Parallel Text Alignment Builder

How to use Parallel Text Alignment Builder

Assemble a human-reviewed parallel text dataset from already segmented documents. One-to-many and many-to-one groupings are explicit, and reused IDs or crossing groups are refused rather than silently rearranged into a plausible translation.

1. Enter source segments json, target segments json, ordered alignment groups json.
2. Choose the settings for parallel text alignment builder, then run the tool.
3. Review the result and any limitations before copying or downloading it.

Good to know

• 200 segments per side, 1,000 characters per segment; 200 monotonic links with 1–10 IDs per side and 250,000 linked text characters. IDs are bounded and unique per side; ordered source inputs define the sequence.
• Only supplied groupings are exported. No translation, sentence splitting, similarity/confidence score, semantic equivalence or Moses/importer compatibility is promised. CSV formula-like cells are neutralized; unlinked IDs remain visible.
• Processing is local. Output is limited to 1,000,000 text characters unless a stricter file limit is stated.