docs(semantics): record iterative retraining study
Co-authored-by: Rocky <hkgood@users.noreply.github.com>
This commit is contained in:
@@ -0,0 +1,481 @@
|
||||
{
|
||||
"automaticPromotionAllowed": false,
|
||||
"baseCorpus": "ModelTraining/ClipboardSemantics/clipboard_semantic_corpus.jsonl",
|
||||
"configuration": {
|
||||
"alpha": 2e-05,
|
||||
"augmentation": "none",
|
||||
"char_max": 5,
|
||||
"char_min": 2,
|
||||
"external_weight": 0.5700000000000001,
|
||||
"hard_example_weight": 1.0,
|
||||
"l1_ratio": 0.0,
|
||||
"max_char_features": 45000,
|
||||
"max_word_features": 22000,
|
||||
"min_df": 2,
|
||||
"pseudo_weight": 0.16499999999999998,
|
||||
"round": 1,
|
||||
"word_max": 2
|
||||
},
|
||||
"openCorpus": "ModelTraining/ClipboardSemantics/open-training-corpus.jsonl",
|
||||
"reason": "Surrogate feature weights are not deployable NLModel assets and cannot bypass the existing macOS acceptance policy.",
|
||||
"requiredCommands": [
|
||||
"python3 Scripts/clipboard_semantics/generate_open_training_corpus.py",
|
||||
"xcrun swift Scripts/clipboard_semantics/train_models.swift --algorithms maxEnt --corpus ModelTraining/ClipboardSemantics/combined-training-corpus.jsonl",
|
||||
"xcrun swift Scripts/clipboard_semantics/evaluate_random_holdout.swift --corpus ModelTraining/ClipboardSemantics/random-holdout-corpus.jsonl"
|
||||
],
|
||||
"schemaVersion": 1,
|
||||
"seed": 20260827,
|
||||
"selectedRound": 1,
|
||||
"thresholds": {
|
||||
"blessing": {
|
||||
"abstained": false,
|
||||
"byLanguage": {
|
||||
"en": {
|
||||
"abstained": false,
|
||||
"metrics": {
|
||||
"f1": 0.9318,
|
||||
"falseNegative": 4,
|
||||
"falsePositive": 2,
|
||||
"precision": 0.9535,
|
||||
"predictedPositive": 43,
|
||||
"recall": 0.9111,
|
||||
"total": 1485,
|
||||
"trueNegative": 1438,
|
||||
"truePositive": 41,
|
||||
"wilsonPrecisionLower95": 0.8454
|
||||
},
|
||||
"threshold": 0.915
|
||||
},
|
||||
"zh-Hans": {
|
||||
"abstained": false,
|
||||
"metrics": {
|
||||
"f1": 0.7568,
|
||||
"falseNegative": 17,
|
||||
"falsePositive": 1,
|
||||
"precision": 0.9655,
|
||||
"predictedPositive": 29,
|
||||
"recall": 0.6222,
|
||||
"total": 1485,
|
||||
"trueNegative": 1439,
|
||||
"truePositive": 28,
|
||||
"wilsonPrecisionLower95": 0.8282
|
||||
},
|
||||
"threshold": 0.9
|
||||
}
|
||||
},
|
||||
"metrics": {
|
||||
"f1": 0.8375,
|
||||
"falseNegative": 23,
|
||||
"falsePositive": 3,
|
||||
"precision": 0.9571,
|
||||
"predictedPositive": 70,
|
||||
"recall": 0.7444,
|
||||
"total": 2970,
|
||||
"trueNegative": 2877,
|
||||
"truePositive": 67,
|
||||
"wilsonPrecisionLower95": 0.8814
|
||||
},
|
||||
"threshold": 0.915
|
||||
},
|
||||
"complaint": {
|
||||
"abstained": false,
|
||||
"byLanguage": {
|
||||
"en": {
|
||||
"abstained": false,
|
||||
"metrics": {
|
||||
"f1": 1.0,
|
||||
"falseNegative": 0,
|
||||
"falsePositive": 0,
|
||||
"precision": 1.0,
|
||||
"predictedPositive": 180,
|
||||
"recall": 1.0,
|
||||
"total": 1485,
|
||||
"trueNegative": 1305,
|
||||
"truePositive": 180,
|
||||
"wilsonPrecisionLower95": 0.9791
|
||||
},
|
||||
"threshold": 0.5
