pelican-nlp 0.1.1__py3-none-any.whl → 0.1.3__py3-none-any.whl

pelican_nlp/Nils_backup/transcription/word_boundaries.py

@@ -0,0 +1,190 @@
+import sys
+import numpy as np
+
+from PyQt5.QtWidgets import (
+    QApplication,
+    QMainWindow,
+    QSplitter,
+    QWidget,
+    QVBoxLayout,
+    QTableWidget,
+    QTableWidgetItem,
+)
+from PyQt5.QtCore import Qt, pyqtSignal
+
+from matplotlib.backends.backend_qt5agg import FigureCanvasQTAgg as FigureCanvas
+from matplotlib.figure import Figure
+
+
+class WaveformCanvas(FigureCanvas):
+    boundary_changed = pyqtSignal(int, str, float)  # index of word, 'start'/'end', new position
+
+    def __init__(self, parent=None):
+        self.fig = Figure(figsize=(5, 4))
+        super(WaveformCanvas, self).__init__(self.fig)
+        self.setParent(parent)
+
+        self.ax = self.fig.add_subplot(111)
+        self.ax.set_title("Waveform with Word Boundaries")
+
+        # Generate a simple sine wave as the waveform
+        t = np.linspace(0, 10, 1000)
+        self.y = np.sin(2 * np.pi * 1 * t)
+        self.ax.plot(t, self.y)
+
+        # Example words with start and end times
+        self.words = [
+            {'word': 'Word1', 'start': 1.0, 'end': 2.0},
+            {'word': 'Word2', 'start': 2.5, 'end': 4.0},
+            {'word': 'Word3', 'start': 4.5, 'end': 6.0},
+            {'word': 'Word4', 'start': 6.5, 'end': 8.0},
+        ]
+
+        # Draw lines for start and end times
+        self.lines = []  # List of dicts: {'line': line_object, 'word_idx': idx, 'type': 'start'/'end'}
+
+        for idx, word in enumerate(self.words):
+            # Start time line
+            start_line = self.ax.axvline(word['start'], color='green', linestyle='--', picker=5)
+            self.lines.append({'line': start_line, 'word_idx': idx, 'type': 'start'})
+
+            # End time line
+            end_line = self.ax.axvline(word['end'], color='red', linestyle='--', picker=5)
+            self.lines.append({'line': end_line, 'word_idx': idx, 'type': 'end'})
+
+        self.dragging_line = None
+        self.prev_x = None
+
+        self.cid_press = self.fig.canvas.mpl_connect('button_press_event', self.on_press)
+        self.cid_release = self.fig.canvas.mpl_connect('button_release_event', self.on_release)
+        self.cid_motion = self.fig.canvas.mpl_connect('motion_notify_event', self.on_motion)
+
+    def on_press(self, event):
+        if event.inaxes != self.ax:
+            return
+        for line_dict in self.lines:
+            line = line_dict['line']
+            contains, _ = line.contains(event)
+            if contains:
+                self.dragging_line = line_dict
+                self.prev_x = event.xdata
+                break
+
+    def on_motion(self, event):
+        if self.dragging_line is None or event.inaxes != self.ax:
+            return
+        dx = event.xdata - self.prev_x
+        x = self.dragging_line['line'].get_xdata()[0] + dx
+        self.dragging_line['line'].set_xdata([x, x])
+        self.prev_x = event.xdata
+        self.fig.canvas.draw_idle()
+
+    def on_release(self, event):
+        if self.dragging_line is not None:
+            # Update the boundary positions
+            idx = self.dragging_line['word_idx']
+            boundary_type = self.dragging_line['type']
+            new_pos = self.dragging_line['line'].get_xdata()[0]
+            self.words[idx][boundary_type] = new_pos
+            self.dragging_line = None
+            self.prev_x = None
+            # Emit signal to update table
+            self.boundary_changed.emit(idx, boundary_type, new_pos)
+
+
+class MainWindow(QMainWindow):
+    def __init__(self):
+        super().__init__()
+        self.setWindowTitle("Waveform Editor")
+        self.resize(800, 600)
+
+        splitter = QSplitter(Qt.Horizontal)
+
+        # Left panel: Waveform
+        self.waveform_widget = QWidget()
+        waveform_layout = QVBoxLayout(self.waveform_widget)
+        self.canvas = WaveformCanvas(parent=self.waveform_widget)
+        waveform_layout.addWidget(self.canvas)
+        splitter.addWidget(self.waveform_widget)
+
+        # Right panel: Table
+        self.table_widget = QTableWidget()
+        self.table_widget.setColumnCount(3)
+        self.table_widget.setHorizontalHeaderLabels(["Word", "Start Time", "End Time"])
+        splitter.addWidget(self.table_widget)
+
+        self.setCentralWidget(splitter)
+
+        # Connect signals
+        self.canvas.boundary_changed.connect(self.on_boundary_changed)
+        self.table_widget.itemChanged.connect(self.on_table_item_changed)
+
+        self.update_table()
+
+    def update_table(self):
+        self.table_widget.blockSignals(True)  # Prevent signals while updating
+        words = self.canvas.words
+        self.table_widget.setRowCount(len(words))
+        for i, word in enumerate(words):
+            self.table_widget.setItem(i, 0, QTableWidgetItem(word['word']))
+            self.table_widget.setItem(i, 1, QTableWidgetItem(f"{word['start']:.2f}"))
+            self.table_widget.setItem(i, 2, QTableWidgetItem(f"{word['end']:.2f}"))
+        self.table_widget.blockSignals(False)
+
+    def on_boundary_changed(self, idx, boundary_type, new_pos):
+        # Update the table
+        self.table_widget.blockSignals(True)  # Prevent recursive updates
+        if boundary_type == 'start':
+            item = self.table_widget.item(idx, 1)
+            if item is not None:
+                item.setText(f"{new_pos:.2f}")
+        elif boundary_type == 'end':
+            item = self.table_widget.item(idx, 2)
+            if item is not None:
+                item.setText(f"{new_pos:.2f}")
+        self.table_widget.blockSignals(False)
+
+    def on_table_item_changed(self, item):
+        row = item.row()
+        col = item.column()
+        words = self.canvas.words
+
+        if col == 1:
+            # Start time changed
+            try:
+                new_start = float(item.text())
+                words[row]['start'] = new_start
+                # Update the line position in the waveform
+                for line_dict in self.canvas.lines:
+                    if line_dict['word_idx'] == row and line_dict['type'] == 'start':
+                        line = line_dict['line']
+                        line.set_xdata([new_start, new_start])
+                        self.canvas.draw()
+                        break
+            except ValueError:
+                pass  # Invalid input, ignore
+        elif col == 2:
+            # End time changed
+            try:
+                new_end = float(item.text())
+                words[row]['end'] = new_end
+                # Update the line position in the waveform
+                for line_dict in self.canvas.lines:
+                    if line_dict['word_idx'] == row and line_dict['type'] == 'end':
+                        line = line_dict['line']
+                        line.set_xdata([new_end, new_end])
+                        self.canvas.draw()
+                        break
+            except ValueError:
+                pass  # Invalid input, ignore
+
+
+def main():
+    app = QApplication(sys.argv)
+    window = MainWindow()
+    window.show()
+    sys.exit(app.exec_())
+
+
+if __name__ == "__main__":
+    main()

