let me explain how it works :
# On my blogging site ( www.hemenparekh.in ) , there are approx. 30,000 ( word ) documents ( blogs –
notes – reports – postings – emails – poems etc )
# During past 18 months, my colleagues from office , uploaded all of these into Personal.ai ( one at a time )
# Personal.ai broke up each document into 4 / 8 , small paragraphs called MEMORY BLOCKS
# So , in my case , altogether , it has created approx. 109,700 memory blocks. Personal.ai has serially
numbered and stored in some database
# While converting each document into Memory Blocks , Personal.ai , also ,
# Found 5 / 7 “ Key Topics “ ( Key words or phrases ) in the document being uploaded
# Generated 4 / 5 “ Key Questions “ pertaining to that document . It also generated “ Appropriate /
Relevant “ ANSWERS for each of those questions
18 months ago , these ” questions” and “ answers” and “ topics “ , were quite irrelevant, but over the
months, matter improved considerably
By Jan 23 , these were really “ relevant “ ! The “ Answers “ corresponded with what I had actually written /
suggested, in that Blog ! It was as if I was ‘ answering “
But, for some reasons, Personal.ai stopped this “ generation “ of questions / answers by around Feb 23 ( they
did vaguely hint re-introduction but nothing happened )
Which implied that their API , was unable to “ deliver “ anything worthwhile when a visitor typed something in
the “ TOPIC “ search bar
Of course, once in a while, typing a TOPIC ( in Topic Search Bar , does throw up some keywords and questions
)
But of course, if a visitor “ typed “ a question ( in the QUESTION BOX ), which was related to subject of any of
my BLOGS , then the API does manage to find ( from those 109,700 memory blocks ) , and deliver an ANSWER
which makes a lot of sense !
But if an “ answer “ does NOT make sense, then the visitor gets disappointed – and may decide not to return –
neither to refer my site to friends !
WAY OUT :
Create a local “ Database “ of “ relevant “ questions and enable visitor to “ fetch 5 questions “ , by clicking a
button .
Next , visitor would “ select “ one of the displayed questions, by clicking.
Selected question gets transferred to the “ Question Box “
Clicking “ SUBMIT “ will fetch the relevant answer from Personal.ai ( API )
HOW TO CREATE A LOCAL DATABASE OF QUESTIONS ?
We found a simple / elegant method
In BARD , type following prompt :
Please just list the " questions " from the following blogs , with blog-title displayed for each set of questions
/ https://myblogepage.blogspot.com/2023/05/census-app-precursor-to-votesapp.html
/ https://myblogepage.blogspot.com/2023/05/rbi-has-200-r
BARD replied :
Sure, here are the questions from the blogs you mentioned:
------------------------------------------------------------------------------
Blog title: Census app - precursor to votesapp?
What is the Census app?
What are the features of the Census app?
How does the Census app collect data?
How is the data from the Census app used?
Is the Census app secure?
What are the privacy implications of the Census app?
Blog title: RBI has 200 rupees notes?
Is it true that the RBI has 200 rupees notes?
If so, why haven't they been released yet?
What would be the benefits of having 200 rupees notes?
What would be the drawbacks of having 200 rupees notes?
When do you think the RBI will release 200 rupees notes?
I hope this is what you were looking for. Let me know if you have any other questions.
=================================================================
Usually , I have found that these BARD-generated questions are quite “ relevant “ to the URL / hyperlink of the
concerned blog
Not only that , since the questions themselves , are “ related “ to the blog to begin with , then the “ answers “
delivered by API , are also VERY GOOD
Unfortunately , for reasons unknown , just now, when I pasted these questions in hemenparekh.ai , it failed to
“ return “ ANY ANSWER at all !
Sharon ( of Personal.ai ) has suggested that, I must “ login “ into my personal.ai ACCOUNT , 2 / 3 times a day
, for the API to remain connected with hemenparekh.ai ( which I did but no response ! )
However , assuming that this is only some temporary issue , it seems , I can use the services of BARD to
generate 50 questions at a time ( BARD accepts only max 10 blog hyperlinks, in one prompt )
By deploying a person for 4 hours daily , we can generate 500 questions / day ! .
There are > 3000 blogs + 4000 notes = 7000 document ( hyperlinks in www.hemenparekh.in )
Hence, we can build a LOCAL database of 35, 000 questions !
USER INTERFACE :
No comments:
Post a Comment