Sentiment_Analysis_3W0624 / reports /evaluation_comparison.json
Lavender825's picture
Deploy 3W 0624 BERT metadata fusion model
38136ed
Raw
History Blame Contribute Delete
18.6 kB
{
"overall_3class_comparison": {
"Baseline_1_TFIDF_LogReg": {
"macro_f1": 0.6065885032485002,
"accuracy": 0.8164444444444444
},
"Baseline_2_BERT_overall_3class": {
"macro_f1": 0.7179307477990319,
"accuracy": 0.9002222222222223
},
"Baseline_3_BERT_ACSA_no_meta__aggregated_to_overall": {
"macro_f1": 0.4793355643956521,
"accuracy": 0.6584444444444445
},
"Proposed_BERT_Meta_Fusion__overall_head": {
"macro_f1": 0.697132916417921,
"accuracy": 0.8935555555555555
},
"Proposed_BERT_Meta_Fusion__aggregated_to_overall_(reference)": {
"macro_f1": 0.5042876450582995,
"accuracy": 0.6951111111111111
}
},
"proposed_per_aspect": {
"per_aspect": {
"SIZE": {
"macro_f1": 0.7995265557571497,
"accuracy": 0.8084444444444444,
"confusion_matrix": [
[
1048,
138,
72
],
[
285,
1854,
254
],
[
41,
72,
736
]
],
"report": {
"Not_Mentioned": {
"precision": 0.7627365356622998,
"recall": 0.8330683624801272,
"f1-score": 0.7963525835866262,
"support": 1258.0
},
"Positive": {
"precision": 0.8982558139534884,
"recall": 0.7747597158378604,
"f1-score": 0.8319497419789096,
"support": 2393.0
},
"Negative": {
"precision": 0.6930320150659134,
"recall": 0.866902237926973,
"f1-score": 0.7702773417059131,
"support": 849.0
},
"accuracy": 0.8084444444444444,
"macro avg": {
"precision": 0.7846747882272339,
"recall": 0.8249101054149869,
"f1-score": 0.7995265557571497,
"support": 4500.0
},
"weighted avg": {
"precision": 0.8216517567655182,
"recall": 0.8084444444444444,
"f1-score": 0.810362832403517,
"support": 4500.0
}
}
},
"MATERIAL": {
"macro_f1": 0.7375608350812924,
"accuracy": 0.7748888888888888,
"confusion_matrix": [
[
885,
196,
99
],
[
307,
2119,
271
],
[
74,
66,
483
]
],
"report": {
"Not_Mentioned": {
"precision": 0.6990521327014217,
"recall": 0.75,
"f1-score": 0.723630417007359,
"support": 1180.0
},
"Positive": {
"precision": 0.8899622007559849,
"recall": 0.78568780126066,
"f1-score": 0.8345805435210712,
"support": 2697.0
},
"Negative": {
"precision": 0.5662368112543963,
"recall": 0.7752808988764045,
"f1-score": 0.6544715447154471,
"support": 623.0
},
"accuracy": 0.7748888888888888,
"macro avg": {
"precision": 0.7184170482372677,
"recall": 0.770322900045688,
"f1-score": 0.7375608350812924,
"support": 4500.0
},
"weighted avg": {
"precision": 0.7950833567640129,
"recall": 0.7748888888888888,
"f1-score": 0.7805518645117191,
"support": 4500.0
}
}
},
"QUALITY": {
"macro_f1": 0.7225683708431877,
"accuracy": 0.7782222222222223,
"confusion_matrix": [
[
1939,
419,
163
],
[
290,
1360,
59
],
[
42,
25,
203
]
],
"report": {
"Not_Mentioned": {
"precision": 0.8538088947600176,
"recall": 0.7691392304641016,
"f1-score": 0.8092654424040067,
"support": 2521.0
},
"Positive": {
"precision": 0.753880266075388,
"recall": 0.7957870099473376,
"f1-score": 0.7742670082550527,
"support": 1709.0
},
"Negative": {
"precision": 0.4776470588235294,
"recall": 0.7518518518518519,
"f1-score": 0.5841726618705037,
"support": 270.0
},
"accuracy": 0.7782222222222223,
"macro avg": {
"precision": 0.695112073219645,
"recall": 0.7722593640877636,
"f1-score": 0.7225683708431877,
"support": 4500.0
},
"weighted avg": {
"precision": 0.793288512065599,
"recall": 0.7782222222222223,
"f1-score": 0.7824682480252049,
"support": 4500.0
}
}
},
"APPEARANCE": {
"macro_f1": 0.686600260772051,
"accuracy": 0.8044444444444444,
"confusion_matrix": [
[
2098,
260,
145
],
[
273,
1404,
131
],
[
31,
40,
118
]
],
"report": {
"Not_Mentioned": {
"precision": 0.8734388009991674,
"recall": 0.8381941669996005,
"f1-score": 0.855453618756371,
"support": 2503.0
},
"Positive": {
"precision": 0.823943661971831,
"recall": 0.7765486725663717,
"f1-score": 0.7995444191343963,
"support": 1808.0
},
"Negative": {
"precision": 0.29949238578680204,
