Skip to main content

Table 5 Model 1 CEM results for optimizing Gaussian prior variance and feature frequency threshold.

From: tmChem: a high performance approach for chemical named entity recognition and normalization

 

Internal Evaluation Set

Official Test Set

Setup

P

R

F

P

R

F

σ = 1.0, c = 0

0.8773

0.8758

0.8766

0.8595

0.8721

0.8657

σ = 2.0, c = 1

0.8758

0.8778

0.8768

0.8973

0.8474

0.8716

σ = 0.5, c = 5

0.8766

0.8731

0.8749

0.9009

0.8462

0.8727

  1. Model 1 results for the CEM task for the Gaussian prior variance (σ) value and feature frequency threshold (c) value, which resulted in the highest micro-averaged precision (P), recall (R) and f-measure (F) on our internal evaluation set and the official test set. The highest value is shown in bold.
  2. Results for the CDI task for each experimental setup on our internal evaluation set and the official test set, as measured by micro-averaged precision (P), recall (R), and f-measure (F). The highest value is shown in bold.