pelican_nlp/Silvia_files/Opensmile/opensmile_feature_extraction.py

@@ -0,0 +1,66 @@
+import csv
+import os
+from typing import List
+
+import audiofile
+import opensmile
+import glob
+
+audio_dir = "C:\\Users\\Acer\\Documents\\Silvia\\UMC_Language and schizophrenia\\Research assistant\\Projects\\Language and sex hormones\\P009"
+files: List[str] = glob.glob('C:\\Users\\Acer\\Documents\\Silvia\\UMC_Language and schizophrenia\\PhD\\Projects\\Interns\\Language and sex hormones\\P009\\*.wav')
+
+
+#extract whole audiofile
+#if wish to be shortended, fill in argument 'duration'
+results = []
+for file in files:
+    print(file)
+    storage = {}
+    signal, sampling_rate = audiofile.read(
+        file,
+        always_2d=True,
+        #duration=
+        #offset=
+    )
+
+#extract eGeMAPSv02 feature set
+    smile = opensmile.Smile(
+        feature_set=opensmile.FeatureSet.eGeMAPSv02,
+        feature_level=opensmile.FeatureLevel.Functionals,
+    )
+
+    ##print(smile.feature_names)
+
+
+    output = smile.process_signal(
+        signal,
+        sampling_rate
+    )
+    print(output)
+
+#save output
+    storage['p_nr'] = file[0:4]
+    for feature in smile.feature_names:
+        storage[feature] = output[feature]
+
+    results.append(storage)
+
+csv_columns = results[0].keys()
+csv_file = "C:\\Users\\Acer\\Documents\\Silvia\\UMC_Language and schizophrenia\\Research assistant\\Projects\\Language and sex hormones\\Opensmile_results.csv"
+
+with open(csv_file, 'w', newline="") as csvfile:
+    writer = csv.writer(csvfile)
+    writer.writerow(csv_columns)
+    for i in range (len(results)): #iterate over particpants' results
+        data = results[i]
+        new_array = []
+        for t in data: #iterate all features in participants' results
+            if t == 'p_nr': #in first column, insert particpant number
+                new_array.append(files[i].split(os.sep)[-1]) #filesplit takes only participant number (and not the whole path) #append inserts data in the empty array
+            else: #for all other columns insert feature value
+                new_array.append(float(data[t]))
+        writer.writerow(new_array)
+
+
+
+

