Hacker News new | ask | show | jobs
Twitter/Weather Mashup I put together over a few nights. (tweetweather.com)
31 points by aviel 6391 days ago
10 comments

This showed up: "Homemade migas this morning; cold brewed coffee & client work the rest of the day"

How about you match only if at least two words are weather-related? Or at least 1 weather related + 1 location related, like "cold in new york."

Thanks! Yea it's definitely imperfect, but I'm limited by the functionality of the twitter search api, what i have been doing and what I can improve on is filtering out of common terms that involve weather related words, so for instance "ice cream", or in your case "cold brewed", which I am about to add to the exclusion list. granted this whole app has only about 15 hours of dev in it, so such a list is going to be FAR from perfect.
Ok. BTW, why are you using the twitter search API instead of Gnip?
I definitely plan on switching to gnip. As i said, tweetweather is just something I put together after work the past week, grabbing from the search api was the quick and easy solution to getting tweets into the system.

Gnip rocks.

Can you explain how you'd use Gnip in this instance? From reading the API docs, it doesn't sound like you can do anything with Gnip + Twitter other than get tweets from / to a particular Twitter user. So you can't get keyword-filtered results from the public timeline, I don't think. Am I missing it?
You're 100% correct. But I'm hoping they'll add the ability to use filters on the whole firehose in the near future.
I concur. There is definitely some room for improvement on the word filtering.

I'm at Pac Sun (205 W Blackstock Rd # 210, Spartanburg, SC)

Otherwise, I think this is an awesome idea. Keep it up.

This rocks. You, or someone, could do a similar Twitter mashup for movie reviews - I just did a few searches for movies and the some of the short reviews are pretty entertaining.

It'd be even cooler (and much, much harder) to do some NLP on the reviews and try to give them a star rating based on your model.

Excellent initiative Aviel. I think it's a great demonstration of the future trends in Twitter mashups!
How did you decide which twits to pick up? Did you just have a list of keywords that were weather related that you picked up? Or did you use some other method?
There is a whole ton of logic in there uses the current forecast, weather alerts, etc then determines based on keywords it finds what terms users might be talking about.

Unforunately weatherbug is struggling right now so it's not working as good as it should be. I'm switching to weather underground as we speak.

Very cool, the ability to post weather via Twitter is something I'd like to add to my social weather startup: www.OtherWeather.com
Found my town from the ip address and then when I tested a post it appeared as fast as I could switch windows and reload.
Wow -- it figured out where I was much better than most systems.

Are you doing geolocation with IP? Or something else?

Just geolocation with ip.
I found it amusing that most of the tweets were by women.
Great idea! Very useful and easy to ue.
I'm in Sydney, asking for weather in Sydney, I don't want to see temperature Fahrenheits.

Other than that, a very nice idea and execution.

Good call. It really shouldn't be hard at all, i can query weather bug with a UnitType parameter depending on the location.

Thanks!

Unfortunately, due to the unreliability of the weather bug api, i've had to switch to the weather underground api which makes this much more difficult. Currently working on this.
Well I guess it can be done easier. Currently, the only countries not using Celsius are United States and Belize. And you are already using GeoIP.