JobLens is cool! This past month we had the same idea -- it's been a fun project:
https://hnjobs.u-turn.dev
ChatGPT does an incredible job parsing, but then lots of effort goes into normalizing and deduping each field. Long story short, your results look quite good to me!
hubraumhugo 774 days ago [-]
Your project looks quite impressive as well, especially the extracted URLs to apply and the candidate profiles, didn't get that far yet :) Automating tedious work like data extraction and transformation is a great use case for LLMs.
boxcarr 774 days ago [-]
Not sure if you found this as well, but gpt-3.5-turbo-016 does a poor job following instructions other than parsing. So, to work around this, we prompt gpt-3.5-turbo with the rules we want applied say an extracted field and then go back to gpt-3.5-turbo-016 to parse with chatgpt functions.
Bottom line, every single post requires approximately 10 different prompts to refine the extraction.
hubraumhugo 773 days ago [-]
yeah I had the same experience, divide and conquer works better than trying to do everything with a single prompt. Are you interested in comparing notes? Feel free to ping me via the email in my profile :)
version_five 774 days ago [-]
Hi, just a thought, I backed out as soon as I saw I had to pick from a giant list of industries. Most people don't have rigid criteria like that, and if the do, a conventional search is easy. Conversational AI has a potential to more naturally arrive at interesting jobs, either through asking for a freeform description of what you want, or maybe a back and forth discussion.
great_psy 773 days ago [-]
I got turned off by that as well, but I submitted the form anyway, and it loads correctly. I guess no fill in means *
hubraumhugo 773 days ago [-]
You're right, I'll improve this.
SCUSKU 774 days ago [-]
Wow this is amazing. This is more or less what I was trying to build https://hnresumetojobs.com to look like, I am reconsidering whether or not to continue now.
I tried using the OpenAI functions API to do structured text extraction, but found that it would hallunicate a lot of things. Do you guys have that problem? How did you guys go about solving that?
SomewhatLikely 774 days ago [-]
Having the 'unchecked' visual actually have a checkmark threw me off. I was thinking, oh goodness, I have to deselect all of these.
OccamsMirror 774 days ago [-]
Ask ChatGPT to improve your prompt to stop it hallucinating. I’ve had a lot of wins doing that.
mmaia 774 days ago [-]
Looks like it's getting some fields wrong. For examples, "Global" location is matching other text not related to the job location.
Still, a good example of OpenAI's functions. I was using direct prompt for YAML parsing before and migrated to functions for my job board https://aijobnetwork.com/
Need a category for design. Right now all design jobs are scattered around, mostly in product management. Also as others have mentioned, having to make all these decisions before searching is hard to get through without abandoning.
rahimnathwani 774 days ago [-]
This is cool. I'm curious about this part:
We leverage this to map all the unstructured job postings into the same unified structure. The new GPT functions feature and the extended context windows are really helpful for this.
Does this mean that, rather than using the gpt-3.5-turbo chat API, and using a system prompt that describes the output format you want (JSON?), you instead describe your output format in the same way you would if you were going to call a function?
I was going to build something similar for funsies but now I don't have to :)
I'm curious the page mentions not being able to scrape sites that require a login but then I saw LinkedIn as an example. Will LinkedIn eventually be supported but require a login or something?
timhigins 774 days ago [-]
Site is down for me, maybe the HN hug of death? Anyone else seeing this?
baxtr 774 days ago [-]
Nice! Is there something similar for freelancing gigs?
Laaas 774 days ago [-]
All I want is a list of jobs that involve functional programming.
ChatGPT does an incredible job parsing, but then lots of effort goes into normalizing and deduping each field. Long story short, your results look quite good to me!
Bottom line, every single post requires approximately 10 different prompts to refine the extraction.
I tried using the OpenAI functions API to do structured text extraction, but found that it would hallunicate a lot of things. Do you guys have that problem? How did you guys go about solving that?
Still, a good example of OpenAI's functions. I was using direct prompt for YAML parsing before and migrated to functions for my job board https://aijobnetwork.com/
I'm also using OpenAI's embeddings for recommendations: https://medium.com/@mauricio/how-to-add-recommendations-to-a...
I'm curious the page mentions not being able to scrape sites that require a login but then I saw LinkedIn as an example. Will LinkedIn eventually be supported but require a login or something?