Rocky
167 lines
4.4 KiB
JSON
167 lines
4.4 KiB
JSON
{
|
|
"containsProjectUserClipboardData": false,
|
|
"corpusSHA256": "34387615f5721dc645edcc7c7feb2bd30b27c362a314aad6159fbc9e5baadf8c",
|
|
"coverage": {
|
|
"blessing": {
|
|
"en": {
|
|
"negative": 13107,
|
|
"positive": 69
|
|
},
|
|
"zh-Hans": {
|
|
"negative": 28002,
|
|
"positive": 17
|
|
}
|
|
},
|
|
"complaint": {
|
|
"en": {
|
|
"negative": 12742,
|
|
"positive": 434
|
|
},
|
|
"zh-Hans": {
|
|
"negative": 27749,
|
|
"positive": 270
|
|
}
|
|
},
|
|
"confirmationDecision": {
|
|
"en": {
|
|
"negative": 13164,
|
|
"positive": 12
|
|
},
|
|
"zh-Hans": {
|
|
"negative": 27962,
|
|
"positive": 57
|
|
}
|
|
},
|
|
"followUpReminder": {
|
|
"en": {
|
|
"negative": 13051,
|
|
"positive": 125
|
|
},
|
|
"zh-Hans": {
|
|
"negative": 27754,
|
|
"positive": 265
|
|
}
|
|
},
|
|
"invitation": {
|
|
"en": {
|
|
"negative": 13115,
|
|
"positive": 61
|
|
},
|
|
"zh-Hans": {
|
|
"negative": 27946,
|
|
"positive": 73
|
|
}
|
|
},
|
|
"question": {
|
|
"en": {
|
|
"negative": 10520,
|
|
"positive": 2656
|
|
},
|
|
"zh-Hans": {
|
|
"negative": 20835,
|
|
"positive": 7184
|
|
}
|
|
},
|
|
"replyable": {
|
|
"en": {
|
|
"negative": 2295,
|
|
"positive": 10881
|
|
},
|
|
"zh-Hans": {
|
|
"negative": 1593,
|
|
"positive": 26426
|
|
}
|
|
},
|
|
"scheduleNegotiation": {
|
|
"en": {
|
|
"negative": 13109,
|
|
"positive": 67
|
|
},
|
|
"zh-Hans": {
|
|
"negative": 28015,
|
|
"positive": 4
|
|
}
|
|
},
|
|
"task": {
|
|
"en": {
|
|
"negative": 12057,
|
|
"positive": 1119
|
|
},
|
|
"zh-Hans": {
|
|
"negative": 27360,
|
|
"positive": 659
|
|
}
|
|
}
|
|
},
|
|
"generatedFromOnlineSources": true,
|
|
"holdoutPolicy": "Frozen online evaluation only. These records are excluded from training, model selection, and threshold calibration.",
|
|
"labelCorrection": {
|
|
"reason": "An initial prelabel audit found obvious surface questions, invitations, schedule alternatives, reminders, and wishes that source intent schemas did not encode. A second audit removed quoted, sarcastic, and substring-only blessing matches. Deterministic label rules were corrected before the frozen evaluation.",
|
|
"supersededCorpusSHA256": [
|
|
"36f736f69e0c9ac42262d2b57640f9889534ffdc6e70da67fda6f41d7211b2f9",
|
|
"baffdfff4b5436810c45d92d28c5885d8ed475b878531220852ae6f2a55011e4"
|
|
]
|
|
},
|
|
"labelingMethod": "Original dataset labels are retained where available. Product intents not present in source schemas use conservative lexical mapping; metrics must therefore be read with per-source results and label-coverage counts.",
|
|
"profile": "comprehensive",
|
|
"recordCount": 41195,
|
|
"seed": 20260831,
|
|
"sentimentCoverage": {
|
|
"en": {
|
|
"negative": 1605,
|
|
"neutral": 9578,
|
|
"positive": 1993
|
|
},
|
|
"zh-Hans": {
|
|
"negative": 11934,
|
|
"neutral": 10191,
|
|
"positive": 5894
|
|
}
|
|
},
|
|
"sources": [
|
|
{
|
|
"dataset": "CFPB via hpe-ai/customer-complaints",
|
|
"license": "CC0-1.0 source / Apache-2.0 mirror",
|
|
"records": 250,
|
|
"url": "https://huggingface.co/datasets/hpe-ai/customer-complaints"
|
|
},
|
|
{
|
|
"dataset": "Meituan-Dianping/ASAP",
|
|
"license": "Apache-2.0",
|
|
"records": 120,
|
|
"revision": "975122a60065240124df62cb4d5dbfd19ed9ef2c",
|
|
"url": "https://github.com/Meituan-Dianping/asap"
|
|
},
|
|
{
|
|
"dataset": "AmazonScience/MASSIVE",
|
|
"license": "CC-BY-4.0",
|
|
"records": 5817,
|
|
"url": "https://huggingface.co/datasets/AmazonScience/massive"
|
|
},
|
|
{
|
|
"dataset": "Verified Enron Intent Dataset",
|
|
"license": "MIT / public Enron source corpus",
|
|
"records": 3537,
|
|
"url": "https://github.com/vseledkin/enron_intent_dataset_verified"
|
|
},
|
|
{
|
|
"dataset": "CPED",
|
|
"license": "Apache-2.0",
|
|
"records": 25049,
|
|
"url": "https://github.com/qftie/CPED"
|
|
},
|
|
{
|
|
"dataset": "Google Research GoEmotions",
|
|
"license": "Apache-2.0",
|
|
"records": 5418,
|
|
"url": "https://huggingface.co/datasets/google-research-datasets/go_emotions"
|
|
},
|
|
{
|
|
"dataset": "Sulak2020 GitHub issues multi-repo dataset",
|
|
"license": "CC-BY-4.0 with per-record attribution",
|
|
"records": 1004,
|
|
"url": "https://huggingface.co/datasets/Sulak2020/github-issues-multirepo-datasets-Sulakshana"
|
|
}
|
|
]
|
|
}
|