o m o u|0
+./eval_wav/whisper_BASIC5000_2638.wav|t o sh i o t o r i , sh i N t a i m o f u j i y u u d e a cl t a g a , k a r e n i w a s o n o sh i g o t o o s u r u k i ry o k u g a a cl t a|0
+./eval_wav/whisper_BASIC5000_0104.wav|z o o k a s u r u j u y o o n i o o j i r u t a m e , a m e r i k a k a r a , gy u u n i k u o ts u i k a s u r u k e e k a k u d a|0
+./eval_wav/whisper_BASIC5000_0338.wav|t o r a w a , sh o o m e ts u sh I k a k a cl t e i r u sh u z o k u d e a r u|0
+./eval_wav/tsuchiya_happy_096.wav|p e N sh i r u b e n i a sh u u , p i cl ts u b a a g u n o , a r e g e e n i i k o o k o o o s o ts u gy o o sh i , k a r i f o r u n i a d a i g a k u , b a a k u r e e k o o n i ny u u g a k u|51|2
+./eval_wav/tsuchiya_happy_097.wav|k o n o g a i n e N n o d o o ny u u n i y o cl t e , s a m a z a m a n a b a r i e e sh o N n o , ry u u t a i n o k o N py u u t a a sh i my u r e e sh o N g a , t a k a i s e e d o d e k a n o o t o n a cl t a|51|2
+./eval_wav/tsuchiya_happy_098.wav|i n u i d o cl k u n i ny u u ky o sh I t e , o o b a a h o o r u s u b e k I k a d o o k a , p a f o o m a N s u g a ch e cl k U s a r e t a|51|2
+./eval_wav/tsuchiya_happy_099.wav|d e by u u w e e t o w a , s u u p a a b a N t a m u ky u u d e w a n a k u , f e z a a ky u u d a cl t a|51|2
+./eval_wav/tsuchiya_happy_100.wav|a a r i i w a , k o o u n o n a k a o , m i n a m i n o b a a j i n i a sh u u , w i N ch e s U t a a ch i k a k u n o , f i cl sh a a z u h i r u m a d e , g u N o sh i r i z o i t a|51|2
+./eval_wav/tsuchiya_angry_096.wav|p e N sh i r u b e n i a sh u u , p i cl ts u b a a g u n o , a r e g e e n i i k o o k o o o s o ts u gy o o sh i , k a r i f o r u n i a d a i g a k u , b a a k u r e e k o o n i ny u u g a k u|51|3
+./eval_wav/tsuchiya_angry_097.wav|k o n o g a i n e N n o d o o ny u u n i y o cl t e , s a m a z a m a n a b a r i e e sh o N n o , ry u u t a i n o k o N py u u t a a sh i my u r e e sh o N g a , t a k a i s e e d o d e k a n o o t o n a cl t a|51|3
+./eval_wav/tsuchiya_angry_098.wav|i n u i d o cl k u n i ny u u ky o sh I t e , o o b a a h o o r u s u b e k I k a d o o k a , p a f o o m a N s u g a ch e cl k U s a r e t a|51|3
+./eval_wav/tsuchiya_angry_099.wav|d e by u u w e e t o w a , s u u p a a b a N t a m u ky u u d e w a n a k u , f e z a a ky u u d a cl t a|51|3
+./eval_wav/tsuchiya_angry_100.wav|a a r i i w a , k o o u n o n a k a o , m i n a m i n o b a a j i n i a sh u u , w i N ch e s U t a a ch i k a k u n o , f i cl sh a a z u h i r u m a d e , g u N o sh i r i z o i t a|51|3
+./eval_wav/uemura_happy_096.wav|p e N sh i r u b e n i a sh u u , p i cl ts u b a a g u n o , a r e g e e n i i k o o k o o o s o ts u gy o o sh i , k a r i f o r u n i a d a i g a k u , b a a k u r e e k o o n i ny u u g a k u|52|2
+./eval_wav/uemura_happy_097.wav|k o n o g a i n e N n o d o o ny u u n i y o cl t e , s a m a z a m a n a b a r i e e sh o N n o , ry u u t a i n o k o N py u u t a a sh i my u r e e sh o N g a , t a k a i s e e d o d e k a n o o t o n a cl t a|52|2
