updating loading script

IMDA - National Speech Corpus/PART3/tmp_clip.wav

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:bee5844aab3b7fc7f612fd752dc4f7ca8ce49b30a5cc8c3b8f116beb9f45c761
-size 787404
+oid sha256:57cbf6bf547727f71d6d0208756677ceea272749468e7877db11f8a918b053a6
+size 530898

imda-dataset-p1.py

@@ -5,7 +5,9 @@ import pandas as pd
 from sklearn.model_selection import train_test_split
 
 _DESCRIPTION = """\
-This new dataset is designed to solve this great NLP task and is crafted with a lot of care.
+Part 1 of the National Speech Corpus. The National Speech Corpus (NSC)
+is the first large-scale Singapore English corpus spearheaded by the 
+Info-communications and Media Development Authority (IMDA) of Singapore.
 """
 
 _CITATION = """\

imda-dataset.py

@@ -1,54 +1,97 @@
 import os
-import glob
 import datasets
-import pandas as pd
 from sklearn.model_selection import train_test_split
+import textgrids
+import soundfile as sf
+import re
+import json
+import tempfile
+import random
+
+def cleanup_string(line):
+
+    words_to_remove = ['(ppo)','(ppc)', '(ppb)', '(ppl)', '<s/>','<c/>','<q/>', '<fil/>', '<sta/>', '<nps/>', '<spk/>', '<non/>', '<unk>', '<s>', '<z>', '<nen>']
+
+    formatted_line = re.sub(r'\s+', ' ', line).strip().lower()
+
+    #detect all word that matches words in the words_to_remove list
+    for word in words_to_remove:
+        if re.search(word,formatted_line):
+            # formatted_line = re.sub(word,'', formatted_line)
+            formatted_line = formatted_line.replace(word,'')
+            formatted_line = re.sub(r'\s+', ' ', formatted_line).strip().lower()
+            # print("*** removed words: " + formatted_line)
+
+    #detect '\[(.*?)\].' e.g. 'Okay [ah], why did I gamble?'
+    #remove [ ] and keep text within
+    if re.search('\[(.*?)\]', formatted_line):
+        formatted_line = re.sub('\[(.*?)\]', r'\1', formatted_line).strip()
+        #print("***: " + formatted_line)
+
+    #detect '\((.*?)\).' e.g. 'Okay (um), why did I gamble?'
+    #remove ( ) and keep text within
+    if re.search('\((.*?)\)', formatted_line):
+        formatted_line = re.sub('\((.*?)\)', r'\1', formatted_line).strip()
+        # print("***: " + formatted_line)
+
+    #detect '\'(.*?)\'' e.g. 'not 'hot' per se'
+    #remove ' ' and keep text within
+    if re.search('\'(.*?)\'', formatted_line):
+        formatted_line = re.sub('\'(.*?)\'', r'\1', formatted_line).strip()
+        #print("***: " + formatted_line)
+
+    #remove punctation '''!()-[]{};:'"\, <>./?@#$%^&*_~'''
+    punctuation = '''!–;"\,./?@#$%^&*~'''
+    punctuation_list = str.maketrans("","",punctuation)
+    formatted_line = re.sub(r'-', ' ', formatted_line)
+    formatted_line = re.sub(r'_', ' ', formatted_line)
+    formatted_line = formatted_line.translate(punctuation_list)
+    formatted_line = re.sub(r'\s+', ' ', formatted_line).strip().lower()
+    #print("***: " + formatted_line)
+
+    return formatted_line
+
+
 
 _DESCRIPTION = """\
-This new dataset is designed to solve this great NLP task and is crafted with a lot of care.
+The National Speech Corpus (NSC) is the first large-scale Singapore English corpus
+spearheaded by the Info-communications and Media Development Authority (IMDA) of Singapore.
 """
 
 _CITATION = """\
 """
 _CHANNEL_CONFIGS = sorted([
-    "CHANNEL0", "CHANNEL1", "CHANNEL2"
+    "Audio Same CloseMic", "Audio Separate StandingMic"
 ])
 
-_GENDER_CONFIGS = sorted(["F", "M"])
-
-_RACE_CONFIGS = sorted(["CHINESE", "MALAY", "INDIAN", "OTHERS"])
+_HOMEPAGE = "https://www.imda.gov.sg/how-we-can-help/national-speech-corpus"
 
