Skip to content

Versuch 4: FIDmove Index (TIBKAT Anteil), GND Schlagwörter, deutsch

runnwerth edited this page Nov 26, 2020 · 3 revisions

Die Daten wurden testweise bereitgestellt.

Kollektion

  • FID-Move-TIBKAT-Metadaten mit GND-Schlagwörtern und Dokumentsprache deutsch
  • 36.478 Metadatensätze

Format

FTX

Lernverfahren

entsprechender Teil in projects.cfg

Skript

FIDmove_GND_de.sh

annif-Evaluation

Precision (doc avg): 0.3663934426229508
Recall (doc avg): 0.9910581222056631
F1 score (doc avg): 0.5014939361168362
Precision (subj avg): 0.0008387737432160627
Recall (subj avg): 0.000989529334263565
F1 score (subj avg): 0.0008840220447053122
Precision (weighted subj avg): 0.8085908913754895
Recall (weighted subj avg): 0.9977678571428571
F1 score (weighted subj avg): 0.867929184807981
Precision (microavg): 0.3663934426229508
Recall (microavg): 0.9977678571428571
F1 score (microavg): 0.5359712230215827
F1@5: 0.7059601373125963
NDCG: 0.9958699263303735
NDCG@5: 0.9928464629488181
NDCG@10: 0.9963440406910891
Precision@1: 0.9918032786885246
Precision@3: 0.8224043715846995
Precision@5: 0.6409836065573771
LRAP: 0.9911902845935552
True positives: 447
False positives: 773
False negatives: 1
Documents evaluated: 122

Evaluation nach Schlagwörtern siehe subject_results/FIDmove_GND_de_OP.txt

Clone this wiki locally