+./eval_wav/uemura_happy_098.wav|i n u i d o cl k u n i ny u u ky o sh I t e , o o b a a h o o r u s u b e k I k a d o o k a , p a f o o m a N s u g a ch e cl k U s a r e t a|52|2
+./eval_wav/uemura_happy_099.wav|d e by u u w e e t o w a , s u u p a a b a N t a m u ky u u d e w a n a k u , f e z a a ky u u d a cl t a|52|2
+./eval_wav/uemura_happy_100.wav|a a r i i w a , k o o u n o n a k a o , m i n a m i n o b a a j i n i a sh u u , w i N ch e s U t a a ch i k a k u n o , f i cl sh a a z u h i r u m a d e , g u N o sh i r i z o i t a|52|2
+./eval_wav/uemura_angry_096.wav|p e N sh i r u b e n i a sh u u , p i cl ts u b a a g u n o , a r e g e e n i i k o o k o o o s o ts u gy o o sh i , k a r i f o r u n i a d a i g a k u , b a a k u r e e k o o n i ny u u g a k u|52|3
+./eval_wav/uemura_angry_097.wav|k o n o g a i n e N n o d o o ny u u n i y o cl t e , s a m a z a m a n a b a r i e e sh o N n o , ry u u t a i n o k o N py u u t a a sh i my u r e e sh o N g a , t a k a i s e e d o d e k a n o o t o n a cl t a|52|3
+./eval_wav/uemura_angry_098.wav|i n u i d o cl k u n i ny u u ky o sh I t e , o o b a a h o o r u s u b e k I k a d o o k a , p a f o o m a N s u g a ch e cl k U s a r e t a|52|3
+./eval_wav/uemura_angry_099.wav|d e by u u w e e t o w a , s u u p a a b a N t a m u ky u u d e w a n a k u , f e z a a ky u u d a cl t a|52|3
+./eval_wav/uemura_angry_100.wav|a a r i i w a , k o o u n o n a k a o , m i n a m i n o b a a j i n i a sh u u , w i N ch e s U t a a ch i k a k u n o , f i cl sh a a z u h i r u m a d e , g u N o sh i r i z o i t a|52|3
+./eval_wav/fujitou_happy_096.wav|p e N sh i r u b e n i a sh u u , p i cl ts u b a a g u n o , a r e g e e n i i k o o k o o o s o ts u gy o o sh i , k a r i f o r u n i a d a i g a k u , b a a k u r e e k o o n i ny u u g a k u|53|2
+./eval_wav/fujitou_happy_097.wav|k o n o g a i n e N n o d o o ny u u n i y o cl t e , s a m a z a m a n a b a r i e e sh o N n o , ry u u t a i n o k o N py u u t a a sh i my u r e e sh o N g a , t a k a i s e e d o d e k a n o o t o n a cl t a|53|2
+./eval_wav/fujitou_happy_098.wav|i n u i d o cl k u n i ny u u ky o sh I t e , o o b a a h o o r u s u b e k I k a d o o k a , p a f o o m a N s u g a ch e cl k U s a r e t a|53|2
+./eval_wav/fujitou_happy_099.wav|d e by u u w e e t o w a , s u u p a a b a N t a m u ky u u d e w a n a k u , f e z a a ky u u d a cl t a|53|2
+./eval_wav/fujitou_happy_100.wav|a a r i i w a , k o o u n o n a k a o , m i n a m i n o b a a j i n i a sh u u , w i N ch e s U t a a ch i k a k u n o , f i cl sh a a z u h i r u m a d e , g u N o sh i r i z o i t a|53|2
+./eval_wav/fujitou_angry_096.wav|p e N sh i r u b e n i a sh u u , p i cl ts u b a a g u n o , a r e g e e n i i k o o k o o o s o ts u gy o o sh i , k a r i f o r u n i a d a i g a k u , b a a k u r e e k o o n i ny u u g a k u|53|3
+./eval_wav/fujitou_angry_097.wav|k o n o g a i n e N n o d o o ny u u n i y o cl t e , s a m a z a m a n a b a r i e e sh o N n o , ry u u t a i n o k o N py u u t a a sh i my u r e e sh o N g a , t a k a i s e e d o d e k a n o o t o n a cl t a|53|3
