BCAI Finance Kor Embedding Pair: Korean Financial Text Pairs for Model Training

45,589 Korean financial text pairs serve as training data for contrastive learning and as a retrieval corpus. The dataset is authored by BCCard and was last updated on June 11, 2026. It likely contains anchor and positive sentence pairs derived from financial documents.

Use Cases

• Train sentence embedding models via contrastive learning based on the described (anchor, positive) pairs.
• Evaluate information retrieval systems using the described corpus of approximately 36,000 unique text chunks.
• Fine-tune language models for financial domain adaptation based on Korean financial text pairs.

Strengths

• Contains 45,589 explicitly stated (anchor, positive) text pairs for training.
• Serves a dual purpose as both training data and a retrieval evaluation corpus.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count for the retrieval corpus is approximate ('≈ 36k unique chunks'), which may limit suitability assessment.

Provenance

Source

BCCard

Freshness

Last updated 2026-06-11 06:52:30