-_HOMEPAGE = "https://huggingface.co/indonesian-nlp/librivox-indonesia"
+_LICENSE = ""
 
-_LICENSE = "https://creativecommons.org/publicdomain/zero/1.0/"
+_PATH_TO_DATA = './IMDA - National Speech Corpus/PART3'
 
-_PATH_TO_DATA = './IMDA - National Speech Corpus/PART2'
-# _PATH_TO_DATA = './PART2/DATA'
+INTERVAL_MAX_LENGTH = 25
 
 class Minds14Config(datasets.BuilderConfig):
     """BuilderConfig for xtreme-s"""
 
     def __init__(
-        self, channel, gender, race, description, homepage, path_to_data
+        self, channel, description, homepage, path_to_data
     ):
         super(Minds14Config, self).__init__(
-            name=channel+gender+race,
+            name=channel,
             version=datasets.Version("1.0.0", ""),
             description=self.description,
         )
         self.channel = channel
-        self.gender = gender
-        self.race = race
         self.description = description
         self.homepage = homepage
         self.path_to_data = path_to_data
 
 
-def _build_config(channel, gender, race):
+def _build_config(channel):
     return Minds14Config(
         channel=channel,
-        gender=gender,
-        race=race,
         description=_DESCRIPTION,
         homepage=_HOMEPAGE,
         path_to_data=_PATH_TO_DATA,
@@ -73,28 +116,24 @@ class NewDataset(datasets.GeneratorBasedBuilder):
     # data = datasets.load_dataset('my_dataset', 'second_domain')
     BUILDER_CONFIGS = []
     for channel in _CHANNEL_CONFIGS + ["all"]:
-        for gender in _GENDER_CONFIGS + ["all"]:
-            for race in _RACE_CONFIGS + ["all"]:
-                BUILDER_CONFIGS.append(_build_config(channel, gender, race))
+        BUILDER_CONFIGS.append(_build_config(channel))
     # BUILDER_CONFIGS = [_build_config(name) for name in _CHANNEL_CONFIGS + ["all"]]
 
-    DEFAULT_CONFIG_NAME = "allallall"  # It's not mandatory to have a default configuration. Just use one if it make sense.
+    DEFAULT_CONFIG_NAME = "all"  # It's not mandatory to have a default configuration. Just use one if it make sense.
 
     def _info(self):
         # TODO: This method specifies the datasets.DatasetInfo object which contains informations and typings for the dataset
         task_templates = None
-        # mics = _CHANNEL_CONFIGS
         features = datasets.Features(
             {
-                "audio": datasets.features.Audio(sampling_rate=16000),
+                "audio": datasets.features.Audio(),
                 "transcript": datasets.Value("string"),
                 "mic": datasets.Value("string"),
                 "audio_name": datasets.Value("string"),
-                "gender": datasets.Value("string"),
-                "race": datasets.Value("string"),
+                "interval": datasets.Value("string")
             }
         )
-        
+
         return datasets.DatasetInfo(
             # This is the description that will appear on the datasets page.
             description=_DESCRIPTION,
@@ -121,31 +160,55 @@ class NewDataset(datasets.GeneratorBasedBuilder):
             else [self.config.channel]
         )
 
-        gender = (
-            _GENDER_CONFIGS
-            if self.config.gender == "all"
-            else [self.config.gender]
-        )
+        json_path = dl_manager.download(os.path.join(self.config.path_to_data, "directory_list.json"))
+        # print(f"json_path: {json_path}")
+        with open(json_path, "r") as f:
+            directory_dict = json.load(f)
+        # print(f"directory_dict: {directory_dict}")
 
