Ask HN: Fake Resume Keywords Finder
6 comments
I’ve struggled with this problem in a large organization as well. Even when it was shown that the resumes being submitted were mostly fake (that is, the majority of the contents were copied verbatim from random resumes on popular resume aggregation sites) the bigger issue was fixing the overall talent pipeline. If the vendors don’t care, and your vendor management doesn’t care, and your HR doesn’t care, then all the tech to detect a fake resume _probably_ won’t help you. IME there didn’t seem to be a lot of inclination or incentive for people to address the root of the problem, but hopefully you have more success.
This!. This exactly is what I go through as well.
Obvious answer to the free service is "no", but can you expand on what you mean by fake?
One of the more inane parts of AI CV scan nonsense is people scatter keywords across the document to avoid being dropped by the autonomous first stage.
One of the more inane parts of AI CV scan nonsense is people scatter keywords across the document to avoid being dropped by the autonomous first stage.
I agree. There is no free service. Fake is when candidates stuff in lot of technologies randomnly in their resume, doesn't add those technologies under any project or real experience in any company (think Tech. skills section vs Project Description section) and sometimes even add experience for the time when they were doing their course.
I've built a SaaS service for this in 2017. After a year, I killed the service to concentrate on another project (which I'm currently continuing). A couple of reasons why I stopped it
> 1. I did not spend any marketing efforts because it was not making any money
> 2. I had doubts if a recruiter will come to this platform only to check a Fake resume
> 3. As I have no experience in recruiting I couldn't decide on what would be the best pricing for the service
> 4. Lastly, I stuck with another idea, that is more generic for the people on the internet to use it
> 1. I did not spend any marketing efforts because it was not making any money
> 2. I had doubts if a recruiter will come to this platform only to check a Fake resume
> 3. As I have no experience in recruiting I couldn't decide on what would be the best pricing for the service
> 4. Lastly, I stuck with another idea, that is more generic for the people on the internet to use it
> 1)...
ML training
> 2) Determine if the resume is fake based on how many keywords are randomnly distributed across the resume
That can be defeated by sorting them. Which in turn can be detected by a category time/distance analysis. I'm working on something similar to match knowledge graphs for my next generation search engine. Nice to see there's another use case.
> 3)...
Trivial
I'd say 4) A quiz on 10% of random technologies the candidate included in their CV. Let them know in advance and you will get shorter lists.
ML training
> 2) Determine if the resume is fake based on how many keywords are randomnly distributed across the resume
That can be defeated by sorting them. Which in turn can be detected by a category time/distance analysis. I'm working on something similar to match knowledge graphs for my next generation search engine. Nice to see there's another use case.
> 3)...
Trivial
I'd say 4) A quiz on 10% of random technologies the candidate included in their CV. Let them know in advance and you will get shorter lists.
There was this 5 yr. experienced candidate, whose resume was sent to us by a preferred sourcing vendor, had 80+ technologies listed in the resume. In one scan we knew it was fake, but nevertheless spent 15 mins on an online first round to reject.
Is there a free service or product which I can use to find 1) whether the resume is fake (based on any AI algorithms) 2) Determine if the resume is fake based on how many keywords are randomnly distributed across the resume 3) Find how many unique technologies are listed, how many are repeated, and things like that.
I tried Standford Named Entity Recognizer[1] to pick some keywords from the text, but the online demo allows only certain text to be scanned.
1 - https://nlp.stanford.edu/software/CRF-NER.html