humaneness-ears-base-medium / gemini_train_statistics.json
ChristophSchuhmann's picture
Release best Gemini-tuned Whisper Base and Small with code, normalization and evaluation
cd9b2d8 verified
Raw History Blame Contribute Delete
17.2 kB
{
"basis": "current training partition only; provisional until final source/annotation completion",
"score_names": [
"emo_Affection",
"emo_Amusement",
"emo_Anger",
"emo_Astonishment_Surprise",
"emo_Awe",
"emo_Bitterness",
"emo_Concentration",
"emo_Confusion",
"emo_Contemplation",
"emo_Contempt",
"emo_Contentment",
"emo_Disappointment",
"emo_Disgust",
"emo_Distress",
"emo_Doubt",
"emo_Elation",
"emo_Embarrassment",
"emo_Emotional_Numbness",
"emo_Fatigue_Exhaustion",
"emo_Fear",
"emo_Helplessness",
"emo_Hope_Enthusiasm_Optimism",
"emo_Impatience_and_Irritability",
"emo_Infatuation",
"emo_Interest",
"emo_Intoxication_Altered_States_of_Consciousness",
"emo_Jealousy_and_Envy",
"emo_Longing",
"emo_Malevolence_Malice",
"emo_Pain",
"emo_Pleasure_Ecstasy",
"emo_Pride",
"emo_Relief",
"emo_Sadness",
"emo_Sexual_Lust",
"emo_Shame",
"emo_Sourness",
"emo_Teasing",
"emo_Thankfulness_Gratitude",
"emo_Triumph",
"vn_AGEV_reg",
"vn_AROU_reg",
"vn_ARSH_reg",
"vn_ATCK_reg",
"vn_BKGN_reg",
"vn_BRGT_reg",
"vn_CHNK_reg",
"vn_CLRT_reg",
"vn_COGL_reg",
"vn_DARC_reg",
"vn_DFLU_reg",
"vn_EMPH_reg",
"vn_ESTH_reg",
"vn_EXPL_reg",
"vn_FOCS_reg",
"vn_FULL_reg",
"vn_GEND_reg",
"vn_HARM_reg",
"vn_METL_reg",
"vn_RANG_reg",
"vn_RCQL_reg",
"vn_REGS_reg",
"vn_RESP_reg",
"vn_ROUG_reg",
"vn_R_CHST_reg",
"vn_R_HEAD_reg",
"vn_R_MASK_reg",
"vn_R_MIXD_reg",
"vn_R_NASL_reg",
"vn_R_ORAL_reg",
"vn_R_THRT_reg",
"vn_SMTH_reg",
"vn_STNC_reg",
"vn_STRU_reg",
"vn_S_ASMR_reg",
"vn_S_AUTH_reg",
"vn_S_CART_reg",
"vn_S_CASU_reg",
"vn_S_CONV_reg",
"vn_S_DRAM_reg",
"vn_S_FORM_reg",
"vn_S_MONO_reg",
"vn_S_NARR_reg",
"vn_S_NEWS_reg",
"vn_S_PLAY_reg",
"vn_S_RANT_reg",
"vn_S_STRY_reg",
"vn_S_TECH_reg",
"vn_S_WHIS_reg",
"vn_TEMP_reg",
"vn_TENS_reg",
"vn_VALN_reg",
"vn_VALS_reg",
"vn_VFLX_reg",
"vn_VOLT_reg",
"vn_VULN_reg",
"vn_WARM_reg",
"genuineness_0_6",
"blend_0_10",
"R_quality",
"eiv_extra:Age",
"eiv_extra:Arousal",
"eiv_extra:Authenticity",
"eiv_extra:Background_Noise",
"eiv_extra:Confident_vs._Hesitant",
"eiv_extra:Gender",
"eiv_extra:High-Pitched_vs._Low-Pitched",
"eiv_extra:Monotone_vs._Expressive",
"eiv_extra:Recording_Quality",
"eiv_extra:Serious_vs._Humorous",
"eiv_extra:Soft_vs._Harsh",
"eiv_extra:Submissive_vs._Dominant",
"eiv_extra:Valence",
"eiv_extra:Vulnerable_vs._Emotionally_Detached",
"eiv_extra:Warm_vs._Cold",
"eiv_extra:score_background_quality",
"eiv_extra:score_content_enjoyment",
"eiv_extra:score_overall_quality",
"eiv_extra:score_speech_quality",
"audiobox:CE",
"audiobox:CU",
"audiobox:PC",
"audiobox:PQ",
"dnsmos:SIG_raw",
"dnsmos:BAK_raw",
"dnsmos:OVRL_raw",
"dnsmos:SIG",
"dnsmos:BAK",
"dnsmos:OVRL",
"dnsmos:P808_MOS",
"burst_count_log1p",
"voiceclap_attribute:Affection",
"voiceclap_attribute:Age",
"voiceclap_attribute:Amusement",
"voiceclap_attribute:Anger",
"voiceclap_attribute:Arousal",
"voiceclap_attribute:Astonishment_Surprise",
"voiceclap_attribute:Authenticity",
"voiceclap_attribute:Awe",
"voiceclap_attribute:Background_Noise",
"voiceclap_attribute:Bitterness",
"voiceclap_attribute:Concentration",