-        race = (
-            _RACE_CONFIGS
-            if self.config.race == "all"
-            else [self.config.race]
-        )
+        train_audio_list = []
+        test_audio_list = []
+        for mic in mics:
+            audio_list = []
+            if mic == "Audio Same CloseMic":
+                audio_list = [x for x in directory_dict[mic] if (x[-5] == "1") ]
+                # train test split speaker 1, append speaker 2 depending on in train or test dataset
+                train, test = train_test_split(audio_list, test_size=0.005, random_state=42, shuffle=True)
+                for path in train:
+                    train_audio_list.append(os.path.join(self.config.path_to_data, mic, path))
+                    s = list(path)
+                    s[-5] = "2"
+                    train_audio_list.append(os.path.join(self.config.path_to_data, mic, "".join(s)))
+                for path in test:
+                    test_audio_list.append(os.path.join(self.config.path_to_data, mic, path))
+                    s = list(path)
+                    s[-5] = "2"
+                    test_audio_list.append(os.path.join(self.config.path_to_data, mic, "".join(s)))
+            elif mic == "Audio Separate IVR":
+                audio_list = [x.split("\\")[0] for x in directory_dict[mic]]
+                print('AUDIO LIST',audio_list)
+                train, test = train_test_split(audio_list, test_size=0.005, random_state=42, shuffle=True)
+                for folder in train:
+                    audios = [os.path.join(self.config.path_to_data, mic, x) for x in directory_dict[mic] if (x.split("\\")[0]==folder)]
+                    train_audio_list.extend(audios)
+                for folder in test:
+                    audios = [os.path.join(self.config.path_to_data, mic, x) for x in directory_dict[mic] if (x.split("\\")[0]==folder)]
+                    test_audio_list.extend(audios)
+            elif mic == "Audio Separate StandingMic":
+                audio_list = [x[:14] for x in directory_dict[mic]]
+                audio_list = list(set(audio_list))
+                train, test = train_test_split(audio_list, test_size=0.005, random_state=42, shuffle=True)
+                for folder in train:
+                    audios = [os.path.join(self.config.path_to_data, mic, x) for x in directory_dict[mic] if (x[:14]==folder)]
+                    train_audio_list.extend(audios)
+                for folder in test:
+                    audios = [os.path.join(self.config.path_to_data, mic, x) for x in directory_dict[mic] if (x[:14]==folder)]
+                    test_audio_list.extend(audios)
 
-        # augment speaker ids directly here
-        # read the speaker information
-        train_speaker_ids = []
-        test_speaker_ids = []
-        # path_to_speaker = os.path.join(self.config.path_to_data, "DOC", "Speaker Information (Part 1).XLSX")
-        path_to_speaker = dl_manager.download(os.path.join(self.config.path_to_data, "DOC", "Speaker Information (Part 2).XLSX"))
-        speaker_df = pd.read_excel(path_to_speaker, dtype={'SCD/PART2': 'str'})
-        for g in gender:
-            for r in race:
-                X = speaker_df[(speaker_df["ACC"]==r) & (speaker_df["SEX"]==g)]
-                X_train, X_test = train_test_split(X, test_size=0.3, random_state=42, shuffle=True)
-                train_speaker_ids.extend(X_train["SCD/PART2"])
-                test_speaker_ids.extend(X_test["SCD/PART2"])
+        random.shuffle(train_audio_list)
+        random.shuffle(test_audio_list)
+        print(f"train_audio_list: { train_audio_list}")
+        print(f"test_audio_list: { test_audio_list}")
 
