clarin-pl/2021-punctuation-restoration

Dataset Overview

Dataset Description

• The 2021‑punctuation‑restoration dataset is primarily used to restore punctuation in the output of automatic speech recognition (ASR) systems.
• It contains Polish text and audio data, divided into two parts: WikiTalks (conversational) and WikiNews (informational).
• The dataset aims to improve the readability of ASR‑generated transcripts and may also enhance performance on other NLP tasks.
• It comprises 1,200 texts, totaling over 240,000 words, spoken by over 100 different native speakers.
• The dataset provides training and test splits, with the test set containing ASR transcriptions of texts from both sources (WikiNews and WikiTalks).

Dataset Composition

• Training Size: 800 texts
• Development Size: 0 texts
• Test Size: 200 texts

Data Format

• Input Format: TSV file containing text ID and lower‑cased input text without punctuation.
• Output Format: Same number of lines as the input file, each line containing the text with punctuation.

Evaluation

• Metrics: Accuracy, recall and F1 score, computed separately for each punctuation mark.
• Final Score: Weighted average of F1 scores across punctuation marks.

Download

• Dataset hosted on GitHub: https://github.com/poleval/2021-punctuation-restoration
• Additional training data and resources available via Google Drive.

License

• Creative Commons – Attribution‑NonCommercial‑NoDerivatives 4.0 International (CC BY‑NC‑ND 4.0)