pelican_nlp/Silvia_files/prosogram/prosogram.py

@@ -0,0 +1,104 @@
+# 1. packages
+import os
+import subprocess
+import shutil
+import glob
+import pandas as pd
+from tqdm import tqdm
+
+# 2. functions
+def extract_profile(wav_path):
+    """
+    
+
+    Args:
+        wav_path (str): pathway to the speech file.
+
+    Returns:
+        profile (DataFrame) : the result of prosogram.
+
+    """
+    
+    # get directory
+    original_cwd = os.getcwd()
+    script_dir = os.path.dirname(os.path.abspath("__file__"))
+
+    # create a filefoder to store temporary file
+    tmp_dir = os.path.abspath("./tmp")
+    shutil.rmtree(tmp_dir, ignore_errors=True)
+    os.makedirs(tmp_dir, exist_ok=True)
+
+    try:
+        # copy all prosogram scripts into the temporary folder 
+        for p in glob.glob(os.path.join(script_dir, "*.praat")):
+            shutil.copyfile(p, os.path.join(tmp_dir, os.path.basename(p)))
+
+        shutil.copyfile("Praat.exe", os.path.join(tmp_dir, "Praat.exe"))
+
+        # copy the speech file into the temporary folder
+        wav_copy_path = os.path.join(tmp_dir, os.path.basename(wav_path))
+        shutil.copyfile(wav_path, wav_copy_path)
+
+        # create a praat work file in the temporary folder with setting
+
+        # use default setting here
+        # if needed, add custome settings in prosogram_variants of job_content
+        # for example, if the custom settings are: g=0.32 dg=20 dmin=0.035
+        # the job_content should be:
+        # """
+        # include prosomain.praat
+        # @prosogram_variants: "file={os.path.abspath(wav_copy_path)} save=yes draw=no 
+        # g=0.32 dg=20 dmin=0.035"
+        # exit
+        # """
+
+        job_contents = f"""include prosomain.praat
+        @prosogram: "file={os.path.abspath(wav_copy_path)} save=yes draw=no"
+        exit"""
+
+        job_path = os.path.join(tmp_dir, "job.praat")
+        with open(job_path, "w") as job_file:
+            job_file.write(job_contents)
+
+        os.chdir(tmp_dir)
+
+        # create a command line to run praat
+        invocation = f"Praat.exe --run {os.path.abspath(job_path)}"
+        status, output = subprocess.getstatusoutput(invocation)
+        # Give a warning when the processing failed (status is not 0)
+        if status != 0:
+            print(output)
+            raise Warning("FAILED: Praat failed! The outputs are printed above.")
+
+        # Read the result file 
+        profile_path = os.path.join(tmp_dir, 
+                                    wav_copy_path.replace(".wav", "") + \
+                                        "_profile_data.txt")
+        profile = pd.read_csv(profile_path, sep="\t")
+    
+    finally:
+        # delete all  contents in the temporary filefolder
+        shutil.rmtree(tmp_dir, ignore_errors=True)
+        os.chdir(original_cwd)
+    
+    return profile 
+
+
+# 4. commands
+path = 'Audios/'
+wav_paths = glob.glob(os.path.join(path, "*.wav"))
+
+# get all data within one data frame
+profile = pd.DataFrame()
+for wav_path in tqdm(wav_paths):
+    profile = pd.concat([profile,
+                         extract_profile(wav_path)])
+
+# add the identifier column
+profile['Audios'] = wav_paths
+profile = profile.reindex(columns=['Audios']+profile.columns.tolist()[1:])
+
+# save as a csv file
+profile.to_csv('Prosogram_results.csv', encoding='utf-8-sig', index=False)
+
+

