Correct dataset card: both annotators scored all 81 pairs (score is their mean); half two is collected web text, not original writing; drop failing raw_sentences viewer config

Merge nursimakgul's 38 scored pairs; dataset is now 81 pairs from two contributors

Add 43 manually scored Turkish STS pairs as parquet, with pairing metadata and build script

Remove pairing protocol

Replace event_id example with a two-newspaper report of the same incident

Add Turkish STS raw sentence pool (527 sentences), pairing protocol and collector

initial commit