DEV Community

Vasil Vasilev
Vasil Vasilev

Posted on

TensorFlow to filter PDF files

Hello everyone,

So I got the idea to filter resumes using Tensorflow but I don't know how to do it o:).. I found some tutorial about training the model to recognize if on the image is dog or cat.

So what I want to do? Training the model with some keywords (let's say some kind of dictionary) -> I want to filter pdf file depend on those keywords and store it different folder.

Any help?

Top comments (6)

Collapse
 
lukaszkuczynski profile image
lukaszkuczynski • Edited

I think its great with stripping data from pdf first.
I was thinking how to employ Elasticsearch some time ago..
There are ways how to easily index PDF docs, then you can use similarity or scoring to search.

Collapse
 
msoedov profile image
Alex Miasoiedov

I would use pdf - to- text and then feed the data to github.com/vi3k6i5/flashtext with annotated dictionary of keywords. Elasticsearch seems to complex solution since you only want some basic filtering like Qa/Devops/Java Dev/Python dev to group resumes by category.

If you still want to play with tensorflow I suggest you to think about what kind of feature you can extract from image and what kind of output to expect. That's 90% of success the rest 10% is just to code up tensorflow model

Collapse
 
vascov profile image
Vasil Vasilev

Hey, Alex thanks for the answer makes sense to me!

I have a question: So what you are saying is to create a dictionary with the keywords and then extract the pdf to text and filter the cv correspondingly? Right?

How will you do the PDF -> txt?

Collapse
 
msoedov profile image
Alex Miasoiedov

Well that's for you to figure out :)

Collapse
 
affinestructure profile image
KYLE RASMUSSEN

Why don't you just do OCR to extract the PDF to txt?

Collapse
 
soi_dev profile image
Thành

I'm also interested in the topic of tensorflow :))