Job-and-Talent-Recommendation-System

Aim
Business Values
Process
Data Collection
Data Preprocessing
Optical Character Recognition
Natural Language Processing
Recommendations
Result
Conclusion
Next Step

Aim

Assist applicants to search for potential jobs
Assist recruiters to find talents

Business Values

Jobmigo brings the job world to you
- Simple to use: Drag and Drop
- Save time
- Tailor-made dashboard

Process

For job seeker: Firstly, Input the CV and then it will be processed by OCR. After that, the CV and job posts will undergo NLP process. And then both CV and job post will be compared by varies methods to find the similarity. Lastly, the system will list out recomendation of jobs.

For recruiter: Firstly, Input the job post and it will undergo NLP process along with CV in database. And then both CV and job post will be compared by varies methods to find the similarity. Lastly, the system will list out recomendation of candidates.

Data Collection

Web Scraping Scraping useful information from websites, including Title, Company, Location, Salary and Post Date etc.

Data Preprocessing

Extract useful content from scraped data. E.g. HK$ 11,000 - HK$ 25, 000/month to 11000, 25000.

Job:

20,137 jobs
Across 5 main industries (including main industry such as finance, banking, logistics, and IT)
Method: Web scraping from Jobsdb

54 CVs
From friends & google search

Optical Character Recognition

Convert a pdf file of text document into a machine encoded text. In other words, it can be further processed as a string format in jupyter notebook.

Natural Language Processing

1.Word tokenization:

2.Stop words removal

3.Lemmatization Converting a word to its base form.
"motivated" to "motivate", "Learning" to "Learn".

4.Bigram Collection Finder
Finding meaning double words.

Recommendations

CountVectorizer
- It converts a set of strings into a sparse matrix - One hot encode the text document

TF-IDF (Term Frequency-Inverse Document Frequency)

Term Frequency, TF: The number of times a term occurs in a document
Inverse Document Frequency, IDF: It measures how important a term is

(Please note that for the word "earth", the IDF should be log(2/1)=0.3)

KNN

Cosine Similarity

Result

Streamlit Demo

edited.mp4

Conclusion

Jobmigo provides a robust recommendation platform for both job seekers and recruiters. The goal is to simplify the job seeking and candidate screening process, this can be achieved with Jobmigo. It is easy to use (just upload and run). For recruiters, it saves time screening CV manually and provides a less bias result; and for job seekers, it provides a tailor made and interactive dashboard with relevant job opportunities.

Next Step

* Further develop recommendation system to return accurate relevant results
* Engage with job posting companies to gain access to database / API
* Invite applicants to join the CV database
* More features on the dashboard

thoadao0301 / job-and-talent-recommendation-system Goto Github PK