ResuMatch

Out of job due to COVID? Find smart job recommendations with a simple resume drop.

Comment

Click here for the slides

Inspiration

Unemployment has come under the spotlight recently with many workers losing jobs due to the COVID-19 pandemic. In particular, total employment in Singapore contracted by 19,900 in the January to March period, facing the sharpest drop since SARS.

Workers in the gig economy are particularly hard hit, while fresh graduates are finding it difficult to enter the job market in these uncertain times. Nevertheless, companies still need workers and are hiring. However, it is difficult for job applicants to find the right jobs because they might simply be looking in the wrong place or even looking for the wrong positions!

Current Solution - MyCareersFuture

There currently exists the MyCareersFuture jobs portal that recommends jobs to job seekers based on the skills they have typed in. These skills are based on predefined keywords inside the jobs portal. When the user selects the skills, the skills are compared with the job postings. However, useful information that goes beyond the keywords will be lost. For one, the skills mostly look for hard skills and not soft skills. Hence a job applicant will only ever be confined to the job title he is searching for, although there may be another job that is more suited to his soft skills. Furthermore, using predefined skills means that the database of skills has to be manually updated, which is not feasible especially in the tech industry where new skills are constantly being created. Hence, a jobseeker possessing a new skill X may not be matched to a job that is looking for new skill X simply because X is too new and has not been added as a skill on the jobs portal.

So why not automatically compare the applicant's uploaded resume with the job description without relying on keywords only, and then provide recommendations based on the fit? This way, the applicant does not have to manually key in his or her skills, nor even log in to the portal. The applicant simply has to upload their resume onto the ResuMatch website, and targetted job recommendations will appear.

What ResuMatch Does

ResuMatch simplifies the process for job applicants. It uses machine learning to automatically extract information from the applicant's uploaded resume, and then do a direct comparison with the full description of the job postings without having to compare by keywords only. In this way, jobseekers are no longer constrained to a single job title because the machine learning algorithm will also recommend new job titles that fit with the applicant's resume. This also means that there is no need to manually update the database of skills when new technologies are released, because the machine learning algorithm does not depend on the keywords but can just do comparisons at the document level. Hence, ResuMatch is able to obtain information beyond the keywords to provide more targeted recommendations.

How ResuMatch Works

The following summarizes the steps taken to provide smart job recommendations to applicants based on their uploaded resume:

  1. When the resume is uploaded onto our website as a PDF, relevant information like the applicant's skills and experience is extracted from the resume.
  2. A Bidirectional Long Short Term Memory Network (BiLSTM) categorises job descriptions into job titles like “business analyst” and “accountant”. The model predicts the top 10 job titles that the skills and experiences listed in the resume are likely to fall under, opening up the job applicant to more job choices.
  3. A request to the Adzuna API is made to get job postings that are relevant to the top 5 job titles found.
  4. With the job postings, a state-of-the-art Natural Language Processing model, the Universal Sentence Encoder, is used to encode the job descriptions into high dimensional vectors and capture the nuances behind each word in the description.
  5. The resume information is also encoded into high dimensional vectors and with cosine similarity, the similarity of the applicant’s resume to the job description is obtained.
  6. The top most similar job postings that match the applicant’s resume is returned.

What's next for ResuMatch

I will implement the functionality to allow companies to create their job postings on this platform to provide even more options to workers. In addition, the job postings will be categorised into different categories including "freelance", "fresh graduates", "interns", and so on to facilitate the search for jobs. Ultimately, I hope to integrate the machine learning algorithm with MyCareersFuture to provide more targeted job recommendations -- all users can simply go onto the website, upload their resume and receive customised job recommendations based on their resume, without having to manually type in their skills. Furthermore, users will also receive email updates of new job postings that are highly compatible with their resume. This means that once users have uploaded their resume onto the portal, they do not have to search for jobs every time but instead will be recommended jobs whenever new compatible job openings are posted, automating the job searching process and opening them up to more job possibilities. The machine learning algorithm can also provide recommendations to applicants to upskill based on their resume.

Share this project:

Updates