Manoj -
See the following API for suggestions, you can use this for partial word as well as full phrases. The authorization is the same as the other API using your api key.
There are two types of suggestions returned. The category = "Topics" indicate the keyword or phrase that matches the input. The category = "Questions" indicates the question containing the keyword or phrase that matches the input.
I will add the density of topics sometime later this week,
Cheers,
Sharon
On Tue, Jan 10, 2023 at 7:26 PM, Hemen Parekh <hcp@recruitguru.com> wrote:
Hello, Sharon,
Last evening, Manoj / Prasad gave me an online demo of the working of the API , within a “ tentative “ U / I designed by Atidan team ( - which, I plan to redesign today )
It worked well, though it took a little longer for answer to appear
What we now need ( on the double ! ) is for you to increase the “ scope “ of the API to ( also ) deliver the following features , which you say, are possible :
# List of topics
# Questions ( 5 / 7 , related to the selected topic ) pop up
# Voice clone ( answer text to be simultaneously read out in my own voice )
During our discussions, it was felt that there is a STRONG CASE , to declare upfront , which “ answers “ were sourced from “ Personal AI of Hemen Parekh “ , AND , which answers were sourced from ChatGPT
In absence of such a DISCLAIMER, we would be “ misleading “ the visitors in believing that Hemen Parekh is “ Omnipresent – Omnipotent – Omniscient “ ( what ChatGPT would soon appear to be ) !
I urge you to incorporate such a disclaimer in case of those “ answers “ where ( largely ), the source is ChatGPT
By a copy of this mail, I am requesting Manoj to write to you , if he wants to further elaborate on the contents of this mail
Regards,
Hemen
Subject: RE: FW: Keywords for sample content
I agree to Sharon, both can be achieved (while typing we can show the questions or after selecting the keyword we can show the questions), since API can return both keywords and questions.
@Sharon Zhang please let us know when you can provision the API for the same.
Regards
Manoj Hardwani
|
Subject: RE: FW: Keywords for sample content
CAUTION: This email originated from outside of the organization. This message might not be safe, use caution. If in doubt, do not open any attachments or links in the message. |
Ya I think that's up to you all how the user gets the topic in there — either typing or clicking would be fine. It's not a required on our side since the topic is sent from your side.
Cheers,
Sharon
On Mon, Jan 09, 2023 at 10:09 PM, Hemen Parekh <hcp@recruitguru.com> wrote:
Hi Sharon,
Once you show the “ list of topics that matches “ the typed prefix, why require that “ user types in any topic “ to see a list of questions ?
Can she not , simply click on any “ displayed topic “ and see the questions pop up ?
hemen
Subject: RE: FW: Keywords for sample content
Both are possible. Although right now our topics are mapped to keywords and not higher level abstractions.
1/ It sounds like the search functionality is our autocomplete functionality where you can type prefixes and we'll show the list of topics that matches. [See below when I type in doc … ] # of occurrence is also part of this payload. [See screenshot 1]
2/ Once the user types in a topic, we can also show the list of questions that matches the topics the AI has learned.
Let me know if this is what u are looking for,
Sharon
On Mon, Jan 09, 2023 at 9:43 PM, Hemen Parekh <hcp@recruitguru.com> wrote:
Hi, Sharon,
Thanks for your prompt reply
1 )….. I am thrilled at the possibility of API delivering the list of “ topics “ ( in my draft U / I , I have used the word “ topic “ – and not “ keyword )
During my meeting with Manoj / Kartavya, it was decided that :
# We will NOT , show alphabets ( A ..B..C.. ), for visitors to select / click, in order to view a “ drop-list “ of topics starting with the clicked
