For a while now, Google has been serving up tweets related to searches as part of its real-time search effort. Now they are making it possible to search the Twitter stream in exactly the way Twitter doesn’t allow — that is, to search for tweets older than a few days. A query like<\/p>\n
cyberwarfare site:twitter.com<\/a><\/pre>\nwill return a bunch of tweets, formatted as Google search results. As of the time I ran this query, it identified 1,380 hits from Twitter. Twitter’s search yielded about 250 tweets, going back to no more than 10 days ago. So far, so good.<\/p>\n
What’s no so good with Google’s UI is that the tweets are returned in order of “relevance” but it is not clear what that means for a single-word query and 140 character documents. Twitter search returned results by date, so there was some transparent organizing principle. In the Google interface, it’s hard to explain why one result is placed above another. Clicking ‘timeline’ doesn’t show any results, as you might expect. To get Google to show results by date, you have to perform the following steps (assuming you’re using iGoogle):<\/p>\n
\n
- Click on “show options…” to open the menu<\/li>\n
- Click “Past 24 hours” or “Past week” or “Past year” links (but not “Latest”) in the time category to show the sort criteria options<\/li>\n
- Click “sorted by date”<\/li>\n<\/ol>\n
The results will now be sorted with the latest item first, but there appears to be no way to sort the list in earliest-first order. Of course you can just go to the last page in the set of pages, and scroll to the bottom, right? Wrong! The documents shown there are not tweets. They are links to lists and to tweeps’ pages that happen to include the text of a tweet that mentions ‘cyberwarfare’. Furthermore, the number of matching documents as reported on the last page (10 as shown on the first page of results) decreased from 1380 to 466.<\/p>\n
It is also possible to filter search results by a specific date range, but when I tried searching for tweets containing the word ‘cyberwarfare’ between 1\/1\/2005 and 1\/22\/2010, I got 380 hits (ordered by relevance). Switching to ‘sorted by date’ mysteriously reduced the number of hits to 369. Clicking on the last page shown reduced the number of matching items further, to 321, although none of these was a tweet. In fact, tweets were returned only on the first page, and only 14 of 50 on that one.<\/p>\n
The problem with indexing tweeps’ home pages is that they change so frequently that the index gets stale almost immediately. It makes much more sense to index the tweets themselves. Google seems to do some of each, but the results are not predictable, not complete, and thus not trustworthy.<\/p>\n
What about Bing?<\/p>\n