"voiceclap_attribute:Confident_vs._Hesitant",
"voiceclap_attribute:Confusion",
"voiceclap_attribute:Contemplation",
"voiceclap_attribute:Contempt",
"voiceclap_attribute:Contentment",
"voiceclap_attribute:Disappointment",
"voiceclap_attribute:Disgust",
"voiceclap_attribute:Distress",
"voiceclap_attribute:Doubt",
"voiceclap_attribute:Elation",
"voiceclap_attribute:Embarrassment",
"voiceclap_attribute:Emotional_Numbness",
"voiceclap_attribute:Fatigue_Exhaustion",
"voiceclap_attribute:Fear",
"voiceclap_attribute:Gender",
"voiceclap_attribute:Helplessness",
"voiceclap_attribute:High-Pitched_vs._Low-Pitched",
"voiceclap_attribute:Hope_Enthusiasm_Optimism",
"voiceclap_attribute:Impatience_and_Irritability",
"voiceclap_attribute:Infatuation",
"voiceclap_attribute:Interest",
"voiceclap_attribute:Intoxication_Altered_States_of_Consciousness",
"voiceclap_attribute:Jealousy_&_Envy",
"voiceclap_attribute:Longing",
"voiceclap_attribute:Malevolence_Malice",
"voiceclap_attribute:Monotone_vs._Expressive",
"voiceclap_attribute:Pain",
"voiceclap_attribute:Pleasure_Ecstasy",
"voiceclap_attribute:Pride",
"voiceclap_attribute:Recording_Quality",
"voiceclap_attribute:Relief",
"voiceclap_attribute:Sadness",
"voiceclap_attribute:Serious_vs._Humorous",
"voiceclap_attribute:Sexual_Lust",
"voiceclap_attribute:Shame",
"voiceclap_attribute:Soft_vs._Harsh",
"voiceclap_attribute:Sourness",
"voiceclap_attribute:Submissive_vs._Dominant",
"voiceclap_attribute:Teasing",
"voiceclap_attribute:Thankfulness_Gratitude",
"voiceclap_attribute:Triumph",
"voiceclap_attribute:Valence",
"voiceclap_attribute:Vulnerable_vs._Emotionally_Detached",
"voiceclap_attribute:Warm_vs._Cold",
"voiceclap_attribute:duration",
"voiceclap_attribute:score_background_quality",
"voiceclap_attribute:score_content_enjoyment",
"voiceclap_attribute:score_overall_quality",
"voiceclap_attribute:score_speech_quality",
"voiceclap_attribute:talking_speed"
],
"count": [
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
39001,
38159,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380,
58380
],
"mean": [
0.5192874272010963,
0.8194929770469395,
0.3966255566974979,
0.4953579993148342,
0.44897225077081243,
0.46447413497773066,
1.6810037684138335,
0.41133949982870355,
1.679496402877696,
0.3649708804385086,
1.2492977046933986,
0.6139773895169456,
0.24770469338814685,
0.7459061322370573,
0.6685337444330276,
0.45397396368619514,
0.15202124015073717,
0.19498115793079762,
0.6957519698526845,
0.2684994861253824,
0.636365193559424,
0.9964542651593041,
0.5949811579308014,
0.12663583418978872,
1.8540253511476465,
0.07341555327166721,
0.04021925316889366,
0.41702637889687927,
0.14842411784857978,
0.3755738266529632,
0.32266187050359474,
0.6275094210345996,
0.47195957519698184,
0.47295306611853016,
0.08124357656731661,
0.0987667009249734,
0.44242891401164874,
0.5102261048304335,
0.2551216169921222,
0.35068516615279116,
3.1353374443302626,
2.7525865022267553,
2.9202295306611767,
2.923723878040449,
3.6355943816375405,
3.2135662898252586,
2.5344295991778205,
3.700702295306596,
1.7232271325796513,
2.998629667694438,
1.3346008907160092,
3.3840698869475836,
3.6343953408701886,
0.06791709489551243,
3.8782631038026807,
3.2260020554984594,
3.433453237410097,
3.2719253168893876,
1.664354230901004,
2.9878725590955955,
3.79321685508737,