+./eval_wav/fujitou_angry_098.wav|i n u i d o cl k u n i ny u u ky o sh I t e , o o b a a h o o r u s u b e k I k a d o o k a , p a f o o m a N s u g a ch e cl k U s a r e t a|53|3
+./eval_wav/fujitou_angry_099.wav|d e by u u w e e t o w a , s u u p a a b a N t a m u ky u u d e w a n a k u , f e z a a ky u u d a cl t a|53|3
+./eval_wav/fujitou_angry_100.wav|a a r i i w a , k o o u n o n a k a o , m i n a m i n o b a a j i n i a sh u u , w i N ch e s U t a a ch i k a k u n o , f i cl sh a a z u h i r u m a d e , g u N o sh i r i z o i t a|53|3

text_normal.py

@@ -0,0 +1,147 @@
+import os
+import pyopenjtalk
+import re
+
+out_path = './text_normal.txt'
+output = []
+j=0
+in_path = '/mnt/vol21/i21_mkazuki/text/transcripts_utf8.txt'
+for i in range(1,101):
+    if i in [1,3,5,6,9,11,12,13,20,21,22,23,28,31,32,33,34,37,41,42,44,45,46,47,48,49,50,52,54,68,70,71,73,74,75,76,77,78,79,80,81,86,87,88,89,97,98,99,100]:
+        continue
+    with open(in_path) as f:
+        lines = f.readlines()
+        for line in lines:
+            strs = line.split(':')
+            if not (os.path.isfile('/mnt/nfs-mnj-archive-03/group/creative/i21_mkazuki/wav/jvs'+format(i, '03')+'/parallel100/'+strs[0]+'.wav')):
+                continue
+            strs[1] = pyopenjtalk.g2p(strs[1])
+            strs[1] = strs[1].replace('pau',',')
+            
+    nonpara_path = '/mnt/nfs-mnj-archive-03/group/creative/i21_mkazuki/wav/jvs_ver1/jvs'+format(i, '03')+'/nonpara30/transcripts_utf8.txt'
+    with open(nonpara_path) as f:
+        lines = f.readlines()
+        for line in lines:
+            strs = line.split(':')
+            if not (os.path.isfile('/mnt/nfs-mnj-archive-03/group/creative/i21_mkazuki/wav/jvs'+format(i, '03')+'/nonpara30/'+strs[0]+'.wav')):
+                continue
+            strs[1] = pyopenjtalk.g2p(strs[1])
+            strs[1] = strs[1].replace('pau',',')
+            
+            if j==0:
+                output.append('./eval_wav/'+strs[0]+'.wav|'+strs[1]+'|'+str(j)+'\n')
+            
+    whisper_path = '/mnt/nfs-mnj-archive-03/group/creative/i21_mkazuki/wav/jvs_ver1/jvs'+format(i, '03')+'/whisper10/transcripts_utf8.txt'
+    with open(whisper_path) as f:
+        lines = f.readlines()
+        for line in lines:
+            strs = line.split(':')
+            if not (os.path.isfile('/mnt/nfs-mnj-archive-03/group/creative/i21_mkazuki/wav/jvs'+format(i, '03')+'/whisper10/'+strs[0]+'.wav')):
+                continue
+            strs[1] = pyopenjtalk.g2p(strs[1])
+            strs[1] = strs[1].replace('pau',',')
+            
+
+    j+=1
+    
+k=1
+
+with open(in_path) as f:
+    lines = f.readlines()
+    for line in lines:
+        strs = line.split(':')
+        strs[1] = pyopenjtalk.g2p(strs[1])
+        strs[1] = strs[1].replace('pau',',')
+        
+        if k>95:
+            output.append('./eval_wav/tsuchiya_normal_'+format(k, '03')+'.wav|'+strs[1]+'|'+str(j)+'\n')
+        k+=1
+        
+k=1        
+with open(in_path) as f:
+    lines = f.readlines()
+    for line in lines:
+        strs = line.split(':')
+        strs[1] = pyopenjtalk.g2p(strs[1])