|
||||
},
|
||||
"zh-Hans": {
|
||||
"abstained": false,
|
||||
"metrics": {
|
||||
"f1": 0.9526,
|
||||
"falseNegative": 9,
|
||||
"falsePositive": 8,
|
||||
"precision": 0.9553,
|
||||
"predictedPositive": 179,
|
||||
"recall": 0.95,
|
||||
"total": 1485,
|
||||
"trueNegative": 1297,
|
||||
"truePositive": 171,
|
||||
"wilsonPrecisionLower95": 0.9143
|
||||
},
|
||||
"threshold": 0.54
|
||||
}
|
||||
},
|
||||
"metrics": {
|
||||
"f1": 0.9764,
|
||||
"falseNegative": 9,
|
||||
"falsePositive": 8,
|
||||
"precision": 0.9777,
|
||||
"predictedPositive": 359,
|
||||
"recall": 0.975,
|
||||
"total": 2970,
|
||||
"trueNegative": 2602,
|
||||
"truePositive": 351,
|
||||
"wilsonPrecisionLower95": 0.9567
|
||||
},
|
||||
"threshold": 0.54
|
||||
},
|
||||
"confirmationDecision": {
|
||||
"abstained": false,
|
||||
"byLanguage": {
|
||||
"en": {
|
||||
"abstained": false,
|
||||
"metrics": {
|
||||
"f1": 0.8848,
|
||||
"falseNegative": 17,
|
||||
"falsePositive": 2,
|
||||
"precision": 0.9733,
|
||||
"predictedPositive": 75,
|
||||
"recall": 0.8111,
|
||||
"total": 1485,
|
||||
"trueNegative": 1393,
|
||||
"truePositive": 73,
|
||||
"wilsonPrecisionLower95": 0.9079
|
||||
},
|
||||
"threshold": 0.5
|
||||
},
|
||||
"zh-Hans": {
|
||||
"abstained": false,
|
||||
"metrics": {
|
||||
"f1": 0.7947,
|
||||
"falseNegative": 30,
|
||||
"falsePositive": 1,
|
||||
"precision": 0.9836,
|
||||
"predictedPositive": 61,
|
||||
"recall": 0.6667,
|
||||
"total": 1485,
|
||||
"trueNegative": 1394,
|
||||
"truePositive": 60,
|
||||
"wilsonPrecisionLower95": 0.9128
|
||||
},
|
||||
"threshold": 0.5
|
||||
}
|
||||
},
|
||||
"metrics": {
|
||||
"f1": 0.8418,
|
||||
"falseNegative": 47,
|
||||
"falsePositive": 3,
|
||||
"precision": 0.9779,
|
||||
"predictedPositive": 136,
|
||||
"recall": 0.7389,
|
||||
"total": 2970,
|
||||
"trueNegative": 2787,
|
||||
"truePositive": 133,
|
||||
"wilsonPrecisionLower95": 0.9372
|
||||
},
|
||||
"threshold": 0.5
|
||||
},
|
||||
"followUpReminder": {
|
||||
"abstained": false,
|
||||
"byLanguage": {
|
||||
"en": {
|
||||
"abstained": false,
|
||||
"metrics": {
|
||||
"f1": 0.8554,
|
||||
"falseNegative": 41,
|
||||
"falsePositive": 6,
|
||||
"precision": 0.9586,
|
||||
"predictedPositive": 145,
|
||||
"recall": 0.7722,
|
||||
"total": 1485,
|
||||
"trueNegative": 1299,
|
||||
"truePositive": 139,
|
||||
"wilsonPrecisionLower95": 0.9127
|
||||
},
|
||||
"threshold": 0.83
|
||||
},
|
||||
"zh-Hans": {
|
||||
"abstained": false,
|
||||
"metrics": {
|
||||
"f1": 0.8354,
|
||||
"falseNegative": 48,
|
||||
"falsePositive": 4,
|
||||
"precision": 0.9706,
|
||||
"predictedPositive": 136,
|
||||
"recall": 0.7333,
|
||||
"total": 1485,
|
||||
"trueNegative": 1301,
|
||||
"truePositive": 132,
|
||||
"wilsonPrecisionLower95": 0.9268
|
||||
},
|
||||
"threshold": 0.505
|
||||
}
|
||||
},
|
||||
"metrics": {
|
||||
"f1": 0.7994,
|
||||
"falseNegative": 113,
|
||||
"falsePositive": 11,
|
||||
"precision": 0.9574,
|
||||
"predictedPositive": 258,
|
||||
"recall": 0.6861,
|
||||
"total": 2970,
|
||||
"trueNegative": 2599,
|
||||
"truePositive": 247,
|
||||
"wilsonPrecisionLower95": 0.9253
|
||||
},
|
||||
"threshold": 0.815
|
||||
},
|
||||
"invitation": {
|
||||
"abstained": false,
|
||||
"byLanguage": {
|
||||
"en": {
|
||||
"abstained": false,
|
||||
"metrics": {
|
||||
"f1": 0.8302,
|
||||
"falseNegative": 24,
|
||||
"falsePositive": 3,
|
||||
"precision": 0.9565,
|
||||
"predictedPositive": 69,
|
||||
"recall": 0.7333,
|
||||
"total": 1485,
|
||||
"trueNegative": 1392,
|
||||
"truePositive": 66,
|
||||
"wilsonPrecisionLower95": 0.8798
|
||||
},
|
||||
"threshold": 0.935
|
||||
},
|
||||
"zh-Hans": {
|
||||
"abstained": false,
|
||||
"metrics": {
|
||||
"f1": 0.6765,
|
||||
"falseNegative": 44,
|
||||
"falsePositive": 0,
|
||||
"precision": 1.0,
|
||||
"predictedPositive": 46,
|
||||
"recall": 0.5111,
|
||||
"total": 1485,
|
||||
"trueNegative": 1395,
|
||||
"truePositive": 46,
|
||||
"wilsonPrecisionLower95": 0.9229
|
||||
},
|
||||
"threshold": 0.505
|
||||
}
|
||||
},
|
||||
"metrics": {
|
||||
"f1": 0.5476,
|
||||
"falseNegative": 111,
|
||||
"falsePositive": 3,
|
||||
"precision": 0.9583,
|
||||
"predictedPositive": 72,
|
||||
"recall": 0.3833,
|
||||
"total": 2970,
|
||||
"trueNegative": 2787,
|
||||
"truePositive": 69,
|
||||
"wilsonPrecisionLower95": 0.8845
|
||||
},
|
||||
"threshold": 0.935
|
||||
},
|
||||
"question": {
|
||||
"abstained": false,
|
||||
"byLanguage": {
|
||||
"en": {
|
||||
"abstained": false,
|
||||
"metrics": {
|
||||
"f1": 0.8445,
|
||||
"falseNegative": 76,
|
||||
"falsePositive": 12,
|
||||
"precision": 0.9522,
|
||||
"predictedPositive": 251,
|
||||
"recall": 0.7587,
|
||||
"total": 1485,
|
||||
"trueNegative": 1158,
|
||||
"truePositive": 239,
|
||||
"wilsonPrecisionLower95": 0.9183
|
||||
},
|
||||
"threshold": 0.725
|
||||
},
|
||||
"zh-Hans": {
|
||||
"abstained": false,
|
||||
"metrics": {
|
||||
"f1": 0.8536,
|
||||
"falseNegative": 76,
|
||||
"falsePositive": 6,
|
||||
"precision": 0.9755,
|
||||
"predictedPositive": 245,
|
||||
"recall": 0.7587,
|
||||
"total": 1485,
|
||||
"trueNegative": 1164,
|
||||
"truePositive": 239,
|
||||
"wilsonPrecisionLower95": 0.9476
|
||||
},
|
||||
"threshold": 0.5
|
||||
}
|
||||
},
|
||||
"metrics": {
|
||||
"f1": 0.8729,
|
||||
"falseNegative": 122,
|
||||
"falsePositive": 26,
|
||||
"precision": 0.9513,
|
||||
"predictedPositive": 534,
|
||||
"recall": 0.8063,
|
||||
"total": 2970,
|
||||
"trueNegative": 2314,
|
||||
"truePositive": 508,
|
||||
"wilsonPrecisionLower95": 0.9296
|
||||
},
|
||||
"threshold": 0.515
|
||||
},
|
||||
"replyableMessage": {
|
||||
"abstained": false,
|
||||
"byLanguage": {
|
||||
"en": {
|
||||
"abstained": false,
|
||||
"metrics": {
|
||||
"f1": 0.948,
|
||||
"falseNegative": 52,
|
||||
"falsePositive": 46,
|
||||
"precision": 0.951,
|
||||
"predictedPositive": 939,
|
||||
"recall": 0.945,
|
||||
"total": 1485,
|
||||
"trueNegative": 494,
|
||||
"truePositive": 893,
|
||||
"wilsonPrecisionLower95": 0.9353
|
||||
},
|
||||
"threshold": 0.7
|
||||
},
|
||||
"zh-Hans": {
|
||||
"abstained": false,
|
||||
"metrics": {
|
||||
"f1": 0.9424,
|
||||
"falseNegative": 62,
|
||||
