Did not receive any answer on the questions below. I think it is important for us developers to understand the direction Twitter is taking in relation to geolocalized searches.
In a nutshell: For geolocalized searches, the streaming API returns a very small fraction (3% in my tests) of what the search API returns. This is because the streaming API only uses the geotagging API to locate tweets, but the search API uses both the geotagging API and the user location field. Depending on your application, both methods can be valuable. In particular, what the search API retrieves make sense in my context but it is not possible to get this using the streaming API. - Can we expect both methods to be supported in the future and can we expect to get a streaming version of what the search API does today? Thanks, Colin On Feb 15, 2:19 pm, Colin Surprenant <[email protected]> wrote: > So basically today we have two options for geo search: > > - use the search api and get results that will include some > incorrectly geolocalized tweets when falling back on the user location > field. > - use the streaming api and retrieve significantly far less tweets but > with a higher degree of confidence in their geolocation using only the > geotagging api. > > Can we expect these two methods to be available concurrently for the > next 3, 6, 12 months? > > I have two problems with this: > > - As developers we are asked to migrate toward the streaming api > instead of using periodic polling, which makes sense. But for geo > search, the streaming api is not a viable alternative for those who > actually prefer/require/want the behaviour of the geo search api. > > - The fact the the streaming api returns far less but more precise > data is not necessarily better, it really depends who you ask. For me, > having lots of geolocalized data that will contain a fraction of > invalid data is far more valuable than having far less but more > accurate data. > > My tests told me the streaming api currently returns only ~3% of the > volume of data the search api produces. If the only difference between > the search api and the streaming api is the usage of the user location > field, then we can certainly say that FAR more people are still only > using their user location field and not using the geotagging api. > > Will you offer an option in the streaming api to fall back or not on > the user location field when evaluating the geolocation of a tweet? > > Thanks, > Colin > > On Feb 14, 2:33 pm, Taylor Singletary <[email protected]> > wrote: > > > > > > > > > Hi Colin, > > > You hit the nail on the head with this observation: > > > > In the doc it says that the streaming API will only return tweets that > > > are created using the Geotagging API (and within the bounding box) but > > > the search API will preferentially use the Geotagging API, but will > > > fall back to the Twitter profile location. > > > The Search API is greedy with those location fields on user's > > profiles. It's not likely this behavior will be emulated in the > > Streaming API with the bright side that you can be more confident in > > the location accuracy in matches on the Streaming API. > > > Thanks, > > Taylor -- Twitter developer documentation and resources: http://dev.twitter.com/doc API updates via Twitter: http://twitter.com/twitterapi Issues/Enhancements Tracker: http://code.google.com/p/twitter-api/issues/list Change your membership to this group: http://groups.google.com/group/twitter-development-talk