pelican_nlp/__init__.py

@@ -1,5 +1,5 @@
 # Version and metadata
-__version__ = "0.1.1"
+from ._version import __version__
 __author__ = "Yves Pauli"
 
 try:

pelican_nlp/_version.py

@@ -0,0 +1 @@
+__version__ = "0.1.3"

pelican_nlp/configuration_files/config_audio.yml

@@ -0,0 +1,150 @@
+# Master Configuration File
+# ========================
+
+# Basic Settings
+# -------------
+input_file: "audio"  # Options: 'text' or 'audio'
+PATH_TO_PROJECT_FOLDER: "/home/yvespauli/PycharmProjects/Silvia"
+language: "dutch"  # Options: 'german', 'english'
+recompute_everything: true  # If false, reuses previously computed results
+output_document_information: true
+# Task Configuration
+# -----------------
+task_name: "monologue"  # Options: 'fluency', 'interview'
+fluency_task: &fluency_flag false  # Flag for fluency-specific settings
+discourse: &discourse_flag false  # Flag for discourse-specific settings
+corpus_names:  # List of task corpora
+  - "interv-removed"
+
+# Session and Subject Settings
+# --------------------------
+multiple_sessions: true
+number_of_subjects: null  # If null, auto-detected
+number_of_speakers: 1
+subject_speakertag: null  # Speaker tag for subject (e.g., "B")
+
+# Document Structure
+# ----------------
+has_multiple_sections: false
+has_section_titles: false
+section_identification: null  # e.g., "Section:"
+number_of_sections: 1  # If null, auto-detected
+
+# Processing Pipeline
+# -----------------
+pipeline_options:
+  quality_check: false
+  clean_text: true
+  tokenize_text: false
+  normalize_text: false
+
+# Audio Processing
+# ---------------
+opensmile_feature_extraction: true
+prosogram_extraction: false
+
+opensmile_configurations:
+  always_2d: true
+  duration: null
+  offset: null
+
+  feature_set: opensmile.FeatureSet.eGeMAPSv02
+  feature_level: opensmile.FeatureLevel.Functionals
+
+
+# Metric Extraction
+# ---------------
+metric_to_extract: null  # Options: 'embeddings', 'logits'
+extract_logits: false
+extract_embeddings: false
+
+# Cleaning Options
+# --------------
+cleaning_options:
+  general_cleaning: true
+  remove_punctuation: false
+  lowercase: true
+  remove_brackets_and_bracketcontent: false
+  remove_timestamps: false
+  timestamp_pattern_example: null  # e.g., "#00:00:23-00#"
+  # Fluency-specific options
+  fluency_task: *fluency_flag
+  word_splitter: ';'
+  remove_hyphens: true
+  remove_duplicates: true
+
+general_cleaning_options:
+  strip_whitespace: true
+  merge_multiple_whitespaces: true
+  remove_whitespace_before_punctuation: true
+  merge_newline_characters: true
+  remove_backslashes: true
+
+# Embedding Options
+# ---------------
+options_embeddings:
+  tokenization_method: "whitespace"  # Options: 'whitespace', 'model'
+  model_name: "fastText"  # Options: 'fastText', 'xlm-roberta-base'
+  pytorch_based_model: false
+  method: "model_instance"
+  max_length: 512
+  clean_embedding_tokens: true
+  remove_punctuation: false
+  lowercase: false
+  keep_speakertags: false
+  semantic-similarity: true
+  window_size: null
+  clean_tokens: true
+  divergence_from_optimality: false
+  output_options:
+    exclude_special_tokens: true
+    remove_'_'_character: true
+    remove_speaker_labels: true
+    remove_punctuation_and_symbols: true
+    remove_brackets_and_content: true
+
+# Logits Options
+# -------------
+options_logits:
+  chunk_size: 128
+  overlap_size: 64
+  tokenization_method: "model"
+  model_name: "DiscoResearch/Llama3-German-8B-32k"
+  remove_punctuation: true
+  lowercase: true
+  keep_speakertags: true
+
+# Analysis Options
+# --------------
+options_semantic-similarity:
+  window_sizes:  # 'all' or window size as integer
+    - 2
+    - 8
+
+options_dis_from_randomness:
+  window_size: 8
+  min_len: null
+  bootstrap: 10000
+  shuffle_mode: 'include0_includeN'
+  parallel_computing: false
+
+# Normalization Options
+# -------------------
+normalization_options:
+  method: "lemmatization"  # Options: 'lemmatization', 'stemming'
+
+# Filename Configuration
+# --------------------
+filename_components:
+  subject: true    # mandatory
+  session: false
+  task: true       # mandatory
+  task_addition: false
+  corpus: true
+  metric: true
+  additional_tags: []
+
+document_information_output:
+  parameters:
+    - subject_ID
+    - recording_length