+        strs[1] = strs[1].replace('pau',',')
+        
+        k+=1
+        
+k=1        
+with open(in_path) as f:
+    lines = f.readlines()
+    for line in lines:
+        strs = line.split(':')
+        strs[1] = pyopenjtalk.g2p(strs[1])
+        strs[1] = strs[1].replace('pau',',')
+        
+        k+=1    
+j+=1
+k=1
+with open(in_path) as f:
+    lines = f.readlines()
+    for line in lines:
+        strs = line.split(':')
+        strs[1] = pyopenjtalk.g2p(strs[1])
+        strs[1] = strs[1].replace('pau',',')
+        
+        if k>95:
+            output.append('./eval_wav/uemura_normal_'+format(k, '03')+'.wav|'+strs[1]+'|'+str(j)+'\n')
+        k+=1
+        
+k=1        
+with open(in_path) as f:
+    lines = f.readlines()
+    for line in lines:
+        strs = line.split(':')
+        strs[1] = pyopenjtalk.g2p(strs[1])
+        strs[1] = strs[1].replace('pau',',')
+        
+        k+=1
+        
+k=1        
+with open(in_path) as f:
+    lines = f.readlines()
+    for line in lines:
+        strs = line.split(':')
+        strs[1] = pyopenjtalk.g2p(strs[1])
+        strs[1] = strs[1].replace('pau',',')
+        k+=1    
+
+j+=1
+k=1
+with open(in_path) as f:
+    lines = f.readlines()
+    for line in lines:
+        strs = line.split(':')
+        strs[1] = pyopenjtalk.g2p(strs[1])
+        strs[1] = strs[1].replace('pau',',')
+        
+        if k>95:
+            output.append('./eval_wav/fujitou_normal_'+format(k, '03')+'.wav|'+strs[1]+'|'+str(j)+'\n')
+        k+=1
+        
+k=1        
+with open(in_path) as f:
+    lines = f.readlines()
+    for line in lines:
+        strs = line.split(':')
+        strs[1] = pyopenjtalk.g2p(strs[1])
+        strs[1] = strs[1].replace('pau',',')
+        
+
+        k+=1
+        
+k=1        
+with open(in_path) as f:
+    lines = f.readlines()
+    for line in lines:
+        strs = line.split(':')
+        strs[1] = pyopenjtalk.g2p(strs[1])
+        strs[1] = strs[1].replace('pau',',')
+        
+        k+=1    
+        
+        
+with open(out_path, 'w') as f:
+    f.writelines(output)

text_normal.txt

@@ -0,0 +1,45 @@
+./eval_wav/TRAVEL1000_0772.wav|d o ch i r a k a t e g a a i t e o r a r e m a s U k a|0
+./eval_wav/BASIC5000_0114.wav|s a N n e N t o i u w a , n a g a i j i k a N d a t o w a t a sh i w a o m o u|0
+./eval_wav/BASIC5000_2638.wav|t o sh i o t o r i , sh i N t a i m o f u j i y u u d e a cl t a g a , k a r e n i w a s o n o sh i g o t o o s u r u k i ry o k u g a a cl t a|0
+./eval_wav/BASIC5000_0104.wav|z o o k a s u r u j u y o o n i o o j i r u t a m e , a m e r i k a k a r a , gy u u n i k u o ts u i k a s u r u k e e k a k u d a|0
+./eval_wav/BASIC5000_0338.wav|t o r a w a , sh o o m e ts u sh I k a k a cl t e i r u sh u z o k u d e a r u|0
+./eval_wav/ONOMATOPEE300_036.wav|h a h a g a d a i d o k o r o d e , y u u h a N n o k a r e e o g u ts u g u ts U t o n i k o N d e i r u|0
+./eval_wav/BASIC5000_0533.wav|sh i m i o k o s u r i o t o s U k o t o g a d e k i n a k a cl t a|0
+./eval_wav/BASIC5000_1260.wav|d a k e d o , o r e n o k o k o r o w a m a d a k a o s u n o n a k a n i i r u|0
