feat(readme): added error handler for the case of missing music-tag library, updated readme

README.md

@@ -1,3 +1,78 @@
 ---
+language:
+- ru
+- en
 license: apache-2.0
+tags:
+- not-for-all-audiences
+- art
+- humour
+- jokes
+annotation_creators:
+- crowdsourced
+- original
+- machine-generated
+language_creators:
+- crowdsourced
+- original
+- machine-generated
+pretty_name: baneks-speech
+size_categories:
+- 10K<n<100K
+task_categories:
+- text-to-speech
+- automatic-speech-recognition
 ---
+
+# Dataset card for baneks-speech
+
+## Table of contents
+
+- [Dataset description](#dataset-description)
+    - [Dataset summary](#dataset-summary)
+- [Dataset structure](#dataset-structure)
+    - [Dataset instance](#dataset-instance)
+    - [Dataset fields](#dataset-fields)
+
+## Dataset description
+
+- **Homepage:** [baneks-speech homepage]()
+- **Repository:** [baneks-speech repository](https://huggingface.co/datasets/zeio/baneks-speech)
+- **Point of contact:** [Zeio Nara](mailto:zeionara@gmail.com)
+- **Dataset version:** `10.10.2023`
+
+### Dataset summary
+
+This dataset contains speech generated for anekdotes parsed from a few vk social network communities using [stc model](https://cloud.speechpro.com/home).
+The dataset corresponds to the **default** configuration from the [baneks](https://huggingface.co/datasets/zeio/baneks) dataset.
+Since the dataset is regularly updated, there is no fixed number of entries, so stay tuned.
+
+## Dataset structure
+
+### Data instance
+
+An example of an entry from the dataset is given below:
+
+```json
+{
+    'text': 'Сидят русский и казах Русский спрашивает: - Слушай, а у тебя жена есть? - Жок - Молодец! Я бы свою тоже сжег нахуй!',
+    'audio': {
+        'path': '/root/.cache/huggingface/datasets/downloads/extracted/e8abea39e83c61e4a60c5a4b0661dc044ae82cfcd84b26f966b87999f73ae92e/00476018.anekdotikategoriib.mp3',
+        'array': array([ 8.09500818e-07,  1.27653129e-06,  4.17583010e-07, ..., -8.52341486e-07,  9.19626189e-07,  1.67368569e-06]),
+        'sampling_rate': 22050
+    },
+    'artist': 'Vladimir_n',
+    'id': 476018,
+    'source': 'anekdotikategoriib'
+}
+```
+
+### Data fields
+
+Each dataset entry therefore consists of the following fields:
+
+- `text` - text representation of the anecdote;
+- `id` - id of the corresponding post;
+- `source` - community name in which the corresponding post has been published;
+- `artist` - identifier of the voice which was used for speech generation;
+- `audio` - audio data read from an `mp3` file.

baneks-speech.py

@@ -5,10 +5,10 @@ from pathlib import Path
 import datasets
 from datasets import Features, Value, Audio
 
-# os.system('pip install music-tag')
-
-from music_tag import load_file
-
+try:
+    from music_tag import load_file
+except Exception:
+    load_file = None
 
 _DESCRIPTION = """
 This dataset contains speech generated for anecdotes from the [baneks dataset](https://huggingface.co/datasets/zeio/baneks)
@@ -21,7 +21,7 @@ _LICENSE = "Apache License Version 2.0"
 _URL = "https://huggingface.co/datasets/zeio/baneks-speech/resolve/main/speech/{first:06d}-{last:06d}.tar.xz"
 
 
-_N_TOTAL = 46_554
+_N_TOTAL = 39_536
 _N_BATCH = 1_000
 
 
@@ -56,7 +56,7 @@ class BaneksSpeech(datasets.GeneratorBasedBuilder):
 
             paths.append(dl_manager.download_and_extract(url))
 
-            break
+            # break # uncomment this line to download only first partition of the dataset which weighs approximately 70 Mb
 
         return [
             datasets.SplitGenerator(
@@ -77,12 +77,10 @@ class BaneksSpeech(datasets.GeneratorBasedBuilder):
                 id_ = int(components[0])
                 source = components[1]
 
-                meta = load_file(file)
-
-                # artist = meta['artist']
-                # text = meta['lyrics']
-
-                # print(id_, source, artist, text)
+                meta = {
+                    'lyrics': 'Here will be lyrics which are loaded from the mp3 file metadata',
+                    'artist': 'Here will be artist which is loaded from the mp3 file metadata'
+                } if load_file is None else load_file(file)
 
                 yield f'{id_:08d}-{source}', {
                     'text': meta['lyrics'],