jobPost-generator

Inspiration

I wanted to do a RNN to originally generate reddit posts, but hearing about access to the icims api I wanted to instead use it for something that was applicable to that api

What it does

It uses a recurrent neural network to do a character by character prediction and generate a job posting, and using the icims api would be able to push the Job Post to the database

How I built it

I used a self built webscraper to grab sample job posts from indeed.com and converted it into jsons. I then used tensorflow and keras to preprocess and build the recurrent neural network. Finally I used flask to build a sample front end that allows users to easily generate and post the job listing onto the icims database

Challenges I ran into

Since I did not have access to a GPU I was unable to test frequently, and I was unable to make a recurrent neural network that had a very low loss score. However, despite that the loss score was less than .1 and still managed to output decent results. Also I had a very small dataset, and because of this I wasnt able to learn more complex relations.

Accomplishments that I'm proud of

I think the biggest accomplishment was that I had done this all by myself and was able to get a decently working neural network up and running

What I learned

I learned how to use recurrent neural networks for text generation, and setting up a neural network to work on a backend

What's next for jobListing-generator

There are definitely some hyperparameters that could be tuned with the recurrent neural network, and I would also like to turn it from a basic recurrent neural network into a LSTM neural network. Finally I also would like to train the neural network on a large dataset since the dataset i trained on was very small

Built With

• beautiful-soup
• flask
• icims-api
• keras
• numpy
• python
• selenium
• tensorflow