Tag Insights ML Router

Model Metadata

72
Training Docs
73.3%
Accuracy
10040
Vocabulary Size
3
Classes

Trained: 2026-03-15T20:43:55.504258+00:00
Labels: commercial, medical_affairs, r_and_d

Auto-Retrain Status

16 more completed analyses needed before the model retrains automatically.

4 new since last training · retrain interval: every 20 completed jobs · total completed: 76

Per-Class Performance

Class Precision Recall F1
Commercial 0.6 1.0 0.75
Medical Affairs 0.8 0.571 0.667
R And D 0.8 0.8 0.8

Top IDF Terms

Terms with the highest inverse document frequency — rare but distinctive terms in the training vocabulary.

interventions that interventions in interstitial interpreted interpretation and interpretability interplay between interplay international journal investor relations investment yet complementary 15 doi 14 12 months 12 2026 113 11 11 regions 11 2025 11 1649

Top Terms: Commercial

Learned logistic regression weights — terms with the strongest positive signal for this audience.

planning
0.2585
narratives
0.2387
assumptions
0.1776
adjustments
0.1591
resilient
0.1572
market
0.1478
organizations
0.1407
internal
0.1388
teams
0.1334
commercial
0.1332
market access
0.1323
commercial teams
0.1323
stock
0.1318
stakeholder
0.1295
kols
0.129

Top Terms: Medical Affairs

Learned logistic regression weights — terms with the strongest positive signal for this audience.

careful
0.1385
evidence interpretation
0.1347
medicine
0.132
review
0.1229
requires
0.1096
observational
0.1085
data can
0.108
dialogue
0.108
stakeholders
0.1039
patterns
0.1027
quickly
0.0993
inconsistent
0.0983
fatty
0.0927
body
0.0924
infections
0.0881

Top Terms: R And D

Learned logistic regression weights — terms with the strongest positive signal for this audience.

learning
0.2273
decision
0.1991
decisions
0.1817
prioritization
0.1787
output
0.1502
portfolio
0.1447
differ from
0.1429
when exploratory
0.139
exploratory
0.1369
work
0.1351
effective
0.1327
more effective
0.1303
collaboration
0.1286
behind
0.1253
assumptions that
0.1244
Processing request…
This can take a few seconds.