tag:blogger.com,1999:blog-6569681.post8919515757330281665..comments2024-01-15T13:17:33.771-08:00Comments on Geeking with Greg: Google Search API for researchGreg Lindenhttp://www.blogger.com/profile/09216403000599463072noreply@blogger.comBlogger3125tag:blogger.com,1999:blog-6569681.post-66746030723096014122007-08-07T15:40:00.000-07:002007-08-07T15:40:00.000-07:00Hey Greg.Good seeing you at lunch today.BTW. We p...Hey Greg.<BR/><BR/>Good seeing you at lunch today.<BR/><BR/>BTW. We provide access to Spinn3r for researchers who need access to a blog crawler. <BR/><BR/>We already have a few researchers using it already.burtonatorhttps://www.blogger.com/profile/08049781702053733725noreply@blogger.comtag:blogger.com,1999:blog-6569681.post-42785506103859852602007-08-06T15:41:00.000-07:002007-08-06T15:41:00.000-07:00I haven't had a chance to look at the APIs yet, bu...I haven't had a chance to look at the APIs yet, but I would be curious to know whether you could get more than 10 results at a time as a result of your query. When you are testing query expansion, you want to know if you missed the target document by just a few docs (i.e. it was ranked 14th or 15th, rather than 9th) or if you missed it by hundreds of documents (it was ranked 104th) or if you missed it by thousands of documents (it was ranked 2243rd). <BR/><BR/>If you can only get 10 docs at a time, and there is a 1 second delay in between requests for the next 10 docs, it will take you forever to determine this sort of information.<BR/><BR/>And sometimes, you really do want to know that the target doc is 104th and not 2243rd, even if a user would never see it at 104th, because that can help you determine whether or not you are even close to being on the right track.Anonymousnoreply@blogger.comtag:blogger.com,1999:blog-6569681.post-66055698476909309012007-08-06T15:22:00.000-07:002007-08-06T15:22:00.000-07:00Their policy sounds reasonable - all of your examp...Their policy sounds reasonable - all of your examples are solvable by querying ahead of time and caching results locally.<BR/><BR/>If a researcher has a requirement for his project to respond in real time for any given input, it sounds like he's no longer researching but instead demoing, (or selling?), which seems like a fine time to move away from a free backend.<BR/><BR/>Also consider that if the average user searches ten times a day (just a guess), google is committing to shoulder a load of nearly 10,000 times that of the average user per researcher, which is not negligible considering the number of researchers out there who would be interested in such a service.Erik Freyhttps://www.blogger.com/profile/16378049476033112229noreply@blogger.com