         # dl_manager is a datasets.download.DownloadManager that can be used to download and extract URLS
         # It can accept any type or nested list/dict and will give back the same structure with the url replaced with path to local files.
@@ -154,23 +217,15 @@ class NewDataset(datasets.GeneratorBasedBuilder):
             datasets.SplitGenerator(
             name=datasets.Split.TRAIN,
             gen_kwargs={
-                "path_to_data": self.config.path_to_data,
-                "speaker_metadata":speaker_df,
-                # "speaker_ids": train_speaker_ids,
-                "speaker_ids":["2001"],
-                "mics": mics,
-                "dl_manager": dl_manager
+                "audio_list": train_audio_list,
+                "dl_manager":dl_manager,
               },
           ),
           datasets.SplitGenerator(
             name=datasets.Split.TEST,
             gen_kwargs={
-                "path_to_data": self.config.path_to_data,
-                "speaker_metadata":speaker_df,
-                # "speaker_ids": test_speaker_ids,
-                "speaker_ids": ["2003"],
-                "mics": mics,
-                "dl_manager": dl_manager
+                "audio_list": test_audio_list,
+                "dl_manager":dl_manager,
             },
         ),
         ]
@@ -178,55 +233,135 @@ class NewDataset(datasets.GeneratorBasedBuilder):
     # method parameters are unpacked from `gen_kwargs` as given in `_split_generators`
     def _generate_examples(
             self,
-            path_to_data,
-            speaker_metadata,
-            speaker_ids,
-            mics,
-            dl_manager
+            audio_list,
+            dl_manager,
         ):
         id_ = 0
-        for mic in mics:
-            for speaker in speaker_ids:
-                # TRANSCRIPT: in the case of error, if no file found then dictionary will b empty
-                d = {}
-                counter = 0
-                while counter < 10:
-                    data = dl_manager.download(os.path.join(path_to_data, "DATA", mic, "SCRIPT", mic[-1]+speaker+str(counter)+'.TXT'))
-                    try:
-                        line_num = 0
-                        with open(data, encoding='utf-8-sig') as f:
-                            for line in f:
-                                if line_num == 0:
-                                    key = line.split("\t")[0]
-                                    line_num += 1
-                                elif line_num == 1:
-                                    d[key] = line.strip()
-                                    line_num -= 1
-                    except:
-                        print(f"{counter}")
-                        break
-                    counter+=1
-                # AUDIO: in the case of error it will skip the speaker
-                # archive_path = os.path.join(path_to_data, "DATA", mic, "WAVE", "SPEAKER"+speaker+'.zip')
-                archive_path = dl_manager.download(os.path.join(path_to_data, "DATA", mic, "WAVE", "SPEAKER"+speaker+'.zip'))
-                # check that archive path exists, else will not open the archive
-                if os.path.exists(archive_path):
-                    audio_files = dl_manager.iter_archive(archive_path)
-                    for path, f in audio_files:
-                        # bug catching if any error?
-                        result = {}
-                        full_path = os.path.join(archive_path, path) if archive_path else path # bug catching here
-                        result["audio"] = {"path": full_path, "bytes": f.read()}
-                        result["audio_name"] = path
-                        result["mic"] = mic
-                        metadata_row = speaker_metadata.loc[speaker_metadata["SCD/PART2"]==speaker].iloc[0]
-                        result["gender"]=metadata_row["SEX"]
-                        result["race"]=metadata_row["ACC"]
-                        try:
-                            result["transcript"] = d[f.name[-13:-4]]
+        for audio_path in audio_list:
+            try:
+                file = os.path.split(audio_path)[-1]
+                folder = os.path.split(os.path.split(audio_path)[0])[-1]
+                # get script_path
+                if folder.split("_")[0] == "conf":
+                    # mic == "Audio Separate IVR"
+                    file_name = folder+'_'+file
+                    script_path = os.path.join(self.config.path_to_data, "Scripts Separate", file_name[:-4]+".TextGrid")
+                elif folder.split()[1] == "Same":
+                    # mic == "Audio Same CloseMic IVR"
+                    script_path = os.path.join(self.config.path_to_data, "Scripts Same", file[:-4]+".TextGrid")
+                elif folder.split()[1] == "Separate":
+                    # mic == "Audio Separate StandingMic":
+                    script_path = os.path.join(self.config.path_to_data, "Scripts Separate", file[:-4]+".TextGrid")
+                script_path = dl_manager.download(script_path)
+            except Exception as e:
+                print(f"error getting script path, {str(e)}")
+                continue
+
+                # LOAD TRANSCRIPT
+            # check that the textgrid file can be read
+
+            try:
+                # tg = textgrid.TextGrid.fromFile(script_path)
+                with open(script_path, "rb") as f:
+                    tg = f.read()
+                    tg_dict = textgrids.TextGrid()
