Swiss Monetary Policy Communication Corpus (SMPC)

SMPC v1.0.4 is a metadata-sanitization successor to SMPC v1.0.3. It contains 2547 text-bearing document-language rows representing 833 SNB source documents, 132 policy events, duplicate diagnostics, leak-aware splits, provenance, and 2547 x 1024 normalized multilingual E5 document embeddings. The scientific payload is byte-identical to v1.0.3: corpus, embeddings, events, labels, splits, and provenance are unchanged. Only two metadata members changed, to remove the depositor's account handle from the archived mirror reference in dataset-metadata.json and CITATION_LEDGER.json, both of which now reference this record. Attribution: Source: Swiss National Bank (snb.ch). Raw SNB HTML/PDF and model weights are not redistributed.