pelican_nlp/configuration_files/config_discourse.yml

@@ -0,0 +1,104 @@
+# Configuration file for discourse protocols
+#=======================================
+input_file: "text" #or 'audio'
+discourse: &discourse_flag true
+#=====================================
+
+#general configurations; always adapt
+PATH_TO_PROJECT_FOLDER: "/home/yvespauli/PycharmProjects/Morteza/"
+language: "german" # Possibly add options for German and English
+
+task_name: "interview" # Give name of task used for creation of the input file (e.g., ['fluency', 'interview'])
+corpus_names:
+  - "schizophrenia"
+
+metric_to_extract: "embeddings" #Possible options: 'logits' or 'embeddings'
+
+number_of_speakers: 3
+subject_speakertag: "B"
+#=========================================================
+
+#Optional configurations; Change with preference. However, default settings recommended
+fluency_task: &fluency_flag false
+cleaning_options:
+  general_cleaning: true # General cleaning options used for most text preprocessing, default: True.
+  remove_brackets_and_bracketcontent: true
+  remove_timestamps: true
+  timestamp_pattern_example: "#00:00:19-0#"
+  remove_punctuation: false
+  lowercase: false
+  #Options for fluency tasks
+  fluency_task: *fluency_flag
+  word_splitter: null
+  remove_hyphens: null
+  remove_duplicates: null
+
+options_logits:
+  chunk_size: 128
+  overlap_size: 64
+  tokenization_method: "model"
+  #method: "model_instance" # Options: model, regex, nltk, etc.
+  model_name: "DiscoResearch/Llama3-German-8B-32k" # Replace with your model instance name
+  remove_punctuation: true
+  lowercase: true
+  keep_speakertags: true
+
+options_embeddings:
+  tokenization_method: "whitespace" #"model" or "whitespace"
+  max_length: 512 #max sequence length
+  model_name: "fastText" #e.g. "fastText", "xlm-roberta-base"
+  pytorch_based_model: false
+  method: "model_instance"
+  remove_punctuation: false
+  lowercase: false
+  keep_speakertags: true
+  clean_embedding_tokens: true
+  output_options:
+    exclude_special_tokens: true
+    remove_'_'_character: true
+    remove_speaker_labels: true
+    remove_punctuation_and_symbols: true
+    remove_brackets_and_content: true
+  semantic-similarity: false
+  window_size: null
+  clean_tokens: false
+  divergence_from_optimality: false
+#================================================================================
+
+#Extra configurations:
+pipeline_options:
+  quality_check: false
+  clean_text: true
+  tokenize_text: false
+  normalize_text: false
+
+general_cleaning_options:
+  strip_whitespace: true
+  merge_multiple_whitespaces: true
+  remove_whitespace_before_punctuation: true
+  merge_newline_characters: true
+  remove_backslashes: true
+
+has_multiple_sections: false #evaluated independently
+has_section_titles: false
+section_identification: null #e.g. "Section:", 'null' if file does not have multiple sections, use pattern that is unlikely to appear in rest of transcript
+number_of_sections: null #if 'null' number of sections automatically detected, however, specifying number recommended if known.
+
+# Options for extract_embeddings
+window_sizes: [2]
+metric_function: cosine_similarity
+aggregation_functions: mean_of_means
+
+normalization_options:
+  method: "lemmatization" #Options: lemmatization or stemming
+#================================================================
+
+#Detail configurations; Changes optional, mostly used for quality checking / error handling
+number_of_subjects: null # Specify number of subjects; if 'null', number of subjects is automatically detected
+multiple_sessions: false # Set to True if multiple sessions per subject
+
+recompute_everything: true #If set to 'false' pelican-nlp will try to reuse previously computed results stored on your drive
+
+create_aggregation_of_results: false
+
+