+./eval_wav/BASIC5000_1643.wav|sh u u h a s u u n o h e e k i N ch i w a , f u r i h a b a n o z o o d a i n i sh I t a g a cl t e g e N j i r u|0
+./eval_wav/LOANWORD128_007.wav|y u u j i N w a o N n a n o k o g a n i g a t e d e , o N n a n o k o t o h a n a sh i o s u r u t o ky o d o cl t e sh i m a u|0
+./eval_wav/BASIC5000_0844.wav|y o n o n a k a n i n a cl t o k u d e k i r u m o n o n a N t e n a i y o|0
+./eval_wav/TRAVEL1000_0481.wav|u e sh i ch i N w a k o k o k a r a n i b u r o cl k U h a n a r e t a t o k o r o d e s U|0
+./eval_wav/BASIC5000_0724.wav|o m i y a g e n i j u u b u N n a m o n o o k a u d a k e n o , j u u b u N n a j i k a N g a a r i m a s U k a|0
+./eval_wav/BASIC5000_2523.wav|o i sh a s a N g a , n e ts u g a s a g a r u m a d e w a a N s e e n i s u r u y o o n i cl t e , i cl t e t a d e sh o|0
+./eval_wav/BASIC5000_2103.wav|k a N b u N n o sh I k e N d e m a N t e N o t o r u s e e t o w a , h o t o N d o i m a s e N|0
+./eval_wav/BASIC5000_1883.wav|h i d a r i t e n o j i N b u ts u g a , s o n o e n o t o o i ts U s e e o k o w a sh I t e i r u|0
+./eval_wav/BASIC5000_1145.wav|j i b u N n o sh i g o t o w a a m a r i s U k i d e w a n a k a cl t a g a , a t a r a sh i k U k a cl t a i e d e , o k a n e o k a r i t e i t a|0
+./eval_wav/UT-PARAPHRASE-sent080-phrase1.wav|t o r i k a k o m u o o z e e n o h o o d o o j i N n i , f u k i N o t o o r i k a k a r u h I t o w a o d o r o i t e i t a|0
+./eval_wav/BASIC5000_1146.wav|h i n o u ch i d o k o r o g a n a i k e e k a k u w a , ky u u k U ts u d a n a a|0
+./eval_wav/BASIC5000_0963.wav|n a m a k e m o n o n o s o n o o t o k o w a , sh i b a sh i b a sh o k u m u o o k o t a r u|0
+./eval_wav/UT-PARAPHRASE-sent026-phrase1.wav|d o o sh o d e w a , g a i b u k a r a n o sh i N ny u u n o k e e s e k i g a n a i k o t o n a d o k a r a , n a i b u n o m o n o n i y o r u h a N k o o t o d a N t e e|0
+./eval_wav/BASIC5000_1445.wav|ch i ch I k a r a w a t a sh i n i , k o N py u u t a a g e e m u g a a t a e r a r e t a|0
+./eval_wav/UT-PARAPHRASE-sent212-phrase1.wav|p u r e cl sh a a m o t a e g a t a i h o d o k I ts u i t o i u|0
+./eval_wav/BASIC5000_1915.wav|ry o o a sh i o sh i cl k a r i t o j i m e N n i ts U k e t e i n a s a i|0
+./eval_wav/TRAVEL1000_0639.wav|a t a r a sh i i j u u gy o o i N n o i ch i b u w a , t e k I s e ts u n a t o r e e n i N g u o u k e t e i n a i k a m o sh i r e m a s e N|0
+./eval_wav/TRAVEL1000_0292.wav|t o r a b e r a a z u ch e cl k u d e o sh i h a r a i d e k i m a s U k a|0
+./eval_wav/BASIC5000_2396.wav|d o o m i t e m o k a r e r a n o k o o d o o w a m i o m u s u N d e i n a i|0
+./eval_wav/TRAVEL1000_0312.wav|sh I k a sh i , m o sh i m i cl k a m a e n i k o o ny u u s u r e b a , hy a k u h a ch i d o r u , n a n a j u u g o s e N t o d e s U|0
+./eval_wav/BASIC5000_1516.wav|g a N t a N n i w a , k i N j o n o j i N j a n i o m a i r i s u r u h I t o g a o o i|0
+./eval_wav/TRAVEL1000_0492.wav|w a t a sh i n o s a i z u w a y o k u w a k a r i m a s e N|0
