Hi HN!
I launched the first version of Resume Worded on HN back in October last year and even though it was just a database of sample resume lines, there was a superb response here. Since then, I’ve been adding to the product every day and I’m super excited to show you guys Score My Resume.
The tool scans your resume for elements that are characteristic of a good resume, and gives you feedback/explanations on how you can improve. These range from simple things like word count (is your resume too short/long?) to a bit more complex things such as checking if you’ve quantified your bullet points, and suggesting similar lines in my database of lines (based on keyword ranking) and metrics (see https://resumeworded.com/metrics).
I think the hardest part was parsing a resume properly into its sections (e.g. people name their work experience, skills, etc sections differently e.g. ‘Work experience’, ‘Work and leadership’, …) and identifying what’s a bullet point (people use hyphens, bullets, or no bullets at all), or what’s a job title. Parsing resumes correctly is a pretty hard problem that modern HR systems seem to struggle at too imo. I used spaCy nlp mainly for pos tagging - I think it is the best nlp library out there (e.g. nltk out-of-the-box is ineffective, while stanford was just too slow) - highly recommend so wanted to give a shoutout. I think my model has a long way to go but I feel it’s getting there. I have a few new features that I’m almost done with but thought i should just ship the product as is before I delay it any longer!
I hope you guys find it useful. Appreciate any and all feedback.
-Rohan
P.S. I’ve tried making the parser flexible enough so that it can read pretty much any template. However, sometimes pdf’s don’t get converted correctly, or your template/format is just not something I expected. So if for any reason your resume fails, try putting your content in this standard resume template from Harvard’s career website (http://hwpi.harvard.edu/files/ocs/files/template_bullet.doc) and reupload. Thanks!
3 comments
[ 2.7 ms ] story [ 16.9 ms ] threadThe tool scans your resume for elements that are characteristic of a good resume, and gives you feedback/explanations on how you can improve. These range from simple things like word count (is your resume too short/long?) to a bit more complex things such as checking if you’ve quantified your bullet points, and suggesting similar lines in my database of lines (based on keyword ranking) and metrics (see https://resumeworded.com/metrics).
I think the hardest part was parsing a resume properly into its sections (e.g. people name their work experience, skills, etc sections differently e.g. ‘Work experience’, ‘Work and leadership’, …) and identifying what’s a bullet point (people use hyphens, bullets, or no bullets at all), or what’s a job title. Parsing resumes correctly is a pretty hard problem that modern HR systems seem to struggle at too imo. I used spaCy nlp mainly for pos tagging - I think it is the best nlp library out there (e.g. nltk out-of-the-box is ineffective, while stanford was just too slow) - highly recommend so wanted to give a shoutout. I think my model has a long way to go but I feel it’s getting there. I have a few new features that I’m almost done with but thought i should just ship the product as is before I delay it any longer!
I hope you guys find it useful. Appreciate any and all feedback.
-Rohan
P.S. I’ve tried making the parser flexible enough so that it can read pretty much any template. However, sometimes pdf’s don’t get converted correctly, or your template/format is just not something I expected. So if for any reason your resume fails, try putting your content in this standard resume template from Harvard’s career website (http://hwpi.harvard.edu/files/ocs/files/template_bullet.doc) and reupload. Thanks!
I definitely think your service is attractive given the price point and (presumably) quick response time. Best of luck to you!
[0] https://talent.works/