These are notes from the “Recent Innovations in Search and Other Ways of Finding Information” panel at BayCHI. It was an absolutely fantastic panel! Very well done, the panelists were great all around (although personally I think I liked the contributions from Udi the most). I’ll have to come back at some points and clean up these notes, they’re in kinda rough shape.

Panelists:

  • Peter Norvig

  • Mark Fletcher

  • Udi Manber

  • Ken Norton

  • Jakob Nielsen

5 minute show and tell by each panelist

Peter - talked about question and answer “what is population of japan” was the example

extracted facts from the web

facts are represented many times, accuracy is a problem

then search by sms - people are using more andmore mobile

maps on the cell phone

then google maps overall - plus ssatelite photos

niagra falls, white horse

predictive seach - google suggest

talked about AJAX in particular - google has been pushing it Jesse Garrett

Google desktop search - example combining both desktop and the web

Ken - Enable people to find, use, share and expand all human knowledge

most people addressing only the find part, yahoo wants to do all

remixing and remaking

desktop/creative commons, YDN, big list of things they did

MyYahooSearch - in beta, released soon gives users more control, allows to save and share, like bookmarking but hopefully a step beyond.

video search - huge ammount of content out there, huge demand, that’s new

mediarss - allows people to submit their content back after remixing

desktop search

Y!Q - bringing search to the point of inspiration - there is context associated with any search, it starts somewhere. Can we bring the search process back into each of those tasks and make the search aware of it example is reading about a court decision, get news related

mobile search - multiple modes of access

CC search - demand from users to find content - licensed for reuse.

flickr - photosharing and community - they will bring in the DNA thinking about how users tag and share and expand on content.

what’s next - personal search, contextual seasrch, multimedia search, desktop search, local search, travel search, social media

next.yahoo.com

Mark - did a walkthrough of bloglines

search/subscribe/share/publish

started bloglines to scratch his own itch, hundred sites a day to visit. information overload, blogcontent is the fastest growing area. 1.6 million articles every day.

Mark has 29 feeds in his own account (at least the one he demoed with)

search - keyword search , if you hit the save button you can subscribe to the query

the service is working to cull the information when you create feeds just for the stuff yo uwant to see (search for bloglines in posts instead of reading all posts to find the ones that mention bloglines)

what’s coming: search friends, filter articles

Udi- Yello pages - launched in January - walk down the street, showed actually moving down the street

“it’s this place right next to that place on main street” - now you can find it, find the name, and then search for the info that you want.

showed the rental suv with the camera on top that took the pictures, gps, laptop. The driver drives and the system does everything else.

Udi spent two days doing it himself in arizona, it was a lot of fun

Jakob- showed a table showing that since 1994 we have almost a 100% change in usage people used to use one word in 1994, and they use 2 words in 2004

probably a continuing growth curve, up to a point

how many people are successful: 42% for all searches low experience 32% and high experience 50% internal search on website 33% and external search 56%

search on websites is a miserable failure. it’s a disgrace that we’ve allowed search on websites to be so bad, intranets are even worse. it’s the most prevalent problem actually

showed a video of a recorded session of someone trying to search, but entered the search text into the wrong field. and then going to search to try to find out what was up. It was AOL search. then she tries to type it into the location bar, and tries www.head.ache.com and it’s not found gets an error page and it gives a search option, she searches there and ends up at msn search

looking at problems across all sites, 11% of them were related to search, biggest percentage of problem

first question is to ask the other pannelists to reply to “search on websites is a miserable failure”

udi says that it’s not a failure, 50% is good. It’s a hard problem. when you go to small sites, where the people do not have the resources of an outfit like AOL they just can’t do it.

mark - there’s a lot of work to be done, but it’s not just search. It was basicproblems like how the browser is laid out.

Peter - the same thing happened to me with the IBM site when he was looking for a thinkpad. And he went to this ‘experimental’ site google, and typed in what he wanted and it was there. Google was more powerfull than IBM. That means somethiing.

Ken - more and more people are using the general internet search with site restricts to do their search. Before you do it on your own, think about using one of the public engines to do it for you

How have searchers evolved besides more words?

Ken - demand for better search tools in vertical areas. People used to be happy with general queries in the search engine, now they want the context to be appropriate to their task. Yahoo is looking at the different categories that users seearch in and trying to figure out what goes in those different areas. How do you make the user aware of it once you ahve it. The results page is a poor place to message to users about features and functionality.

Peter - second that, veritcal is very important. growth in international really stands out. Majority is outside of US, ,not in english.

Mark - others from Ask told him - their query length is getting shorter. people realize it’s not JUST natural language at Ask. Tail has grown. Majority of queries are completelly unique. Expectations have grown, searches are successful, people search more, feedback loop.

Udi - the size of the search box actually influences the length of the search terms. also thinks veritcal is imporatnt. Opensearch, syndicated search for opening search and sharing it. Anybody can take search results and put it on A9.

Jakob - People use search cause they are selfish. they don’t want to suffer theinterfaces that sites give them, that applies to both site search and general search engines. If sites did use more standard and humane stuff, site map, richer navigation, that might not be so. This has to be standardized navigation, for example site maps. It has to be a standard widget, not a one off trick

How do you give access to the different verticfals without turning into a portal?

Ken - the web search box IS the interface. that’s what people want. Don’t direct people to different areas, just figure out the conntext and provide them with it in place. Rather than asking users to adapt ask the search engine to be smarter. Is this a lcoal query? Find the users intent. Users have gotten confident. That used to be the greatest problem, but it’s not the case any more. They expect good results.

Udi - Some queries work well and others dont. “what’s the population of india” works even without any natural language. “Waht’s the best TV to buy” does not. How do we balance ease of use and powerful tools?

Peter - we’ve failed in terms of letting the user to have a dialog. search is a command line without a documented interface. How do we put advice in place to help guide users to what they want. We’re getting there but it’s nothing like walking into a library and getting help from the librarian.

Asked how many people have read advanced help, show of hands - lots had, atypical crowd

Are folksonomy/tagging a fundamentally different set of data than what is there?

Udi - it’s all just metadata, it’s our job to deal with it. it’s not any different, the hope is that people will provide much data

Ken - it’s remarkable in that it’s just like what we already deal with. Right now tagging is done by web authors, and now tagging we can collect from a larger group of people. The more metadata the better. The more we can get from users and not authors it should make it better info.

Jakob - we won’t get much mroe info from users. with taggintg hte barrier is lower, but it’s still there. we have to get passive data out of people, eyeball time and metrics from the desktop

Ken - if they find value they’ll do it.

jakob - okay, but it has to be more inline with the work they do already, like spam marking in collaborative filters

Udi - mapr maps flickr to a map view. lots of images in Maine, cause people tag with “me” (as in myself), but mapr thinks it’s a location.

there are questions from the audience now. And these are just the cool bits

Udi - a9 keeps context. If you do a search twice, a few months apart. the second time you get a message that says “these results are new, you haven’t seen them before”, and “these are the items you clicked on last time you did it”

Someone asked when people don’t use search, for what kinds of queries does a user typically not revert to the search interface? This answer from Jakob is great:

Jakob - people do not search when they’re loyal customers of someone. If you’re a loyal customer of Amazon you don’t search for books on Google, you search on Amazon.

Someone asked if the search wars were going to be defined in terms of branding and not tech. All seemed to think it was still a tech race. In that regard the industry is still evolving. People have thought the search battle was over previously, but that wasn’t correct. It’s not correct now, there is a lot more evolution to go.