+./eval_wav/tsuchiya_normal_096.wav|p e N sh i r u b e n i a sh u u , p i cl ts u b a a g u n o , a r e g e e n i i k o o k o o o s o ts u gy o o sh i , k a r i f o r u n i a d a i g a k u , b a a k u r e e k o o n i ny u u g a k u|51
+./eval_wav/tsuchiya_normal_097.wav|k o n o g a i n e N n o d o o ny u u n i y o cl t e , s a m a z a m a n a b a r i e e sh o N n o , ry u u t a i n o k o N py u u t a a sh i my u r e e sh o N g a , t a k a i s e e d o d e k a n o o t o n a cl t a|51
+./eval_wav/tsuchiya_normal_098.wav|i n u i d o cl k u n i ny u u ky o sh I t e , o o b a a h o o r u s u b e k I k a d o o k a , p a f o o m a N s u g a ch e cl k U s a r e t a|51
+./eval_wav/tsuchiya_normal_099.wav|d e by u u w e e t o w a , s u u p a a b a N t a m u ky u u d e w a n a k u , f e z a a ky u u d a cl t a|51
+./eval_wav/tsuchiya_normal_100.wav|a a r i i w a , k o o u n o n a k a o , m i n a m i n o b a a j i n i a sh u u , w i N ch e s U t a a ch i k a k u n o , f i cl sh a a z u h i r u m a d e , g u N o sh i r i z o i t a|51
+./eval_wav/uemura_normal_096.wav|p e N sh i r u b e n i a sh u u , p i cl ts u b a a g u n o , a r e g e e n i i k o o k o o o s o ts u gy o o sh i , k a r i f o r u n i a d a i g a k u , b a a k u r e e k o o n i ny u u g a k u|52
+./eval_wav/uemura_normal_097.wav|k o n o g a i n e N n o d o o ny u u n i y o cl t e , s a m a z a m a n a b a r i e e sh o N n o , ry u u t a i n o k o N py u u t a a sh i my u r e e sh o N g a , t a k a i s e e d o d e k a n o o t o n a cl t a|52
+./eval_wav/uemura_normal_098.wav|i n u i d o cl k u n i ny u u ky o sh I t e , o o b a a h o o r u s u b e k I k a d o o k a , p a f o o m a N s u g a ch e cl k U s a r e t a|52
+./eval_wav/uemura_normal_099.wav|d e by u u w e e t o w a , s u u p a a b a N t a m u ky u u d e w a n a k u , f e z a a ky u u d a cl t a|52
+./eval_wav/uemura_normal_100.wav|a a r i i w a , k o o u n o n a k a o , m i n a m i n o b a a j i n i a sh u u , w i N ch e s U t a a ch i k a k u n o , f i cl sh a a z u h i r u m a d e , g u N o sh i r i z o i t a|52
+./eval_wav/fujitou_normal_096.wav|p e N sh i r u b e n i a sh u u , p i cl ts u b a a g u n o , a r e g e e n i i k o o k o o o s o ts u gy o o sh i , k a r i f o r u n i a d a i g a k u , b a a k u r e e k o o n i ny u u g a k u|53
+./eval_wav/fujitou_normal_097.wav|k o n o g a i n e N n o d o o ny u u n i y o cl t e , s a m a z a m a n a b a r i e e sh o N n o , ry u u t a i n o k o N py u u t a a sh i my u r e e sh o N g a , t a k a i s e e d o d e k a n o o t o n a cl t a|53
+./eval_wav/fujitou_normal_098.wav|i n u i d o cl k u n i ny u u ky o sh I t e , o o b a a h o o r u s u b e k I k a d o o k a , p a f o o m a N s u g a ch e cl k U s a r e t a|53
+./eval_wav/fujitou_normal_099.wav|d e by u u w e e t o w a , s u u p a a b a N t a m u ky u u d e w a n a k u , f e z a a ky u u d a cl t a|53
+./eval_wav/fujitou_normal_100.wav|a a r i i w a , k o o u n o n a k a o , m i n a m i n o b a a j i n i a sh u u , w i N ch e s U t a a ch i k a k u n o , f i cl sh a a z u h i r u m a d e , g u N o sh i r i z o i t a|53