Systematic Reviews

Table 3 Entity recognition performance across machine learning models

From: Evaluation of a prototype machine learning tool to semi-automate data extraction for systematic literature reviews

Model	Relaxed			Strict
Model	Precision, %	Recall, %	F₁ score, %	Precision, %	Recall, %	F₁ score, %
SLR 1
BiLSTM+linear	68	59	63	46	39	42
BiLSTM+CRF	75	53	62	53	38	44
BERT+linear	67	67	67	46	46	46
BERT+CRF	74	65	69	52	46	49
Pretrained BERT+linear	68	72	70	48	50	49
Pretrained BERT+CRF	74	72	73	53	52	52
SLR 2
BiLSTM+linear	69	58	63	47	45	46
BiLSTM+CRF	73	56	63	55	42	48
BERT+linear	59	61	59	44	45	43
BERT+CRF	66	58	61	50	45	46
Pretrained BERT+linear	63	67	64	47	50	48
Pretrained BERT+CRF	70	71	70	56	56	55

Bold indicates the best-performing model. The 95% confidence intervals for the F₁ scores are included within ± 0.5 percentage points of the estimates given
BERT Bidirectional encoder representations from transformers, BiLSTM Bidirectional long-short-term memory, CRF Conditional random field, SLR Systematic literature review

Back to article page

ISSN: 2046-4053

Contact us

Submission enquiries: Access here and click Contact Us
General enquiries: info@biomedcentral.com