Hacker Newsnew | past | comments | ask | show | jobs | submitlogin

Twitter's firehose is extremely valuable data, just not at the lofty valuations the company was given.


What exactly is valuable about it? I keep hearing this but it is just a jumble of collective stream of consciousness. Has anyone that has read Ulysses actually made any sense of it? Twitter is the collective analog of Ulysses.


>it is just a jumble of collective stream of consciousness

Much of it isn't even that. A huge proportion are not even real humans but are instead accounts being operated by one of the myriad of pieces of software that enable automatic posting on behalf of a human.

Perhaps my view is skewed as we have a travel blog and know many other travel bloggers but a vanishingly small percentage are real humans actually typing their own tweets in real time.


However, most of that software can be filtered and controlled via the API. Part of the reason Twitter is so hostile to third party apps (with exceptions) that duplicate its own functionality is so they can get analytics about actual users and not marketing bots.


Do you know if they publicly release analytics about humans Vs bots? I would be interested to know the proportion of tweets from actual humans.


Trend forecasting, market analysis -- mineable data for purposes other than just display (their own) ads.

Shazam [1], a company that makes an app to identify songs recorded through the microphone, makes money off of figuring out which songs get a lot of interest. Twitter can do something similar, if they don't already.

[1] https://www.shazam.com/company


It can be pretty useful for sentiment analysis across large populations - for a lot of different purposes/clients


Do we have sentiment analysis techniques that actually work now? Last I checked they weren't so great.


It is insightful to point out the widely held belief that Google Reader would never be cancelled because its an incredibly tightly curated data set of what real humans actively seek out and hard working devoted humans actively filtered the wheat from the chaff. All for free, for google to monetize.

Everyone knows, that everyone knows, that its a very valuable dataset. But nobody is apparently interested in paying for it or otherwise monetizing it. Shades of "emperor has no clothes". The end result is Google Reader got cancelled.

If twitter was not standalone and was a mere subproject of a real company, it would likewise be long gone. They have no plan B so they'll have to keep doing what they're doing until they run out of money or are purchased, pretty much a Yahoo situation. Its like a decades old portapottie at a concert venue, everyone goes there does not mean its worth anything to anyone.

It would be a nice, cool world to live in where sentiment analysis is a thing. Its such an obvious and tasty idea, that unfortunately is completely wrong.


It's junk for a lot of use cases.

1. As another user mentioned, more tweets are from bots.

2. Lots of sarcasm and rhetorical questions in tweets.

3. Nobody tweets about their hemorrhoids. Or abortion. Unless you're GoPro, good luck finding insights.


> 3. Nobody tweets about their hemorrhoids. Or abortion.

There's an informal support group that tweets about perinatal depression, OCD, and psychosis. This includes people who've taken the choice to have abortion (or even sterilisation) rather than go through what they went through before. ((Watch out, this is just an informal group and there's not much in the way of clinical or data governance.))

https://twitter.com/PNDandMe

Here's a british men's health magazine advertising itself with advice about hemorrhoid prevention: https://twitter.com/MensHealthMag/status/752076400102895616

I pretty much only tweet about treatment for mental illness.

I'm not sure how useful this kind of stuff is to other people.


It can, but at what price? Parent was questioning this data being "extremely valuable" - is sentiment analysis at $.001 per user an attractive proposition to a client? Likely. Is it still a viable product at $.01, $.10, $1 per user?




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: