Datasets:

mah92
/

Mana-FA-Complete-Sentences-22KHz-Dataset

Exception:    SplitsNotFoundError
Message:      The split names could not be parsed from the dataset config.
Traceback:    Traceback (most recent call last):
                File "/usr/local/lib/python3.14/site-packages/datasets/inspect.py", line 286, in get_dataset_config_info
                  for split_generator in builder._split_generators(
                                         ~~~~~~~~~~~~~~~~~~~~~~~~~^
                      StreamingDownloadManager(base_path=builder.base_path, download_config=download_config)
                      ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                  )
                  ^
                File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/folder_based_builder/folder_based_builder.py", line 250, in _split_generators
                  raise ValueError(
                      "`file_name`, `*_file_name`, `file_names` or `*_file_names` must be present as dictionary key in metadata files"
                  )
              ValueError: `file_name`, `*_file_name`, `file_names` or `*_file_names` must be present as dictionary key in metadata files
              
              The above exception was the direct cause of the following exception:
              
              Traceback (most recent call last):
                File "/src/services/worker/src/worker/job_runners/config/split_names.py", line 68, in compute_split_names_from_streaming_response
                  for split in get_dataset_split_names(
                               ~~~~~~~~~~~~~~~~~~~~~~~^
                      path=dataset,
                      ^^^^^^^^^^^^^
                      config_name=config,
                      ^^^^^^^^^^^^^^^^^^^
                      token=hf_token,
                      ^^^^^^^^^^^^^^^
                  )
                  ^
                File "/usr/local/lib/python3.14/site-packages/datasets/inspect.py", line 340, in get_dataset_split_names
                  info = get_dataset_config_info(
                      path,
                  ...<6 lines>...
                      **config_kwargs,
                  )
                File "/usr/local/lib/python3.14/site-packages/datasets/inspect.py", line 291, in get_dataset_config_info
                  raise SplitsNotFoundError("The split names could not be parsed from the dataset config.") from err
              datasets.inspect.SplitsNotFoundError: The split names could not be parsed from the dataset config.

Need help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.

Mana FA — Complete Sentences (22.05 kHz)

A Persian single-speaker corpus of 15,000 complete sentences derived from MahtaFetrat/Mana-TTS (CC-0, 114 h, 44.1 kHz, single speaker).

Why this dataset exists

ManaTTS is distributed as short machine-cut chunks: the audio was segmented without regard to sentence boundaries, so most chunks stop in the middle of a sentence. In the subset we had been using, 68.4 % of the chunks ended without any sentence-final punctuation, and their continuation was frequently a different file (e.g. 10-102 ends with «...در دسترس بودن بریل ملزوماتی» and 10-103 continues «دارد که مهم‌ترین آن...»). Training a TTS model on such fragments teaches it to truncate sentences.

What was done

Downloaded the full ManaTTS release (77 parquet parts, 87,017 chunks, ~114 h).
Merged consecutive chunks within a chapter (ordered by chunk number) into one utterance, closing a unit at the first chunk whose transcript ends with sentence-final punctuation (. ! ? ؟ …). 50 ms of silence is inserted at a chunk join (the original split was at a pause).
Quality filter using the corpus' own match quality field: the sentence-closing chunk must be HIGH; units whose every chunk is HIGH were taken first (8,190), the remainder were chosen with the fewest MIDDLE chunks and the lowest CER.
Duration window 2–40 s. No text normalisation or filtering beyond the original transcripts; 2 duplicate source ids were disambiguated with a _v2 suffix.
Resampled 44.1 kHz → 22.05 kHz with soxr (VHQ quality), float64 intermediate, PCM_16 mono output — anti-aliased, no aliasing products at the new Nyquist.

wav_01/ ... wav_15/   15,000 mono PCM_16 WAV files, 22,050 Hz (1,000 per folder —
                       the Hub rejects more than 10,000 files in one directory)
metadata.csv          file|transcript|duration|match_quality|CER|n_source_chunks

15,000 utterances, 53.6 hours, median 10.99 s (2.0–40.3 s)
11,273 utterances were assembled from 2+ source chunks, 3,727 are single-chunk
Every transcript ends with sentence-final punctuation
Speaker: the single narrator of ManaTTS (not a new recording)

Limitations

No phone-level alignments are included; runs of merged chunks may contain a short (50 ms) seam.
The text is the corpus' transcript as-is (no ezafe diacritisation).
The source chunks carry their own quality label; the label of the closing chunk is reported in metadata.csv, internal chunks may be MIDDLE.

Licence

The source corpus is CC-0; this derivative is released under CC-0 as well. Please cite the original ManaTTS work if you use this data.

Downloads last month: 45

Total file size:

8.51 GB

Mana FA — Complete Sentences (22.05 kHz)

Why this dataset exists

What was done

Contents

Limitations

Licence