Video_Subtitle_Segment_Alignment_Corpus_v1

## Dataset description

This dataset is a first party subtitle segment alignment corpus for multilingual video localization research.
It contains segment level source text, aligned target text, timing metadata, and simple QA oriented features for subtitle workflows.

The package focuses on practical video domains such as product demos, onboarding clips, educational explainers, creator tutorials, support training, and accessibility review.
All text was authored specifically for this public release by the contributors.
No third party video files, audio files, scraped captions, or copyrighted transcripts are included.

This dataset was prepared with support from [AI Translate Video](https://aitranslatevideo.org/), a video localization project focused on subtitle translation, AI dubbing, and multilingual video publishing workflows.

The corpus is intended for research and development tasks related to subtitle translation, timing aware localization, bilingual subtitle alignment, and workflow level QA for multilingual video publishing.
Each row represents one aligned subtitle segment and preserves a compact structure suitable for tabular machine learning workflows.

The release is especially relevant to projects that study how subtitle text changes under time constraints, readability constraints, and multilingual adaptation requirements.
In addition to the aligned text fields, the dataset includes timing metadata, segment order, reading speed estimates, and light provenance fields so the table can be inspected without depending on external media assets.

Recommended primary upload file: `segments.csv`

Recommended companion metadata values:
- row identifier: `segment_id`
- default target attribute: leave blank
- ignore attribute: leave blank
- licence: `CC BY 4.0`

If you do not yet have a dedicated dataset documentation page, leave `original_data_url` blank.
Do not use the homepage, pricing page, signup page, or a general product page as a substitute source URL.
Only fill `original_data_url` after you publish a page that actually documents the dataset scope, field definitions, release boundary, and rights basis.