Shuklendu
18/06/2012
Job Alert Recommendation System
I have written to you about this in my emails of 08/09 June.
I enclose some of my older notes on the same subject.
Not everything that I wrote in those notes may be valid today, but the fundamentals of
▶ Predictive / Personalized Job-Alerts
based on
▶ Development of a “Match Index” for each newly arriving job-advt
based on
▶ Keywords – and their weightages – contained in past Jobs Applied,
remains.
Regards
Hemen Parekh
“My Comments on Pg: 42”
In my emails (June 08/09, 2012) to Shuklendu, I have laid down the “logic” to be followed in “Job Alerts Recommendation” algorithm, which will be based on “Jobs Applied History” of each candidate.
I have suggested to highlight/hyperlink (to Wikipedia?), “keywords” found in each recommended advt.
Obviously, these are the very same keywords which that candidate has “seen” in previous advts — and may be the keywords that made him “Apply” against those previous advts.
So, when he once again “sees” those same keywords in next job advt, his (pleasant?) memory will make him “Like” this next advt!
My attempt should be to “trigger” in the brain of the jobseeker, “pleasant/happy” memories, upon seeing a job-alert “recommended” by our algorithm.
When he “sees” the SAME keywords, will it make him feel “GOOD”?
Notes from “World Wide Mind” (Author: Michael Chorost)
Dated: 18/06/2012
Pg = 42
The implication for brain implants is this: to make someone “see” something, it may be sufficient to trigger a memory of having seen a similar object before.
As strange as this sounds, it happens all the time.
In dreams, none of the dreamer’s sensory input is actually coming from the senses. It is all assembled, willy-nilly, from memories.
Dreams are an extreme case, but even in waking life, a great deal of brain’s sensory experience is actually internally generated by raiding memories.
This doesn’t mean that the external world does not exist, or is irrelevant; when you see a door and open it, much of your experience is coming from sensory information. But quite a lot of detail is filled in, such as information about shapes, colors and textures……
That’s a tip-off that the brain’s stored expectations — that is, its memories — shape experience to a much greater extent than commonly realized.
My Comments on Pg: 47 / 49
For my purpose,
▶ Analyzed Portion of data
≡ keywords found in “Jobs Applied History”
▶ Unanalyzed Portion of data
≡ keywords found in jobs NOT APPLIED yet.
I am looking for a “pattern” of keywords, in (say), 100 advts a candidate has “Applied for” earlier.
The higher the “frequency”, the higher the “weightage” for a given keyword.
So, it tells me, what advts (containing keywords with high weightages) will “appeal” to him.
“Weightages” are my “prior knowledge”.
If he has liked these “keywords” in the past, then he is likely to “like” them in future as well.
This is how we will try to “decipher his intention”.
Pg = 49
Note that in order for blackboxing to work, you have to know ahead of time, what you are looking for.
You have to be looking for a neural pattern or a P300, with the prior knowledge of what they mean in a given situation.
The algorithms are inherently closed-ended.
The machine knows that a given pattern means Add and another one means Subtract……
A pattern-matching methodology, by its very nature, can tell us nothing at all about the subject’s conscious experience……
……algorithms can model the information-flow within a brain, and from that make predictions about what the brain is attempting to do.
They have become pretty good at detecting and executing simple intentions.
No comments:
Post a Comment