People can't spell "engineer" or "developer" even when applying for a visa

A couple weeks ago I published a visualization of H1B salaries in the software industry. You should check it out, here.

It was a smashing success with some 46k visitors. Yay!

But let's talk about how fuzzy the data was. Especially the job title part. Like, people just don't care enough to spell O.o

A visa application is no joke, right? It's an official document, that's going to be read by government employees. It's the kind of thing where submitting a week before the deadline is late.

And your document is judged. Harshly. People hire lawyers just to make sure all their paperwork is in order. That the t's are crossed and the i's are dotted.

You wouldn't expect spelling mistakes to make it through, right?

Wellp ...

I found many spellings for engineer. Everything from the correct engineer to the silly eingineeerr. But most got the first three letters right. The normalization regex is just /eng|enig|ening|eign/.

Well ... maybe counting on the first three letters is pushing it.

The more troubling fact is that people haven't mastered spaces very well. Or my scraping script hasn't. But a large part of the data assumes engineer is a grammatical prefix (or suffix) to whatever your real job entails.

I saw everything from engineerprogrammer to engineerjava.

But let's give people some credit, engineer is a pretty darn difficult word to spell. It's practically latin throws the usual English phonetics out the door.

The word developer though ... Here's the regex I had to use: /develop|dvelop|develp|devlp|devel|deelop|devlop|devleo|deveo/

Yeah ... I don't even know. Like, seriously, I just don't know. How?

Everything from developer to development, both of which are good, to the silly developor, and devloper.

And once more, a bunch of datapoints using it as a prefix ... maybe that's my bad though. Surely it's my bad. Surely it just means .split() isn't the best word tokenizer.

Surely.

Perhaps most interesting fact, though, is that the 81,122 visas in my dataset include 3,558 different job titles. Counting misspellings.

3,558 different job titles. Job titles I normalized to just 11 categories.

That's a hell of a lot of ways to say "person creating business value by getting computers to do stuff". No matter which way you cut it.

PS: if you take away rejected visas, there are only 3,472 job titles left. I wonder how many vanished due to spelling.