I was thinking about this on the way to work this morning while Steve Martin was talking about his experiences with twitter on NPR.
The problem with search right now is that PageRank is not at all a good indicator of individualizing social search results, and contrary to Facebook's, reddit's, or the new, hot social network's goals, the social network ecosystem is constantly evolving. In ten years, I hope, for the sake of progress, that the social networks we have now are mostly dead, relegated to the core of dedicated users similar to sites like metafilter, MySpace and kuro5hin.
What I'm awaiting is a dedicated, personalized (but not 'personalized' or Personalized) social search engine that has enough implicit value for user buy-in. A search engine with enough context to make the facebook wall obsolete.
This hybrid search that Google is attempting wont have the value that they think it does - Wikipedia already does that, and better, without ads. Google needs to step back and think about what made their approach to search a killer app ten years ago and try and understand how to apply that value to the search needs of people today, instead of trying to apply their model from ten years ago to the search needs of people today.
>"In practice, that means that over the next few months you will begin to see subtly different results to searches"
Every time I notice subtle changes to my search results with Google, it's because they are worse [in the sense that they are less relevant to the actual terms I have searched for].
>'"the most important thing about an artist is his greatest works."'
That doesn't mean they are the most relevant results for my search terms.
Perhaps there are dozens or hundreds of improvements over the years you haven't noticed because they were better? People are going to notice things getting subtly worse, not things getting subtly better. Even if they are.
There might of been, but in general I have found that Google's search engine pays much less attention to the actual terms I entered than to what it expected me to enter, compared to the more literal behavior years ago.
When Google search first started suggesting that I had mistyped a term, it still returned results for the mistyping, and just notified me of the possibility. Now, it assumes that I mistyped a search term and returns results of what it thinks I requested. [edit] At times this is helpful but the times when it isn't more than offset any benefit [/edit]
In the old days, Google search would actual give me results from all over the world for my search terms, now it filters them for the US - searching for a foreign term or something that one knows is on a foreign site is a real PITA, these days.
It's all driven, not so much by relevance to me, but by relevance to advertisers. There's little money to be made if "German Bakery" returns a website in Berlin compared to if it returns a map to the one in my local Publix supermarket.
Did you search for Monet[1]? The artworks are actually listed at the bottom (as noted in the article, actually) so they aren't really even suggesting they are the most relevant results. I don't see why offering additional information is seen as a bad thing? It's not like it shows up whenever Monet is in the query[2].
>As further outlined in a Wall Street Journal story on the subject on Thursday, a search for “Lake Tahoe” will produce “key attributes” that Google knows about the lake, including its altitude, location, salt content and average temperature. The query “What are the 10 largest lakes in California?,” meanwhile, would yield an answer instead of just links to relevant sites.
I'm curious what will happen to queries like, 'Thai'. Will they show me local Thai Restaurants? Will they show me travel deals to Thailand? Will these 'attributes' tell me something of Thai culture. God I would hate to have to be working on the query parser with these NLP type changes coming down the path.
>"It’s unclear how the plan overlaps, if at all, with Google’s other recent search initiative, Google plus Your World, which integrates Google+ results into standard searches."
I would have to assume that going forward any tweak to the Google user experience will in some way benefit their social initiative. It seems to matter a lot to them.
Well Google, a word of warning that will probably never be seen. I have loved you almost since you started. You have been a very good companion, which never crashed, (almost) never disappointed me for many years. Then you changed, you thought that adding your dumb immediate search after every keystroke would be a good idea (I hate being wasteful). Then you thought that adding some social component in it would be a good idea but I didn't care much about it. Now if there's more "Social bullshit flavor" in your result, I'm out, I am going to find another Search engine that suits better my needs.
Joke apart, Google has always been the way I search for KNOWLEDGE. Because of this "Social trend", it tends to become more and more aimed at searching for TRENDS. And they are two totally different things and pretty much... incompatible. Don't mix it up Google!
For any other additional information about search or more relevant results, I'm in!
While I really do like the social search feature (seeded with the Hacker News Google+ circle, I see things other people believe are relevant rather than relying on SEO), I agree with you on the instant search. Especially since they enabled it in Chrome's URL bar.
If I type "pho" and hit enter, I expect to be taken to search results about "pho", but instead Google automatically fills in the rest and searches for "phone". I then have to backspace and do another search just for pho. Instant search on Google.com is infuriating as well in that it doesn't work reliably. It will guess three or four times as to what I'm searching for as I type, then give up and say "hit enter to perform your search". That's what I want to do anyway! In the meantime, I've been distracted by a flashing page while I'm typing.
Sounds like it's time for someone to build pg's search engine for the top 10,000 hackers. Google's entering the uncanny valley; it's ai will be alright, but it won't be faster or better than my ability to parse a page of relevant links from 'old Google'.
Just saying that there are folks building search engines, blekko.com and duckduckgo.com are examples.
Many people share your dislike for the privacy aspects (or lack there of) to which Google has gone.
