0
$\begingroup$

I have a dataset of short job titles (e.g., 'marketing manager', 'system administrator', etc.) and their respective Census occupation code (e.g., 1006 Computer systems analysts). I am interested in building a model that can classify each job title into an occupation. Unfortunately, the job titles consist of only 1 or 2 words, so the feature extraction methods that I've worked with before (mainly tf-idf) do not make much sense for this task. I'd appreciate any ideas for how I might approach this!

$\endgroup$

1 Answer 1

0
$\begingroup$

You could try representing titles with word embeddings (e.g., word2vec, GloVe), and either averaging or concatenating them.

Depending on how large your vocabulary is, you could also look at term frequency on its own. I'm assuming job titles would be at least somewhat standardized, e.g., most titles corresponding to "Marketing manager" will have words like "marketing", "advertising", etc.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.