alphabet
# Instead, we will provide a SEARCH BAR , where a visitor can type .
As soon as she types ( say ) D, all the topics starting with letter D will appear. As soon as she type the next letter ( say )
“ O “ , all the topics, starting with “ Do “ ( Eg. Document, Domain, Dormant..etc ), will appear , with number of occurrence in brackets
2 )…… I do not want the autocomplete feature ,
BUT,
Clicking on a topic ( in the drop-list below search bar ), should transfer that word ( say “ Document “ ) , into the Question Box . At this stage, we
have following options :
# Visitor is advised to frame / type a question which contains the topic “ Document “
OR
# We, ourselves, pop up a few questions containing that word , and ask visitor to select / click any of those questions
Since many visitors are not “ adapt “ at framing questions that make any sense, I prefer the second option, - visitor don’t have to struggle to frame a question
Choosing the second option also has the advantage of eliminating the “ starting inertia “
Once, a visitor has successively selected 2 / 3 , “ prepared “ questions, and received relevant answers, her confidence goes up
Can this be done ?
Regards,
hemen
Subject: Re: FW: Keywords for sample content
Hey Hemen —
1/ We can probably fetch the list of keywords for you and possibly expose that through an API to show top keywords. I can gather a list to see if it would match what you are thinking.
2/ If by Search by keyword you mean the autocomplete feature that we have here — we can also expose that API.
On Mon, Jan 09, 2023 at 8:10 PM, Hemen Parekh <hcp@recruitguru.com> wrote:
Suman,
Please take a look at the keywords list ( = “ Topics “ on my Personal.ai page )
Despite Manoj putting in a lot of “ parsing “ efforts over the past week , this list cannot be used ( as it is ) in My Digital Avatar U / I , for the following reasons :
It covers only “ 50 % of the content “ from Blogger site “ back up “ taken about a month ago ( mostly “ text files “ of my blogs / notes / reports )
Ø To extract keywords from the balance 50 % could take another one week
Ø The list does not carry , within brackets alongside , the number of times ( frequency ) each keyword has been used by me in my Digital Content
Ø Since the “ Frequency of Occurrence “ is missing, it is not possible to re-arrange the list in the “ Descending Order of Occurrence “
Ø Only through such re-arrangement , the IMPORTANT keywords ( topics ) would rise to the TOP of the list, with unimportant words sinking to the bottom ( enabling some cut-off )
Ø If , over the past 30 years , I have used a particular word very frequently, it is a clear indication that the “ related text “, denotes my “ Area of Knowledge “
Ø In my Digital Avatar U / I , we want to direct ( in a subtle way ) , our visitors to click / select , such TOP RANKING keywords ( topics ), where the probability of getting a good / relevant answer, is much better . Without rearranging ( in descending order of occurrence ), this cannot be achieved
Ø Then there is the problem of > How often can we take a back-up ? How often can we repeat the “ parsing / rearranging “ exercise ?
Ø On top of this, keywords extracted by Manoj, will NOT perfectly match with the TOPIC words that you display ( on my ai page ), simply because you extract from 100006 Memory Blocks
Ø It is not desirable to have TWO , totally different DATABASES ( one of Keywords with Manoj and another of Topics with Sharon ), which need to be SYNCHRONIZED again and again , even if done programmatically . This like wearing two wrist-watches, one running fast and another running slow – and having to synchronize these every 2 hours !
Ø It is also important to remember that, for the past 2 month, I am uploading many TEXT documents on my persona.ai page which are NOT being uploaded on my blogger site. Hence Keywords and Topics can never get synchronized !
In light of the foregoing, I urge you to consider the following :
Ø Manoj give up effort to parse / extract keywords and store in a LOCAL DATABSE for displaying in U / I ( attached )
Ø In the “ Search for Keywords “ feature ( left column of U / I ), the TOPICS get delivered from Personal.ai , through API
Ø Obviously, this might cover ( may be ) , 5,000 TOPICS ( to be arranged ALPHABETICALLY and in DESCENDING ORDER of frequency
Today evening at 6 pm ( IST ), Manoj is giving me a DEMO of the working of QnA API ( Micosoft Team )
If you ( or any member of your team ) wants to watch, here is that link :
Microsoft Teams meeting
Join on your computer, mobile app or room device
Click here to join the meeting
Download Teams | Join on the web
Or call in (audio only)
No comments:
Post a Comment