+                    tg_dict.parse(tg)
+                for key in tg_dict.keys():
+                    tg = tg_dict[key]
+            except UnicodeDecodeError:
+                try:
+                    with open(script_path, "rb") as f:
+                        tg = f.read()
+                        decoded = tg.decode('utf-16')
+                        encoded = decoded.encode('utf-8')
+                        tg_dict = textgrids.TextGrid()
+                        tg_dict.parse(encoded)
+                    for key in tg_dict.keys():
+                        tg = tg_dict[key]
+                except Exception as e:
+                    print(f"error reading textgrid file, {script_path}, {str(e)}")
+                    continue
+            except TypeError:
+                try:
+                    with open(script_path, "rb") as f:
+                        tg = f.read()
+                        decoded = tg.decode('utf-8-sig')
+                        encoded = decoded.encode('utf-8')
+                        tg_dict = textgrids.TextGrid()
+                        tg_dict.parse(encoded)
+                    for key in tg_dict.keys():
+                        tg = tg_dict[key]
+                except Exception as e:
+                    print(f"error reading textgrid file, {script_path}, {str(e)}")
+                    continue
+            except Exception as e:
+                print(f"error reading textgrid file, {script_path}, {str(e)}")
+                continue
+            # LOAD AUDIO
+            # check that archive path exists, else will not open the archive
+            audio_path = dl_manager.download(audio_path)
+            if os.path.exists(audio_path):
+                try:
+                    with open(audio_path, 'rb') as f:
+                        data, sr = sf.read(f)
+                        if sr != 16000:
+                            print(f'sample rate: {sr}')
+                            continue
+                    # data, sr = sf.read(audio_path)
+                    result = {}
+                    i = 0
+                    intervalLength = 0
+                    intervalStart = 0
+                    transcript_list = []
+                    # filepath = os.path.join(self.config.path_to_data, f'tmp_clip{id_}.wav')
+                    # filepath = dl_manager.download(filepath)
+                    tempWavFile = tempfile.mktemp('.wav')
+                    while i < (len(tg)-1):
+                        transcript = cleanup_string(tg[i].text)
+                        if intervalLength == 0 and len(transcript) == 0:
+                            intervalStart = tg[i+1].xmin
+                            i+=1
+                            continue
+                        intervalLength += tg[i].xmax-tg[i].xmin
+                        if (tg[i].xmax-tg[i].xmin) > INTERVAL_MAX_LENGTH:
+                            print(f"Interval is too long: {tg[i].xmax-tg[i].xmin}")
+                            intervalLength = 0
+                            intervalStart = tg[i+1].xmin
+                            transcript_list = []
+                            i+=1
+                            continue
+                            # spliced_audio = data[int(tg[i].xmin*sr):int(tg[i].xmax*sr)]
+                            # sf.write(tempWavFile, spliced_audio, sr)
+                            # result["transcript"] = transcript
+                            # result["interval"] = "start:"+str(tg[i].xmin)+", end:"+str(tg[i].xmax)
+                            # result["audio"] = {"path": tempWavFile, "bytes": spliced_audio, "sampling_rate":sr}
+                            # result["audio_name"] = audio_path
+                            # yield id_, result
+                            # id_+= 1
+                            # intervalLength = 0
+                        else:
+                            if (intervalLength + tg[i+1].xmax-tg[i+1].xmin) < INTERVAL_MAX_LENGTH:
+                                if len(transcript) != 0:
+                                    transcript_list.append(transcript)
+                                i+=1
+                                continue
+                            if len(transcript) == 0:
+                                spliced_audio = data[int(intervalStart*sr):int(tg[i].xmax*sr)]
+                            else:
+                                transcript_list.append(transcript)
+                                spliced_audio = data[int(intervalStart*sr):int(tg[i].xmax*sr)]
+
+                            sf.write(tempWavFile,spliced_audio, sr )
+                            # sf.write(filepath, spliced_audio, sr)
+                            result["interval"] = "start:"+str(intervalStart)+", end:"+str(tg[i].xmax)
+                            result["audio"] = {"path": tempWavFile, "bytes": spliced_audio, "sampling_rate":sr}
+                            result["transcript"] = ' '.join(transcript_list)
+                            result["audio_name"] = audio_path
                             yield id_, result
-                            id_ += 1
-                        except:
-                            print(f"unable to find transcript")
-                        
-                        
+                            id_+= 1
+                            intervalLength=0
+                            intervalStart=tg[i+1].xmin
+                            transcript_list = []
+                        i+=1
+                except:
+                    continue

