add scripts

anton_old_script.py

@@ -0,0 +1,54 @@
+import pandas as pd
+import torchaudio
+from glob import glob
+import os
+import csv
+from tqdm import tqdm
+import pandas
+
+sentence_bounds = {'!', '.', ';', '?', '…'}
+
+files = list(sorted(glob("earnings22/media/*.mp3")))
+metadata = []
+for audio_file in tqdm(files):
+    file_id = audio_file.split("/")[-1].split(".")[0]
+    nlp_file = f"earnings22/aligned/{file_id}.nlp"
+
+    speech, sr = torchaudio.load(audio_file)
+    with open(nlp_file, "r") as nlp:
+        start, end = None, None
+        sentence = ""
+        segment_id = 0
+        csvreader = csv.DictReader(nlp, delimiter="|")
+        for row in csvreader:
+            punct = row["punctuation"].strip()
+            sentence += row["token"]
+            if punct:
+                sentence += punct
+            sentence += " "
+            if start is None and row["ts"].strip():
+                start = float(row["ts"]) - 0.1
+            if row["endTs"].strip():
+                end = float(row["endTs"]) + 0.1
+            if punct in sentence_bounds and start is not None and end is not None:
+                segment = speech[:, int(start*sr):int(end*sr)+1].contiguous()
+
+                os.makedirs(f"earnings22/segmented/{file_id}", exist_ok=True)
+                torchaudio.save(f"earnings22/segmented/{file_id}/{segment_id}.wav",
+                                segment, sr, encoding="PCM_S", bits_per_sample=16)
+
+                data_row = {
+                    "source_id": f"{file_id}",
+                    "segment_id": segment_id,
+                    "file": f"{file_id}/{segment_id}.wav",
+                    "start_ts": start,
+                    "end_ts": end,
+                    "sentence": sentence.strip()
+                }
+                metadata.append(data_row)
+
+                start, end = None, None
+                sentence = ""
+                segment_id += 1
+    pd.DataFrame(metadata).to_csv("earnings22/segmented/metadata.csv", index=False)
+

generate_metadata_file.py

@@ -0,0 +1,140 @@
+import csv
+import argparse
+from pathlib import Path
+
+
+SEPARATORS = {";", "!", ".", "?", "…"}  #, '-', '–', ':'}
+
+
+# logging.basicConfig(format='%(asctime)s %(levelname)s: %(message)s', level=logging.INFO)
+
+
+def get_args():
+    parser = argparse.ArgumentParser("Prepare sharded audio archives")
+    parser.add_argument(
+        "--source_dir",
+        help="directory with original word-level transcriptions files",
+        type=str,
+        required=True,
+    )
+    parser.add_argument(
+        "--output_filename",
+        required=True,
+        type=str,
+    )
+    parser.add_argument(
+        "--split_on_speaker",
+        action='store_true',
+    )
+    return parser.parse_args()
+
+
+def generate_metadata_for_file(file, writer, split_on_speaker, n_filtered_sentences):
+    audio_id = file.name.split(".")[0]
+    with open(file) as csvfile:
+        tokens = list(csv.DictReader(csvfile, delimiter="|"))
+        # start_time, end_time = None, None
+        sentence_tokens = []
+        sentence_text = ""
+        segment_id = 0
+        previous_speaker = tokens[0]["speaker"]
+        for token in tokens:
+            punct = token["punctuation"].strip()
+            speaker = token["speaker"]
+            sentence_tokens.append(token)
+            if split_on_speaker and speaker != previous_speaker and len(sentence_tokens) > 1:
+                start_time, end_time = sentence_tokens[0]["ts"].strip(), sentence_tokens[-2]["endTs"].strip()
+                if start_time == "":
+                    sentence_tokens = []
+                    sentence_text = ""
+                    previous_speaker = speaker
+                    n_filtered_sentences += 1
+                    continue
+                if end_time == "":
+                    sentence_tokens = []
+                    sentence_text = ""
+                    previous_speaker = speaker
+                    n_filtered_sentences += 1
+                    continue
+                writer.writerow({
+                    "audio_id": audio_id,
+                    "segment_id": segment_id,
+                    "start_time": start_time,
+                    "end_time": end_time,
+                    "speaker": previous_speaker,
+                    "text": sentence_text.strip(),
+                })
+                sentence_tokens = [sentence_tokens[-1]]
+                sentence_text = sentence_tokens[-1]["token"]
+                segment_id += 1
+            else:
+                sentence_text += token["token"]
+            if punct:
+                sentence_text += punct
+            sentence_text += " "
+
+            if punct in SEPARATORS:
+                start_time, end_time = sentence_tokens[0]["ts"].strip(), sentence_tokens[-1]["endTs"].strip()
+                if start_time == "":
+                    sentence_tokens = []
+                    sentence_text = ""
+                    previous_speaker = speaker
+                    n_filtered_sentences += 1
+                    continue
+                if end_time == "":
+                    sentence_tokens = []
+                    sentence_text = ""
+                    previous_speaker = speaker
+                    n_filtered_sentences += 1
+                    continue
+                if len(list(set([t["speaker"] for t in sentence_tokens]))) > 1:
+                    if split_on_speaker:
+                        print("multiple speakers! they should have been filtered on the prev step, maybe something went wrong")
+                        print(audio_id, sentence_tokens)
+                    else:
+                        print("multiple speakers, filtering them")
+                        print(audio_id, sentence_tokens)
+                        sentence_tokens = []
+                        sentence_text = ""
+                        previous_speaker = speaker
+                        n_filtered_sentences += 1
+                        continue
+
+                writer.writerow({
+                    "audio_id": audio_id,
+                    "segment_id": segment_id,
+                    "start_time": start_time,
+                    "end_time": end_time,
+                    "speaker": speaker,
+                    "text": sentence_text.strip(),
+                })
+                sentence_tokens = []
+                sentence_text = ""
+                segment_id += 1
+
+            previous_speaker = speaker
+
+        return n_filtered_sentences
+
+
+def generate_metadata(source_dir, output_filename, split_on_speaker):
+    transcription_files = sorted(list(Path(source_dir).glob("**/*.nlp")))
+    with open(output_filename, "w") as csv_writer:
+        writer = csv.DictWriter(csv_writer, fieldnames=["audio_id", "segment_id", "start_time", "end_time", "speaker", "text"])
+        writer.writeheader()
+        n_filtered_sentences = 0
+        for transcription_file in transcription_files:
+            n_filtered_sentences = generate_metadata_for_file(transcription_file, writer, split_on_speaker, n_filtered_sentences)
+    print("sentences filtered: ", n_filtered_sentences)
+
+def main():
+    args = get_args()
+    generate_metadata(
+        source_dir=args.source_dir,
+        output_filename=args.output_filename,
+        split_on_speaker=args.split_on_speaker
+    )
+
+
+if __name__ == "__main__":
+    main()