Hello,
Does anyone know of any models that can extract job skills from ads for electrical engineer, hardware engineer, embedded engineer (unfortunaltely I can’t include links of such ads)
Thanks in advance
Pavel
Hi Pavel
Could you give more context as to why gpt-3.5/4 is not an option?
Hi Tony,
Thanks for feedback.
I’m looking for some “trained” model. I’ve already tried some models that accept online service (i.e. one can past job announce text online and run extraction) but for these particular job profiles these models were unable to extract skills (e.g. “C/C++”, “Embedded Programming”, “firmware”, “USB”, “I2C”, “SPI”, “UART”, “layout”, “Altium”, “PCB”, “DSP”, etc.)
Sincerely,
Pavel.
1 Like
Hi Pavel - what is the specific problem you are facing?
Based on experience this should be possible to achieve with GPT 3.5 or GPT 4 with the right prompt. There are a couple of different routes you can take here. In the prompt, you could either include examples of the types of job skills that should be extracted or - even better - a long list of skills the model should choose from.
You could likely yield even better results with a finetuned GPT model, whereby your training pairs are the ad as input and the list of identified job skills in the desired format as your output.
Hi Jr.2509,
Maybe I didn’t clearly state the problem.
There is a list of about 100 links with job offers.
I want to use Python to extract content from links and then analyze it to extract professional skills. The wabscraping part is already done, it works well. What remains to be done is the extraction of skills. I don’t plan to use chatGPT for this, but rather an API in Python. As mentioned I have already tested some models that I found on the web. One of them I found on HuggingFace - jobbert_skill_extraction. From the name, this model is apparently supposed to do what I want, but when I apply the content of an online job offer, this model does nothing meaningful - no doubt it was trained on the other types of job offers. Here is the model link:
https//huggingface_co/jjzha/jobbert_skill_extraction
And here is one of job ads from my list:
www_linkedin_com/jobs/view/3693329848/
You can try and see how it works
Sincerely,
Pavel.
Hi Pavel
As @jr.2509 @raymondyeh mentioned (and I dub this), most probably you won’t need a special model for that. Prompt engineering in gpt-4 will do the magic. 3.5 will probably do as well, but with a bit more of prompt engineering work.
Once again, perhaps I didn’t clearly explain the problem, but I didn’t find anything in common with my problem in your link.
In short my task is as follows:
Input: about 100 (or more) links with job announces
Output: data structure (e.g. Python dictionary) that displays the number of occurrences of a particular skill: {“skill”: occurrences}
Sincerely,
Pavel
Hi Tony,
I’m not familiar with Prompt engineering. Can you just confirm that this approach is capable of accomplishing the task I briefly outlined in my previous post. If so, I will acquit with this “Rapid Engineering” in more detail.
Sincerely,
Pavel.
Correct.
You can write a simple python script that will go through the list of you job announces and using gpt-3.5/gpt-4 extract skills from each and then count them.
That will solve your problem.
Well … I just tried in chatGPT 3.5:
Can you extract job skills from the following announce:
Keyrock was founded in 2017 and has quickly grown from 3 to over 100 people today. As an industry recognized liquidity provider and market maker, Keyrock is a leading European creator of algorithmic trading technologies in the digital asset space. With VC backing including Series B investment from Ripple, a constantly expanding global client base, and revenues growth 13x in 2021, Keyrock is one of Belgium’s fastest-growing fintech companies.
We’re currently active on over 80 exchanges and count numerous marketplaces and asset issuers among our clients, including some of the best-known companies in the digital assets space.
Keyrock’s beating heart is its technical team filled with crypto and DeFi natives, software engineers, digital scientists and academics, who themselves are supported by university professors, industry-leading experts and a strong management team. Over the years Keyrock has become a company that is both idealistic and practical.
We are seeking a highly skilled FPGA Lead to join our team of crypto market makers. The successful candidate will be responsible for designing, implementing, and optimizing high-speed trading algorithms on FPGA hardware.
Key Responsibilities:
Design and develop efficient algorithms for execution on FPGA hardware
Optimize existing algorithms to increase performance and reduce latency
Work closely with other developers to integrate FPGA solutions into our overall trading platform
Stay current with industry trends and advancements in FPGA technology
Strong knowledge of working and navigating within a complex AWS environment
Qualifications:
Bachelor’s or Master’s degree in Computer Science, Electrical Engineering, or a related field
5+ years of experience in FPGA development and optimization
Strong understanding of digital logic design and VHDL/Verilog programming
Experience with high-speed trading algorithms and low-latency FPGA solutions
Excellent problem-solving skills and ability to work in a fast-paced environment
If you are a talented and motivated FPGA developer looking to make a significant impact in the crypto market, we want to hear from you. Please submit your resume and a cover letter highlighting your relevant experience.
If you are a talented and motivated FPGA developer looking to make a significant impact in the crypto market, we want to hear from you.
Here is output:
Job Skills Extracted from the Announcement:
- FPGA Development and Optimization:
- 5+ years of experience in FPGA development and optimization.
- Strong understanding of digital logic design.
- Proficiency in VHDL/Verilog programming.
- Algorithm Design and Optimization:
- Design and develop efficient algorithms for execution on FPGA hardware.
- Optimize existing algorithms to increase performance and reduce latency.
- Experience with high-speed trading algorithms.
- Technical Knowledge:
- Bachelor’s or Master’s degree in Computer Science, Electrical Engineering, or a related field.
- Strong knowledge of working and navigating within a complex AWS environment.
- Collaboration and Integration:
- Work closely with other developers to integrate FPGA solutions into the overall trading platform.
- Industry Awareness:
- Stay current with industry trends and advancements in FPGA technology.
- Problem-Solving Skills:
- Excellent problem-solving skills.
- Adaptability:
- Ability to work in a fast-paced environment.
- Motivation and Talent:
- Talented and motivated FPGA developer.
- Desire to make a significant impact in the crypto market.
- Educational Background:
- Bachelor’s or Master’s degree in Computer Science, Electrical Engineering, or a related field.
- Communication:
- Submission of resume and cover letter highlighting relevant experience.
These skills and qualifications indicate a need for a candidate with a strong technical background, specific expertise in FPGA development and optimization, a deep understanding of digital logic design, and experience in high-speed trading algorithms within the context of the crypto market.
Frankly speaking I expected something completely different … something like that:
“VHDL”, “Verilog”, “FPGA” “AWS” and that’s all … without any bla-bla-bla
This is exactly where prompt engineering skill comes in.
This is exactly where prompt engineering skill comes in.
Can you share a link where I can find information on how to proceed
I personally like website called LearnPrompting. Try starting there.
I love what you’re trying to do, and understand your point on the output not being what you expected. You definitely need a skills taxonomy so that Chat GPT can extract skills from. How to train it to identify the skills is what I don’t know. I wonder how are you progressing with this.
For an open taxonomy you might want to check ESCO, they have open APIs that you could include in your script.
I work with L&D technology, and I believe that a bot that analyzes tasks instead of entire job descriptions would be much more useful. Also as a result, I’d think of a bundle of skills, I mean, what works is having a set of skills to perform a task, and not only some of them.
Looking forward to read how you are evolving!