Datasets:

THUgewu
/

SpokenArabicDigits

license: cc-by-4.0
task_categories:
  - time-series-forecasting
tags:
  - tsfile
  - timeseries
  - time-series
  - classification
pretty_name: SpokenArabicDigits (TsFile)
size_categories:
  - 10K<n<100K
configs:
  - config_name: default
    data_files:
      - split: train
        path: SpokenArabicDigits_train.tsfile
      - split: test
        path: SpokenArabicDigits_test.tsfile

SpokenArabicDigits (TsFile)

Apache TsFile version of the SpokenArabicDigits UEA classification subset of thuml/Time-Series-Library.

Overview

MFCC of spoken Arabic digits from 88 speakers; classify the digit.

Train samples: 6,599 • Test samples: 2,199.
Dimensions (channels): 13 • Series length: 39.
Classes: 10.

Each sample is an independent multivariate series. TRAIN and TEST are stored as two separate TsFiles (SpokenArabicDigits_train.tsfile / SpokenArabicDigits_test.tsfile).

Schema (TsFile structure)

sample_index (TAG) — one device per sample. Query one sample with WHERE sample_index=0.
Time (INT64) — within-sample position (0..38); a frame index, not a wall-clock timestamp (the sktime .ts source has none).
dim_0 … dim_12 (FIELD, FLOAT) — the 13 channels at each position.
class_label (FIELD, STRING) — the sample's class, repeated on every row.

Converted from the sktime .ts format. Nothing is dropped: every dimension, time point, and label is preserved.

Usage

Read the .tsfile files with the Apache TsFile Java or Python SDK.

Source & license

Original dataset: https://huggingface.co/datasets/thuml/Time-Series-Library (subset SpokenArabicDigits)
Author / publisher: thuml (Tsinghua University)
Paper: https://arxiv.org/abs/2407.13278
License: CC BY 4.0