"recall": 0.6243386243386243,
"f1-score": 0.40480274442538594,
"support": 189.0
},
"accuracy": 0.8044444444444444,
"macro avg": {
"precision": 0.6656249495859335,
"recall": 0.7463604879681989,
"f1-score": 0.686600260772051,
"support": 4500.0
},
"weighted avg": {
"precision": 0.8294470045910427,
"recall": 0.8044444444444444,
"f1-score": 0.8140632080530185,
"support": 4500.0
}
}
},
"STYLE": {
"macro_f1": 0.7160335416761091,
"accuracy": 0.806,
"confusion_matrix": [
[
1880,
242,
118
],
[
290,
1585,
150
],
[
39,
34,
162
]
],
"report": {
"Not_Mentioned": {
"precision": 0.851063829787234,
"recall": 0.8392857142857143,
"f1-score": 0.8451337379186334,
"support": 2240.0
},
"Positive": {
"precision": 0.8516926383664697,
"recall": 0.782716049382716,
"f1-score": 0.815748841996912,
"support": 2025.0
},
"Negative": {
"precision": 0.3767441860465116,
"recall": 0.6893617021276596,
"f1-score": 0.48721804511278194,
"support": 235.0
},
"accuracy": 0.806,
"macro avg": {
"precision": 0.693166884733405,
"recall": 0.7704544885986966,
"f1-score": 0.7160335416761091,
"support": 4500.0
},
"weighted avg": {
"precision": 0.8265767678080969,
"recall": 0.806,
"f1-score": 0.813219381907331,
"support": 4500.0
}
}
},
"VALUE": {
"macro_f1": 0.6463455740801278,
"accuracy": 0.814,
"confusion_matrix": [
[
2944,
324,
172
],
[
231,
630,
46
],
[
35,
29,
89
]
],
"report": {
"Not_Mentioned": {
"precision": 0.9171339563862928,
"recall": 0.8558139534883721,
"f1-score": 0.8854135338345864,
"support": 3440.0
},
"Positive": {
"precision": 0.6408952187182095,
"recall": 0.6945975744211687,
"f1-score": 0.6666666666666666,
"support": 907.0
},
"Negative": {
"precision": 0.2899022801302932,
"recall": 0.5816993464052288,
"f1-score": 0.3869565217391304,
"support": 153.0
},
"accuracy": 0.814,
"macro avg": {
"precision": 0.6159771517449318,
"recall": 0.7107036247715898,
"f1-score": 0.6463455740801278,
"support": 4500.0
},
"weighted avg": {
"precision": 0.840130627156933,
"recall": 0.814,
"f1-score": 0.8243763490852736,
"support": 4500.0
}
}
}
},
"overall": {
"mean_macro_f1": 0.7181058563683197,
"mean_accuracy": 0.7976666666666666
}
},
"acsa_no_meta_per_aspect": {
"per_aspect": {
"SIZE": {
"macro_f1": 0.723496978308766,
"accuracy": 0.7164444444444444,
"confusion_matrix": [
[
1035,
152,
71
],
[
686,
1445,
262
],
[
49,
56,
744
]
],
"report": {
"Not_Mentioned": {
"precision": 0.5847457627118644,
"recall": 0.8227344992050875,
"f1-score": 0.6836195508586526,
"support": 1258.0
},
"Positive": {
"precision": 0.8741681790683605,
"recall": 0.6038445465942331,
"f1-score": 0.7142857142857143,
"support": 2393.0
},
"Negative": {
"precision": 0.6908077994428969,
"recall": 0.8763250883392226,
"f1-score": 0.7725856697819314,
"support": 849.0
},
"accuracy": 0.7164444444444444,
"macro avg": {
"precision": 0.7165739137410405,
"recall": 0.7676347113795144,
"f1-score": 0.723496978308766,
"support": 4500.0
},
"weighted avg": {
"precision": 0.7586645430509182,
"recall": 0.7164444444444444,
"f1-score": 0.7167120762023909,
"support": 4500.0
}
}
},
"MATERIAL": {
"macro_f1": 0.6869602136807388,
"accuracy": 0.7111111111111111,
"confusion_matrix": [
[
819,
273,
88
],
[
535,
1896,
266
],
[
72,
66,
485
]
],
"report": {
"Not_Mentioned": {
"precision": 0.5743338008415148,
"recall": 0.6940677966101695,
"f1-score": 0.6285495011511896,
"support": 1180.0
},
"Positive": {
"precision": 0.8483221476510067,
"recall": 0.7030033370411568,
"f1-score": 0.7688564476885644,
"support": 2697.0
},
"Negative": {
"precision": 0.5780691299165673,
"recall": 0.7784911717495987,
"f1-score": 0.6634746922024624,
"support": 623.0
},
"accuracy": 0.7111111111111111,
"macro avg": {
"precision": 0.6669083594696962,
"recall": 0.7251874351336417,
"f1-score": 0.6869602136807388,
"support": 4500.0
},
"weighted avg": {
"precision": 0.7390612855879498,
"recall": 0.7111111111111111,
"f1-score": 0.7174753297814658,
"support": 4500.0
}
}
},
"QUALITY": {