2.2750256937307367,
2.4556354916067002,
2.0331963001027913,
2.978639945186707,
2.220897567660148,
2.792069201781382,
3.365741692360377,
1.7864850976361832,
3.414234326824244,
2.753151764302858,
3.1715998629667737,
3.1539910928399952,
3.529427886262416,
0.723466940733134,
2.358667351832805,
0.7155704008222022,
3.024734498115809,
3.2268242548818384,
2.6172147995889055,
1.706543336759175,
2.593114080164422,
2.298681055155869,
0.6772524837273053,
1.5894484412469976,
0.7427714970880375,
2.112949640287793,
1.4278691332648055,
0.5706235011990417,
2.762778348749599,
2.4926515930112942,
2.878143199725919,
2.8546762589928028,
2.9047276464542575,
1.950736553614257,
2.440733127783501,
3.232168550873606,
4.376892771497031,
7.572395579600515,
0.7693877485195274,
3.1276252688748736,
1.7975117851024482,
2.9206331998079174,
0.4244207237190548,
1.2638457512362304,
0.18546744358696285,
1.6562414840420085,
2.012678580510897,
2.427450719181729,
1.0305236693673285,
0.40552226346753134,
0.39699793928476956,
0.39619846046379814,
1.3613356041653706,
0.23008181889474008,
3.499185580987178,
4.65496650034127,
2.802598769030134,
1.8963655511701567,
5.467054878725572,
6.219847980472767,
1.78165686211031,
7.089394044193112,
3.8498586445700322,
3.8741952291880803,
3.464723351832813,
3.4255701135662737,
3.7690470659472406,
3.0630566831106463,
3.6434493018156706,
0.6911627115926797,
0.5192874272010963,
3.046046147886795,
0.8194929770469395,
0.3966255566974979,
1.8033007885401788,
0.4953579993148342,
2.9673965281637833,
0.44897225077081243,
0.42774576940390335,
0.46447413497773066,
1.6810037684138335,
1.461671296478791,
0.41133949982870355,
1.679496402877696,
0.3649708804385086,
1.2492977046933986,
0.6139773895169456,
0.24770469338814685,
0.7459061322370573,
0.6685337444330276,
0.45397396368619514,
0.15202124015073717,
0.19498115793079762,
0.6957519698526845,
0.2684994861253824,
0.1481757374138329,
0.636365193559424,
1.6936685263957523,
0.9964542651593041,
0.5949811579308014,
0.12663583418978872,
1.8540253511476465,
0.07341555327166721,
0.04021925316889366,
0.41702637889687927,
0.14842411784857978,
2.073102347687031,
0.3755738266529632,
0.32266187050359474,
0.6275094210345996,
2.410305775592271,
0.47195957519698184,
0.47295306611853016,
1.0453307331945967,
0.08124357656731661,
0.0987667009249734,
0.3493214684057118,
0.44242891401164874,
0.5024435844148578,
0.5102261048304335,
0.2551216169921222,
0.35068516615279116,
-0.07671741054314579,
1.3852912880492998,
0.17436333947897234,
10.24294253519524,
3.4360303790571716,
4.73948690762626,
2.7390484171397644,
1.9350600422551738,
12.9722604064592
],
"std": [
0.9275509934462636,
1.1199536967386903,
0.8876714039103254,
0.8777417965953798,
0.9136679438380443,
0.9352020825374199,
0.8532260804844294,
0.7958664757742199,
1.034469784685504,
0.8991313749213092,
1.089972052452089,
1.0450669978420806,
0.7033534711122825,
1.1993687071727375,
0.8501536093242995,
0.9062922673554012,
0.5432606109052269,
0.49567380682192186,
1.0374120771862074,
0.7688286378586915,
1.1139542563326688,
1.0143224595515967,
1.0539923080983777,
0.5552511981769991,
0.9988860191471728,
0.3784024798686873,
0.346310391382999,
0.856239575514884,
0.5909113077863566,
0.8990990149350612,
0.7478989451961275,
0.8815397693916925,
0.9090830325094228,
0.9754637418540082,
