kms-hackathon

Introduction

Vision Statement

For unemployed people with low-level skills, the EASY product is a job search consultant that not only connects employers with those who are qualified for the jobs but also helps the unemployed explore their potentials. Unlike online application tracking systems which require domain-specific knowledge and resume writing skill, our product’s focus is to provide a simple but effective UI that brings a better UX to all users, making the application process fast and convenient.

Features

We provide Job Recommendation based on the needs of the users, they can do one of the following action to get the job recommendation from our system:

• Do a quick survey quick our Chat Bot
• Update CV to our system, we support CV Parsing feature and evaluate to give the best recommendation
• Describe yourself to our bot and we will provide the best jobs for you based on your profile

Setup

Pre-requisite:

• Python 3.6+
• Node 8+
• npm 5++
• Prepare the model.bin file (since this file is too large, please contact the contributors of this repo for more infomation)
• Linux machine is recommended

Setup MongoDB :

Run with Docker:

docker run --rm -p 27017:27017 bitnami/mongodb

Install required Python packages for the back-end server (using virtualenv is recommended)

cd webapp
pip install -r requirements.txt

Install required yarn packages for front-end server:

cd frontend
yarn install

Start the application:

Run the back-end server:

cd webapp
python main.py

Run the front-end server:

yarn start

Screenshots

Technology stack

Front-end

• React 16

Back-end

• Flask
• MongoDB
• Docker

We used Flask for fast prototyping the RESTful API server so the Front-end application can communicate to and as a connection between the logic handlers with the UI. The logic handlers (powered by various ML models to handle both users' messages and CV sent from the users) were written in Python, so we could leverage that and the integration process was really fast.

Machine Learning Techniques applied

• Text Classification: State-of-the-art Text Classification based on sub-word feature, supporting Vietnamese.
• Text Clustering: Kmean cluster- unsupervised learning algorithm- is not required any annotations. Thus, the system could be improved significantly by adding more training data.
• We also took advantage of the binary decision tree to give pieces of advice for the job seeker.