ziodos / cv-parser-via-customized-ner

NER implementation by spaCy, data annotation by Dataturks.

CV parser

This project is implementing customized Name Entity Recognition (NER) by spaCy. Recreate 11 classes of name entities and train the model.

tech stack

• Amazon S3
• Dataturks
• Heroku
• python flask
• spaCy

customized name entities description

Name: Name of user
Contact: Contact of user, like phone number, email address
Location: Location where user based in or where user want to relocate
Title: Job title user has ever had
Company: Company user has ever worked at
TimePeriod: Period of the time that can be found in CV
Degree: Education
College: Education
Skills: Technical stacks, soft-skills
Years: Preset a year long, like "3+ years"
Others: Links like LinkedIn profile or others

Train your own model

If you want to train your own model, I also wrote a command line tool to train model easily

Installation

python version: 3.7.4

git clone https://github.com/samuelTyh/CVparser.git
cd CVparser
pip install -r requirements.txt

Configuration setting

Set testing data size, iteration times, and early stopping times in config.py

# default hyperparameter in config.py

test_size = 0.3
n_iter = 300
early_stopping = 50

Train

# predict without training
$ python app_cmd.py -f file_you_want_to_parse.pdf -M lib/model_choose_optional/

# train model first then predict
$ python app_cmd.py -f file_you_want_to_parse.pdf

demo site

https://cvparser-demo.herokuapp.com/

Press Choose file and upload buttons to upload CV (only support English version)
Wait for the result from model prediction