0.46452910060514957,
0.5177313545053387,
0.8777784993195329,
0.9446382746562131,
0.7235127405869907,
0.8254622029290207,
0.48815533340997097,
1.0081745316317827,
0.704309507894119,
0.8446813065137221,
0.6147612508910077,
0.6663401328416759,
0.8225350034427309,
0.9374344440333365,
1.1056699615442602,
0.7156055511146882,
1.3066613675473895,
0.7905356681376416,
0.805157813783411,
0.29322919836517575,
0.7750588820333559,
0.5560118485616042,
1.8624343284362765,
0.8450345220845084,
0.765521373210316,
0.9544024833565046,
0.6733526568778027,
1.3526406201073573,
1.0494463859970398,
1.0036731577791953,
0.9442344473340155,
1.0279521359597226,
0.7615382644820817,
0.6375629584173483,
0.591413041607472,
0.5668344764084486,
0.780299461968861,
0.877558048809578,
1.1182991605989123,
0.9957811218714696,
1.1320008865029352,
1.2914777469284786,
1.1641413954144584,
1.461497888182547,
1.3664380770622713,
1.4570292409850658,
1.2450690715368136,
1.4149330334727896,
1.4064215427935223,
1.0304609645985237,
1.4531504105872122,
1.1641866625818793,
1.2333107185130365,
1.36531482276621,
1.1202901433185866,
0.8157170267685178,
0.9860264495378019,
1.253983790325984,
0.6535416646674022,
0.47409955346718574,
0.8798330365001463,
1.371165503377374,
0.6964303277886761,
0.755500876905032,
1.150546328691376,
0.3437508103385477,
0.485307810571245,
0.7219584851126697,
0.29442738253085027,
0.4049086158849841,
0.5351744073356577,
1.0525658353834653,
0.3580291520141125,
0.5939040895379637,
0.5557086437676062,
0.5391045271464353,
0.39553965092462434,
0.3735075244288823,
1.0786123681311364,
0.5463615153339553,
0.463316066408491,
0.44621996612743514,
0.3820256364734862,
0.4151596525259533,
0.1118751723257326,
0.8663128517994356,
0.9096225297560756,
0.7269194109373684,
0.9122801125337442,
0.641860512284228,
0.8319068241026344,
0.7184776183686608,
0.46497244281437566,
0.6508405397923435,
0.5209071344939292,
0.4543288522399254,
0.5829596392132322,
0.9275509934462636,
0.4948563951478026,
1.1199536967386903,
0.8876714039103254,
0.7215692462474885,
0.8777417965953798,
0.2537341312660219,
0.9136679438380443,
0.2668640078062361,
0.9352020825374199,
0.8532260804844294,
0.5388474147921921,
0.7958664757742199,
1.034469784685504,
0.8991313749213092,
1.089972052452089,
1.0450669978420806,
0.7033534711122825,
1.1993687071727375,
0.8501536093242995,
0.9062922673554012,
0.5432606109052269,
0.49567380682192186,
1.0374120771862074,
0.7688286378586915,
0.9730075512252295,
1.1139542563326688,
0.3460445167614073,
1.0143224595515967,
1.0539923080983777,
0.5552511981769991,
0.9988860191471728,
0.3784024798686873,
0.346310391382999,
0.856239575514884,
0.5909113077863566,
0.6108176908974051,
0.8990990149350612,
0.7478989451961275,
0.8815397693916925,
0.44193650382263355,
0.9090830325094228,
0.9754637418540082,
0.6080649815039615,
0.46452910060514957,
0.5177313545053387,
0.41765448867011085,
0.8777784993195329,
0.38068228139485094,
0.9446382746562131,
0.7235127405869907,
0.8254622029290207,
0.9248001870897204,
0.5582800985269258,
0.5156901162403411,
5.712554910550741,
0.3178340658345531,
0.209620872511189,
0.32364017186116056,
0.07038423458982083,
5.1465783345930385
],
"default_training_normalization": "checkpoint_normalization.json; preserves existing output units; emotions remain 0..4"
}