You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Currently, we're getting lots of phone numbers being sent in, together with some additional context e.g. "Scam caller pretending to be OCBC".
We vote this as scam today, but conceptually they should not send us messages like that because they'll form the database of stuff that can match to other new messages that come in. Theres an argument that this should be voted info, if you look at it from the matching perspective.
So we might need to design a new flow for phone number reporting. This will require UX effort too.
The text was updated successfully, but these errors were encountered:
Probably should create a new message category in the usersHandler.ts, so there will be "image", "text", and a new "phone".
Phone messages should be saved in DB as just the phone number plain. So in cases where ppl send in like "61234567 fake OCBC" we should just save 61234567, and save the "fake OCBC" in some field that captures additional context. For voting, we should show 1) the phone number, and 2) any additional text
The exact handling logic, TBD. probably should be something like this:
use LLM/regex to determine whether theres a phone number and then determine whether the message is predominantly about reporting/checking a phone number (this one probably need genAI). Because a message like
Easy money
$1000 a month
...
contact 61234567
should not fall into the category of phone number, but a message like
61234567 fake OCBC
should. And in the latter category, we should also use LLM to auto-capture the additional context.
Then if no additional context was provided, we should ask the user for it.
Then optionally, trigger a phone number checking API (but I need to negotiate this with police or whatever first, so this part can skip for now, but jsut to note conceptually)
Then, as mentioned earlier, to trigger the voting process with both the phone number and the additional context if any.
Issue
Currently, we're getting lots of phone numbers being sent in, together with some additional context e.g. "Scam caller pretending to be OCBC".
We vote this as scam today, but conceptually they should not send us messages like that because they'll form the database of stuff that can match to other new messages that come in. Theres an argument that this should be voted info, if you look at it from the matching perspective.
So we might need to design a new flow for phone number reporting. This will require UX effort too.
The text was updated successfully, but these errors were encountered: