Hacker News new | ask | show | jobs
Ask HN: Do you think this is a good startup idea?
9 points by avin_regmi 2121 days ago
I'm working on ML model to classify NLP dataset. I'm thinking about online web app where a user would upload CSV file and possible labels. Back end would use ML to classify NLP texts into possible labels. Would you use service like this? Would you pay for it? If yes, how much?
8 comments

It's a good idea but you will run into the following issues:

* Your users won't have enough data

* Your users will have bad data

* Your users will have sensitive data that they can't upload

* Your model would need to be tweaked for best outcomes

Hahaha, this is all true. Cleaning the data will be very tricky.
I would use it. But only if it worked ;) My idea is that I want a model that will basically detect if my kid is whining or not. As a parent, I hate whining, but I also hate monitoring all the time for whining, and would love a robot to tell my kid "you're whining". If I could set up a smart speaker, or a raspberry pi, or a iphone or android app or something to just be listening all the time and tell my kid to use a normal voice, that would be awesome. So I would pay for it, but I wouldn't pay absurd amounts of money. I would want to test it out and see that it works. If it did, I would need an easy way to export the model into iOS/android and python. If I knew it was going to work, and I wasn't going to run into stupid problems (I get a model but I can't figure out how to run it unless I'm inside your web GUI or something for example), then I would pay a couple hundred dollars. If I didn't know if it was going to work or not, then I would probably only pay 20 or 30 dollars.
This is a hilarious idea. I would love a little robot that could tell whining apart from legitimate pain crying, because god knows toddlers do enough of both, and help discourage the former.
Beware second-order effects. I can see this going horribly wrong, raising a generation of believably-sounding whiney 'victims' suffering from whatever they imagine could get them something good.
Oh damn, so you mean they would be whining without me realizing it? And I would entertain them, not knowing I should shut them out like a child?

I think that's just the current world, friend.

Why would you need ml?

Sampling the background for the correct pitch would be enough

Kids are smart enough to beat a pitch detector.
Solves your problem of whiney pitches. If the kid goes back to normal tone it becomes just a complaint.
I'm going to assume you have little kids. If the kid is over, say 12, then a simple comparison of age and a boolean is all that is needed.
This already exists and all the big cloud platforms have this: https://cloud.google.com/natural-language/automl/docs/quicks...
This requires you to upload a labeled dataset. My idea was more of like you enter instances of text without label and ML model will label those texts for you.
It is a useful idea, but to be honest, it might not be a good startup idea.

if you want to raise vc money, the thing they care the most is the potential market size. this sounds like a niche market, so you will have to target big enterprise users, not hobbyists. big enterprise users can easily hire a data scientist team to build models, unless your solution is cheaper and flexible enough to handle their needs well. You will also need to take care of data security. And you need to have sales or you need to do sales.

if this is a bootstrap/indie hacker idea that doesn't need vc money, I still don't think it is a good idea. personally, I won't choose anything that requires a GPU infrastructure , it is too costly, and hence risky.

I made the same online web app where a user uploads a CSV file and the back-end in uses ML to classify data and produce insights and charts automatically

Link: https://www.youtube.com/watch?v=rVQOGRPiNJs

SEOs would use this. Tailor it to SEO and I can get it in front of a lot of them
That's really good to know. Why do SEOs use this? Would you be able to connect them with me?
do you know your potential users? ask them but not HN readers :)
is it similar to monkeylearn.com