"macro_f1": 0.6921391557031549,
"accuracy": 0.7277777777777777,
"confusion_matrix": [
[
1961,
448,
112
],
[
536,
1112,
61
],
[
42,
26,
202
]
],
"report": {
"Not_Mentioned": {
"precision": 0.7723513194170933,
"recall": 0.777865926219754,
"f1-score": 0.775098814229249,
"support": 2521.0
},
"Positive": {
"precision": 0.7011349306431274,
"recall": 0.6506729081334114,
"f1-score": 0.6749620637329287,
"support": 1709.0
},
"Negative": {
"precision": 0.5386666666666666,
"recall": 0.7481481481481481,
"f1-score": 0.6263565891472869,
"support": 270.0
},
"accuracy": 0.7277777777777777,
"macro avg": {
"precision": 0.6707176389089625,
"recall": 0.7255623275004378,
"f1-score": 0.6921391557031549,
"support": 4500.0
},
"weighted avg": {
"precision": 0.7312838383821326,
"recall": 0.7277777777777777,
"f1-score": 0.7281445681469509,
"support": 4500.0
}
}
},
"APPEARANCE": {
"macro_f1": 0.6134243985372673,
"accuracy": 0.7042222222222222,
"confusion_matrix": [
[
1902,
472,
129
],
[
558,
1161,
89
],
[
45,
38,
106
]
],
"report": {
"Not_Mentioned": {
"precision": 0.7592814371257485,
"recall": 0.7598881342389133,
"f1-score": 0.7595846645367412,
"support": 2503.0
},
"Positive": {
"precision": 0.6947935368043088,
"recall": 0.6421460176991151,
"f1-score": 0.6674331704512791,
"support": 1808.0
},
"Negative": {
"precision": 0.3271604938271605,
"recall": 0.5608465608465608,
"f1-score": 0.41325536062378165,
"support": 189.0
},
"accuracy": 0.7042222222222222,
"macro avg": {
"precision": 0.5937451559190726,
"recall": 0.6542935709281964,
"f1-score": 0.6134243985372673,
"support": 4500.0
},
"weighted avg": {
"precision": 0.715222552222505,
"recall": 0.7042222222222222,
"f1-score": 0.7080144112598379,
"support": 4500.0
}
}
},
"STYLE": {
"macro_f1": 0.651480797158933,
"accuracy": 0.7168888888888889,
"confusion_matrix": [
[
1722,
406,
112
],
[
563,
1356,
106
],
[
47,
40,
148
]
],
"report": {
"Not_Mentioned": {
"precision": 0.7384219554030875,
"recall": 0.76875,
"f1-score": 0.7532808398950132,
"support": 2240.0
},
"Positive": {
"precision": 0.7524972253052165,
"recall": 0.6696296296296296,
"f1-score": 0.7086490723804547,
"support": 2025.0
},
"Negative": {
"precision": 0.40437158469945356,
"recall": 0.6297872340425532,
"f1-score": 0.4925124792013311,
"support": 235.0
},
"accuracy": 0.7168888888888889,
"macro avg": {
"precision": 0.6317635884692525,
"recall": 0.6893889545573942,
"f1-score": 0.651480797158933,
"support": 4500.0
},
"weighted avg": {
"precision": 0.7273109741667446,
"recall": 0.7168888888888889,
"f1-score": 0.7195786412327918,
"support": 4500.0
}
}
},
"VALUE": {
"macro_f1": 0.6489411814960196,
"accuracy": 0.8326666666666667,
"confusion_matrix": [
[
3129,
155,
156
],
[
331,
534,
42
],
[
46,
23,
84
]
],
"report": {
"Not_Mentioned": {
"precision": 0.892470051340559,
"recall": 0.9095930232558139,
"f1-score": 0.9009501871580766,
"support": 3440.0
},
"Positive": {
"precision": 0.75,
"recall": 0.5887541345093715,
"f1-score": 0.6596664607782582,
"support": 907.0
},
"Negative": {
"precision": 0.2978723404255319,
"recall": 0.5490196078431373,
"f1-score": 0.38620689655172413,
"support": 153.0
},
"accuracy": 0.8326666666666667,
"macro avg": {
"precision": 0.6467807972553636,
"recall": 0.6824555885361075,
"f1-score": 0.6489411814960196,
"support": 4500.0
},
"weighted avg": {
"precision": 0.8435380988214732,
"recall": 0.8326666666666667,
"f1-score": 0.8348168397604616,
"support": 4500.0
}
}
}
},
"overall": {
"mean_macro_f1": 0.6694071208141467,
"mean_accuracy": 0.7348518518518518
}
},
"note": "The Proposed model jointly trains per-aspect heads and an overall sentiment head on the shared fused representation. The overall_head result is the primary overall metric. The aggregated_to_overall result (voting from per-aspect predictions) is included for reference. Baseline 3 vs Proposed isolates the marginal value of metadata cross-attention fusion."
}