can you also benchmark this script and see how it compares to watson:
def analyzeFeeling(text) {
if text contains "scared, worried and frightened" then fear = 0.7
if text contains "happy" then anger = 0.1
if text contains "confident" then confident = 0.8
if text contains "conscientious" then conscientiousness = 0.4
return (fear, confident, anger, conscientiousness)
}
I think it's a joke that the analysis provided here by an incredibly complex cloud machine learning service doesn't seem very different from applying an incredibly naive bag of words.
As I discovered when writing an IRC bot that pretended to be human, you can go a really long way with a bunch of cleverly-written regular expressions...
As someone who writes resumes/CVs for people and has had a long career as a tech recruiter, this was interesting on a number of levels.
Lots of resume writers don't even bother with summary statements, but I always use them, and my reasoning is that it is an opportunity to set the lens through which the rest of the document is consumed. If we use certain words in the summary that stick with the reader, they will apply those words to your experience in an almost subliminal way.
I don't know that tone is something I look for, at least not consciously - unless the tone is clearly and overwhelmingly aggressive or overly passive.
One thing that I think this summary leaves out is quantifying the skills - how senior are you?
When I look at the writer's blog, there are mentions of lots of seemingly interesting home automation projects and "lifehacking" type things that differentiate the writer from many in the industry. A mention of those things (tech hobbyist with project experience in IoT and home automation) would get my attention much more in a summary than tone and confidence.
This is very clever! But it requires trusting that what the API spits back is actually reflective of how your words are perceived.
In theory you could test whether an "optimized" CV actually leads to better results. But in practice that's very challenging for one person to do, since the sample size is small and you're testing live with your own career...
It also assumes that the recruiters are looking for this personality profile that the author is shooting for. Maybe they don't want someone with 100% confidence, because such people could also be cocky and stubborn.
That is cool. Most of the sentiment analysis APIs I saw just give a -1 to +1 score based on positive or negative feeling but looks like Watson APIs are much more elaborate.
...but could you have just done this with some basic free open source tools (e.g. a few lines of Python).
With too many of these "look what I did with IBM/Watson" the answer is usually very much yes and thus the challenge with IBM's long term value proposition with their platform. It's cool stuff, but you very much don't need IBM/Watson to do it.
Very clever approach to CV optimization; dare I say, even disruptive. Any time the product of one "robot" goes against the product of another "robot", makes me smile. ;-)
I think this is a cool, and a bit frightening use of cloud/A.I.. Of course, as misja111 mentioned, the CV should reliably play well in HR groups that use the same analytics, and that is where it gets a bit frightening. I would hate to be saddled with someone who has gamed the system with an optimized CV/Resume that was easily green-lit through analytics, but I would hope that some manual filtering is still in place. Then again, it isn't like lying on a CV/Resume is a NEW thing...
How is it different from a candidate who is related to the CEO etc.? If anything (at least in software) for now it would indicate a highly capable and inventive candidate.
I rather hope that a blog entry about optimising one's CV would do even more for one's prospects than an optimised CV!
The next step, of course, would to write software which made changes to the CV; then one could iterate until getting the perfect CV blurb. It'd work in theory anyway …
