56
github
7y

Manual Data Entry: Most boring job
This reminds me of one conversation with one of my faculty..
Faculty: Why not try some Machine Learning Project?
Me: Cool. Any ideas you have already thought
Faculty: Comes up with a really noble idea
Me: Awesome idea. But we need data
Faculty: Don't worry. I will get it. Just help me setup Hadoop (see the irony.. no data yet, and he wants big data setup)
Me: But we don't have data. Let's focus of data collection, Sir
Faculty: I will get it. Don't worry. Trust me.
( I did setup for him twice coz he formatted the system on which I did the setup first time)
After 6 months,
Me: (same question) Sir, Data??
Faculty: I got it.
Me: Great. Give me, I can start looking into it from today.
Faculty: Actually, it's in a register written manually in a different language (which even I can't understand) I will hire data entry guys to convert it into English digital contents.
Me: *facepalm*
Road to Manual data entry to Big Data
Dedicating this pencil to the individuals keeping the register up to date and Sir in hopes of converting it into big data..
Long way to go..

Comments
  • 9
    Manually entering enough data to train a machine learning project? Wowow noty.
  • 4
    @Ashkin manually translating + entering translated data.. double sided sword to instantly kill this project..
  • 1
    @illusion466 I doubt even any NN say CNN could be used, because the content in the ledger was not constant. Tons of NA fields, and decomposing individual rows to feed as single image with a target class(supervised classification problem) will itself be a challenge..
  • 1
    @runfrodorun I think learning directly with be easy for now, instead of creating a NN and then learning from it..
Add Comment