imda_nsc_part2.py

@@ -0,0 +1,234 @@
+import os
+import glob
+import datasets
+import pandas as pd
+from sklearn.model_selection import train_test_split
+
+_DESCRIPTION = """\
+Part 2 of the National Speech Corpus. The National Speech Corpus (NSC)
+is the first large-scale Singapore English corpus spearheaded by the 
+Info-communications and Media Development Authority (IMDA) of Singapore.
+"""
+
+_CITATION = """\
+"""
+_CHANNEL_CONFIGS = sorted([
+    "CHANNEL0", "CHANNEL1", "CHANNEL2"
+])
+
+_GENDER_CONFIGS = sorted(["F", "M"])
+
+_RACE_CONFIGS = sorted(["CHINESE", "MALAY", "INDIAN", "OTHERS"])
+
+_HOMEPAGE = "https://www.imda.gov.sg/how-we-can-help/national-speech-corpus"
+
+_LICENSE = "https://creativecommons.org/publicdomain/zero/1.0/"
+
+# _PATH_TO_DATA = './IMDA - National Speech Corpus/PART2'
+_PATH_TO_DATA = './PART2'
+
+class Minds14Config(datasets.BuilderConfig):
+    """BuilderConfig for xtreme-s"""
+
+    def __init__(
+        self, channel, gender, race, description, homepage, path_to_data
+    ):
+        super(Minds14Config, self).__init__(
+            name=channel+gender+race,
+            version=datasets.Version("1.0.0", ""),
+            description=self.description,
+        )
+        self.channel = channel
+        self.gender = gender
+        self.race = race
+        self.description = description
+        self.homepage = homepage
+        self.path_to_data = path_to_data
+
+
+def _build_config(channel, gender, race):
+    return Minds14Config(
+        channel=channel,
+        gender=gender,
+        race=race,
+        description=_DESCRIPTION,
+        homepage=_HOMEPAGE,
+        path_to_data=_PATH_TO_DATA,
+    )
+
+# TODO: Name of the dataset usually matches the script name with CamelCase instead of snake_case
+class NewDataset(datasets.GeneratorBasedBuilder):
+    """TODO: Short description of my dataset."""
+
+    VERSION = datasets.Version("1.1.0")
+
+    # This is an example of a dataset with multiple configurations.
+    # If you don't want/need to define several sub-sets in your dataset,
+    # just remove the BUILDER_CONFIG_CLASS and the BUILDER_CONFIGS attributes.
+
+    # If you need to make complex sub-parts in the datasets with configurable options
+    # You can create your own builder configuration class to store attribute, inheriting from datasets.BuilderConfig
+    # BUILDER_CONFIG_CLASS = MyBuilderConfig
+
+    # You will be able to load one or the other configurations in the following list with
+    # data = datasets.load_dataset('my_dataset', 'first_domain')
+    # data = datasets.load_dataset('my_dataset', 'second_domain')
+    BUILDER_CONFIGS = []
+    for channel in _CHANNEL_CONFIGS + ["all"]:
+        for gender in _GENDER_CONFIGS + ["all"]:
+            for race in _RACE_CONFIGS + ["all"]:
+                BUILDER_CONFIGS.append(_build_config(channel, gender, race))
+    # BUILDER_CONFIGS = [_build_config(name) for name in _CHANNEL_CONFIGS + ["all"]]
+
+    DEFAULT_CONFIG_NAME = "allallall"  # It's not mandatory to have a default configuration. Just use one if it make sense.
+
+    def _info(self):
+        # TODO: This method specifies the datasets.DatasetInfo object which contains informations and typings for the dataset
+        task_templates = None
+        # mics = _CHANNEL_CONFIGS
+        features = datasets.Features(
+            {
+                "audio": datasets.features.Audio(sampling_rate=16000),
+                "transcript": datasets.Value("string"),
+                "mic": datasets.Value("string"),
+                "audio_name": datasets.Value("string"),
+                "gender": datasets.Value("string"),
+                "race": datasets.Value("string"),
+            }
+        )
+        
+        return datasets.DatasetInfo(
+            # This is the description that will appear on the datasets page.
+            description=_DESCRIPTION,
+            # This defines the different columns of the dataset and their types
+            features=features,  # Here we define them above because they are different between the two configurations
+            # If there's a common (input, target) tuple from the features, uncomment supervised_keys line below and
+            # specify them. They'll be used if as_supervised=True in builder.as_dataset.
+            supervised_keys=("audio", "transcript"),
+            # Homepage of the dataset for documentation
+            homepage=_HOMEPAGE,
+            # License for the dataset if available
+            license=_LICENSE,
+            # Citation for the dataset
+            citation=_CITATION,
+            task_templates=task_templates,
+        )
+
+    def _split_generators(self, dl_manager):
+        # TODO: This method is tasked with downloading/extracting the data and defining the splits depending on the configuration
+        # If several configurations are possible (listed in BUILDER_CONFIGS), the configuration selected by the user is in self.config.name
+        mics = (
+            _CHANNEL_CONFIGS
+            if self.config.channel == "all"
+            else [self.config.channel]
+        )
+
+        gender = (
+            _GENDER_CONFIGS
+            if self.config.gender == "all"
+            else [self.config.gender]
+        )
+
+        race = (
+            _RACE_CONFIGS
+            if self.config.race == "all"
+            else [self.config.race]
+        )
+
+        # augment speaker ids directly here
+        # read the speaker information
+        train_speaker_ids = []
+        test_speaker_ids = []
+        # path_to_speaker = os.path.join(self.config.path_to_data, "DOC", "Speaker Information (Part 1).XLSX")
+        path_to_speaker = dl_manager.download(os.path.join(self.config.path_to_data, "DOC", "Speaker Information (Part 2).XLSX"))
+        speaker_df = pd.read_excel(path_to_speaker, dtype={'SCD/PART2': 'str'})
+        for g in gender:
+            for r in race:
+                X = speaker_df[(speaker_df["ACC"]==r) & (speaker_df["SEX"]==g)]
+                X_train, X_test = train_test_split(X, test_size=0.3, random_state=42, shuffle=True)
+                train_speaker_ids.extend(X_train["SCD/PART2"])
+                test_speaker_ids.extend(X_test["SCD/PART2"])
+
+        # dl_manager is a datasets.download.DownloadManager that can be used to download and extract URLS
+        # It can accept any type or nested list/dict and will give back the same structure with the url replaced with path to local files.
+        # By default the archives will be extracted and a path to a cached folder where they are extracted is returned instead of the archive
+        return [
+            datasets.SplitGenerator(
+            name=datasets.Split.TRAIN,
+            gen_kwargs={
+                "path_to_data": self.config.path_to_data,
+                "speaker_metadata":speaker_df,
+                "speaker_ids": train_speaker_ids,
+                # "speaker_ids":["2001"],
+                "mics": mics,
+                "dl_manager": dl_manager
+              },
+          ),
+          datasets.SplitGenerator(
+            name=datasets.Split.TEST,
+            gen_kwargs={
+                "path_to_data": self.config.path_to_data,
+                "speaker_metadata":speaker_df,
+                "speaker_ids": test_speaker_ids,
+                # "speaker_ids": ["2003"],
+                "mics": mics,
+                "dl_manager": dl_manager
+            },
+        ),
+        ]
+
+    # method parameters are unpacked from `gen_kwargs` as given in `_split_generators`
+    def _generate_examples(
+            self,
+            path_to_data,
+            speaker_metadata,
+            speaker_ids,
+            mics,
+            dl_manager
+        ):
+        id_ = 0
+        for mic in mics:
+            for speaker in speaker_ids:
+                # TRANSCRIPT: in the case of error, if no file found then dictionary will b empty
+                d = {}
+                counter = 0
+                while counter < 10:
+                    data = dl_manager.download(os.path.join(path_to_data, "DATA", mic, "SCRIPT", mic[-1]+speaker+str(counter)+'.TXT'))
+                    try:
+                        line_num = 0
+                        with open(data, encoding='utf-8-sig') as f:
+                            for line in f:
+                                if line_num == 0:
+                                    key = line.split("\t")[0]
+                                    line_num += 1
+                                elif line_num == 1:
+                                    d[key] = line.strip()
+                                    line_num -= 1
+                    except:
+                        print(f"{counter}")
+                        break
+                    counter+=1
+                # AUDIO: in the case of error it will skip the speaker
+                # archive_path = os.path.join(path_to_data, "DATA", mic, "WAVE", "SPEAKER"+speaker+'.zip')
+                archive_path = dl_manager.download(os.path.join(path_to_data, "DATA", mic, "WAVE", "SPEAKER"+speaker+'.zip'))
+                # check that archive path exists, else will not open the archive
+                if os.path.exists(archive_path):
+                    audio_files = dl_manager.iter_archive(archive_path)
+                    for path, f in audio_files:
+                        # bug catching if any error?
+                        result = {}
+                        full_path = os.path.join(archive_path, path) if archive_path else path # bug catching here
+                        result["audio"] = {"path": full_path, "bytes": f.read()}
+                        result["audio_name"] = path
+                        result["mic"] = mic
+                        metadata_row = speaker_metadata.loc[speaker_metadata["SCD/PART2"]==speaker].iloc[0]
+                        result["gender"]=metadata_row["SEX"]
+                        result["race"]=metadata_row["ACC"]
+                        try:
+                            result["transcript"] = d[f.name[-13:-4]]
+                            yield id_, result
+                            id_ += 1
+                        except:
+                            print(f"unable to find transcript")
+                        
+