{ "access_conditions": "Public authoritative HTTPS retrieval without registration or individually signed agreement. Upstream use and attribution conditions remain applicable. This acquisition recipe grants no additional publication rights.", "checksum_provenance": "locally_observed_from_authoritative_https", "citation": "Carletta et al. (2005), The AMI Meeting Corpus: A Pre-announcement.", "license": "CC-BY-4.0", "metadata_url": "https://groups.inf.ed.ac.uk/ami/download/", "notes": "The author download chooser supplied these exact HTTPS URLs. Manual annotation ZIP CRC and all selected word timestamps were validated against the actual native session durations. The full annotation package remains provenance; native audio is limited to the three specified sessions.", "provenance_manifest": "docs/curation/speech-selections/bootstrap-provenance.json", "quality_exclusions": { "exclude": [ "missing or malformed required target with no recoverable authoritative annotation", "audio/annotation identifier mismatch or invalid time bounds", "author-withdrawn examples", "undecodable or truncated files" ], "retain": [ "background noise", "difficulty, accents, child speech or dysarthria severity", "low agreement when valid annotator distributions remain available", "overlap within recordings or across datasets; annotate overlap and group related examples" ], "task_specific": "Mask only invalid task targets when other targets remain valid; do not infer poor-label status from acoustic difficulty." }, "release": "manual annotations 1.6.2 and author Mix-Headset sessions; observed 2026-10-28", "selection": "Exactly ES2002a (custom train), IS1000a (custom validation), TS3003a (custom test), complete author-provided Mix-Headset WAV sessions. Their 12 global speaker IDs are pairwise disjoint and frozen in ami.json. Preserve source-native 16 kHz audio and the original mixture view; no local downmix or utterance clipping. This custom three-scenario subset is not the standard full AMI benchmark split.", "source_url": "https://groups.inf.ed.ac.uk/ami/download/", "subset_policy": { "downmix": false, "extraction": "TORGO: stream all compressed parent parts but materialize only selected speaker paths. Other sources: bounded safe ZIP extraction or direct native files. No concatenated archive or resampling.", "extraction_workspace_bytes": 344864249, "family_hard_ceiling_bytes": 50000000000, "full_parent_extraction": true, "manifest": "docs/curation/speech-selections/ami.json", "manifest_required": true, "native_audio": true, "pcm_bytes_limit": 139550378, "resample": false, "selected_native_audio_count": 3, "selection": "Exactly ES2002a (custom train), IS1000a (custom validation), TS3003a (custom test), complete author-provided Mix-Headset WAV sessions. Their 12 global speaker IDs are pairwise disjoint and frozen in ami.json. Preserve source-native 16 kHz audio and the original mixture view; no local downmix or utterance clipping. This custom three-scenario subset is not the standard full AMI benchmark split.", "stored_bytes_limit": 139550584, "validation_receipt": "docs/curation/speech-selections/ami-validation.json" } }