This is a series of articles where I talk about the considerations and procedures we followed while adding a certain feature to SupportFinity platform.
In this article, I will talk about the journey of implementing the CV parsing feature.
What is parsing?
Maybe the term “parsing” is not familiar to some readers, it is simply the process of analyzing a sequence of symbols, often in the form of text or code to understand its structure and meaning according to a formal grammar or syntax. It is a fundamental concept in computer science, linguistics, and various other fields.
Generally, modern parsing is done using Natural Language Processing (NLP). In NLP, parsing refers to the analysis of human language sentences or text to extract grammatical structure and meaning. This can involve breaking down a sentence into its constituent parts, such as nouns, verbs, and phrases, and determining the relationships between these parts. Parse trees or dependency graphs are often used to represent the syntactic structure of a sentence.
In our case, we need to read a candidate’s resume/CV then use it to autofill the profile web page or job application. This should save a huge amount of time and effort for the user.
Why parsing utilities don’t work well?
If you tried applying to a job online before, most probably you have seen a button asking you to upload your CV to autofill the application for you.
While this feature is not new, the efficiency of reading a CV then autofill back an online form for a job applicant is never accurate. This is mainly because CVs come with different flavors and formats, and the way content is written is always different. On the other side, forms come in different shapes and changing mandatory inputs, so we have two moving targets!!
We started to drill down to the core components of any job application, the main reason is that we want to make it easier for the job applicant to apply quickly with details that are sufficient enough for the hiring company on the other side, that was a real challenge.
The ultimate goal of any hiring company is to get their posts to reach to as many viewers as possible and receive as many high quality applications as possible. This usually doesn’t happen! too many applicants is not very good news -especially if they are largely mismatching with the job in place- and of course too few applicants is not good news either.
Two goals we wanted to achieve.
So simply we wanted to attain these two goals:
1- First goal, for applicants, Easy, single and quick experience.
2- Second goal, for companies, High quality applications with sufficient information in each application.
Technologies we used.
OpenAI offers now one of the most sophisticated and efficient NLP models that we decided to use, we also integrated it with our proprietary data algorithm to fine-tune the outcome from in order to achieve stronger matching while focusing on the two goals stated above.
Testing the outcomes
The first goal (for applicants) is the main one we were focusing on, as it is related to the direct experience of the user before and after using the CV parsing utility.
The only way for us to measure the usage improvement was to roll-out the CV parsing feature and compare the average time of registration over a month with the same previous period without this feature.
The improvement was really substantial, we could achieve 47%.2 average time reduction using CV parsing utility!
Here is the comparison graph of the average time reduced before and after adding the CV parsing utility.
We could achieve %47.2 average time reduction using CV parsing utility
How to use it?
2- Follow the steps, use the Parse CV button to save time 😉
Thanks for reading, my next post will be about the email integration feature using the powerful technology we developed at SupportFinity.
You can always email me on email@example.com for any product feedback.