🪐 spaCy Project: Entities extraction from resume sections (NER/SpanCat).

This project uses spaCy (NER) / (Span Categorizer) to extract resume entities

📋 project.yml

The project.yml defines the data assets required by the project, as well as the available commands and workflows. For details, see the spaCy projects documentation.

⏯ Commands

The following commands are defined by the project. They can be executed using spacy project run [name]. Commands are only re-run if their inputs have changed.

Command	Description
preprocess	Preprocess different sections data and merge it in two training and validation files
show-stats-raw	Show different stats about raw data
show-stats-training	Show different stats about training data
convert	Convert data to spaCy format
debug	debug training data
train	Train spaCy pipline
evaluate	Evaluate the trained model and export metrics
clean-preprocessed	Delete the preprocessed files
clean	Clean corpus and training data files
resume-training	Resume training spaCy ner pipline with the transformer architecture

⏭ Workflows

The following workflows are defined by the project. They can be executed using spacy project run [name] and will run the specified commands in order. Commands are only re-run if their inputs have changed.

Workflow	Steps
all	preprocess → show-stats-training → convert → train → evaluate

⏭ Train data summary

⏭ Trained model validation summary

The pipline trained is NER and the config used for training is transformer-distilbert.

📊 Results:

📝 Validation Summary:

Validation metrics:

• P: Precision
• R: Recall
• F: F1 Score

P	R	F
0.97	0.96	0.97

📝 Entities Details:

PERSON

P	R	F
1	0.99	0.99

TITLE

P	R	F
0.97	0.97	0.97

ADDRESS

P	R	F
0.97	0.97	0.97

POSITION

P	R	F
0.96	0.93	0.95

ORG

P	R	F
0.96	0.95	0.95

DATE

P	R	F
0.98	0.98	0.98

DIPLOMA

P	R	F
0.97	0.95	0.96

INSTITUTE

P	R	F
0.96	0.96	0.96

LOC

P	R	F
0.97	0.98	0.98

Birth/Age

P	R	F
1	1	1

SKILL

P	R	F
0.97	0.96	0.96