[disclaimer: per my bio I work at blekko.com and I consider myself friends with Gabe over at duckduckgo :-) doesn't take away from the fact that there are exemplars of people building new search engines]
Interesting. I wonder if I happened to see some test of this a long while back? You see, I was looking up the specific gravity of a whole bunch of different things for something (I couldn't find them all on Wolfram Alpha).
Then, out of the blue, Google started giving me the answers directly, saying that it had calculated some number by averaging X sources. But that's not the odd part.
The odd part? The Google search page then asked me if that number was correct. I've never seen that before or since.
Do you mean something like the "Best guess for..."? In that case, I've seen it on a bunch of factual queries. Try searching for something like the "Size of France".
Please provided a picture when discussing google results. It's really difficult to know what your point is when I'm not sure what Google returns to you. The above links seems like a normal search result to me (without any factual google-provided results)
But I saw something different back then, because it had a little yes/no selection asking if I thought the information was accurate. I have not seen that question except for that one time.
When I saw it originally, the question about accuracy was right next to the search box. I didn't click on anything to get it. But that's interesting. I guess it got integrated at some point.
Similar, but I seem to remember the wording being somewhat different.
And after I had done a lot of searches, Google asked me how accurate the information was. As I recall, it seemed fairly reasonable for that particular query.
26 comments
[ 4.1 ms ] story [ 73.0 ms ] threadThe problem with search right now is that PageRank is not at all a good indicator of individualizing social search results, and contrary to Facebook's, reddit's, or the new, hot social network's goals, the social network ecosystem is constantly evolving. In ten years, I hope, for the sake of progress, that the social networks we have now are mostly dead, relegated to the core of dedicated users similar to sites like metafilter, MySpace and kuro5hin.
What I'm awaiting is a dedicated, personalized (but not 'personalized' or Personalized) social search engine that has enough implicit value for user buy-in. A search engine with enough context to make the facebook wall obsolete.
This hybrid search that Google is attempting wont have the value that they think it does - Wikipedia already does that, and better, without ads. Google needs to step back and think about what made their approach to search a killer app ten years ago and try and understand how to apply that value to the search needs of people today, instead of trying to apply their model from ten years ago to the search needs of people today.
Every time I notice subtle changes to my search results with Google, it's because they are worse [in the sense that they are less relevant to the actual terms I have searched for].
>'"the most important thing about an artist is his greatest works."'
That doesn't mean they are the most relevant results for my search terms.
When Google search first started suggesting that I had mistyped a term, it still returned results for the mistyping, and just notified me of the possibility. Now, it assumes that I mistyped a search term and returns results of what it thinks I requested. [edit] At times this is helpful but the times when it isn't more than offset any benefit [/edit]
In the old days, Google search would actual give me results from all over the world for my search terms, now it filters them for the US - searching for a foreign term or something that one knows is on a foreign site is a real PITA, these days.
It's all driven, not so much by relevance to me, but by relevance to advertisers. There's little money to be made if "German Bakery" returns a website in Berlin compared to if it returns a map to the one in my local Publix supermarket.
[1] https://www.google.com/search?q=monet [2] https://www.google.com/search?q=monets+birthplace
Sounds like WolframAlpha.
I would have to assume that going forward any tweak to the Google user experience will in some way benefit their social initiative. It seems to matter a lot to them.
Joke apart, Google has always been the way I search for KNOWLEDGE. Because of this "Social trend", it tends to become more and more aimed at searching for TRENDS. And they are two totally different things and pretty much... incompatible. Don't mix it up Google!
For any other additional information about search or more relevant results, I'm in!
If I type "pho" and hit enter, I expect to be taken to search results about "pho", but instead Google automatically fills in the rest and searches for "phone". I then have to backspace and do another search just for pho. Instant search on Google.com is infuriating as well in that it doesn't work reliably. It will guess three or four times as to what I'm searching for as I type, then give up and say "hit enter to perform your search". That's what I want to do anyway! In the meantime, I've been distracted by a flashing page while I'm typing.
For that matter, who I am going to share my emails with on Gmail? Get Google Plus out of my face.
Just saying that there are folks building search engines, blekko.com and duckduckgo.com are examples.
Many people share your dislike for the privacy aspects (or lack there of) to which Google has gone.
[disclaimer: per my bio I work at blekko.com and I consider myself friends with Gabe over at duckduckgo :-) doesn't take away from the fact that there are exemplars of people building new search engines]
Google is heading the false direction.
Then, out of the blue, Google started giving me the answers directly, saying that it had calculated some number by averaging X sources. But that's not the odd part.
The odd part? The Google search page then asked me if that number was correct. I've never seen that before or since.
But I saw something different back then, because it had a little yes/no selection asking if I thought the information was accurate. I have not seen that question except for that one time.
And after I had done a lot of searches, Google asked me how accurate the information was. As I recall, it seemed fairly reasonable for that particular query.
OMG, what kind of sophisticated AI did they use to accomplish this? I'm thinking random forests mixed with neural networks and SVMs.