"falsePositive": 46,
|
||||
"precision": 0.9505,
|
||||
"predictedPositive": 929,
|
||||
"recall": 0.9344,
|
||||
"total": 1485,
|
||||
"trueNegative": 494,
|
||||
"truePositive": 883,
|
||||
"wilsonPrecisionLower95": 0.9346
|
||||
},
|
||||
"threshold": 0.695
|
||||
}
|
||||
},
|
||||
"metrics": {
|
||||
"f1": 0.9454,
|
||||
"falseNegative": 115,
|
||||
"falsePositive": 90,
|
||||
"precision": 0.9517,
|
||||
"predictedPositive": 1865,
|
||||
"recall": 0.9392,
|
||||
"total": 2970,
|
||||
"trueNegative": 990,
|
||||
"truePositive": 1775,
|
||||
"wilsonPrecisionLower95": 0.9411
|
||||
},
|
||||
"threshold": 0.7
|
||||
},
|
||||
"scheduleNegotiation": {
|
||||
"abstained": false,
|
||||
"byLanguage": {
|
||||
"en": {
|
||||
"abstained": false,
|
||||
"metrics": {
|
||||
"f1": 1.0,
|
||||
"falseNegative": 0,
|
||||
"falsePositive": 0,
|
||||
"precision": 1.0,
|
||||
"predictedPositive": 45,
|
||||
"recall": 1.0,
|
||||
"total": 1485,
|
||||
"trueNegative": 1440,
|
||||
"truePositive": 45,
|
||||
"wilsonPrecisionLower95": 0.9213
|
||||
},
|
||||
"threshold": 0.5
|
||||
},
|
||||
"zh-Hans": {
|
||||
"abstained": false,
|
||||
"metrics": {
|
||||
"f1": 0.967,
|
||||
"falseNegative": 1,
|
||||
"falsePositive": 2,
|
||||
"precision": 0.9565,
|
||||
"predictedPositive": 46,
|
||||
"recall": 0.9778,
|
||||
"total": 1485,
|
||||
"trueNegative": 1438,
|
||||
"truePositive": 44,
|
||||
"wilsonPrecisionLower95": 0.8547
|
||||
},
|
||||
"threshold": 0.93
|
||||
}
|
||||
},
|
||||
"metrics": {
|
||||
"f1": 0.978,
|
||||
"falseNegative": 1,
|
||||
"falsePositive": 3,
|
||||
"precision": 0.9674,
|
||||
"predictedPositive": 92,
|
||||
"recall": 0.9889,
|
||||
"total": 2970,
|
||||
"trueNegative": 2877,
|
||||
"truePositive": 89,
|
||||
"wilsonPrecisionLower95": 0.9085
|
||||
},
|
||||
"threshold": 0.91
|
||||
},
|
||||
"task": {
|
||||
"abstained": false,
|
||||
"byLanguage": {
|
||||
"en": {
|
||||
"abstained": false,
|
||||
"metrics": {
|
||||
"f1": 0.8403,
|
||||
"falseNegative": 89,
|
||||
"falsePositive": 14,
|
||||
"precision": 0.9509,
|
||||
"predictedPositive": 285,
|
||||
"recall": 0.7528,
|
||||
"total": 1485,
|
||||
"trueNegative": 1111,
|
||||
"truePositive": 271,
|
||||
"wilsonPrecisionLower95": 0.9192
|
||||
},
|
||||
"threshold": 0.835
|
||||
},
|
||||
"zh-Hans": {
|
||||
"abstained": false,
|
||||
"metrics": {
|
||||
"f1": 0.8278,
|
||||
"falseNegative": 98,
|
||||
"falsePositive": 11,
|
||||
"precision": 0.9597,
|
||||
"predictedPositive": 273,
|
||||
"recall": 0.7278,
|
||||
"total": 1485,
|
||||
"trueNegative": 1114,
|
||||
"truePositive": 262,
|
||||
"wilsonPrecisionLower95": 0.9293
|
||||
},
|
||||
"threshold": 0.745
|
||||
}
|
||||
},
|
||||
"metrics": {
|
||||
"f1": 0.8637,
|
||||
"falseNegative": 153,
|
||||
"falsePositive": 26,
|
||||
"precision": 0.9562,
|
||||
"predictedPositive": 593,
|
||||
"recall": 0.7875,
|
||||
"total": 2970,
|
||||
"trueNegative": 2224,
|
||||
"truePositive": 567,
|
||||
"wilsonPrecisionLower95": 0.9365
|
||||
},
|
||||
"threshold": 0.755
|
||||
}
|
||||
}
|
||||
}
|
||||
Reference in New Issue
Block a user