Summary - Google Searchology presentation 16th May 2007
Google’s Searchology™ press event offered a fascinating insight into ‘The mind of Google’ - past, present and future.
[ Watch the video here ]
To save you time: Here is a ‘jump menu’ / ‘timesheet’ to key parts of the event (the whole video is nearly 3 hours long).
The following text indexes this presentation, see below, marking key points on the time line in hours:minutes and seconds (hh:mm:ss) based on the Realplayer High video version. It summarises the live webcast of Google’s press conference held at its Mountain View, Calif. headquarters on Wednesday, May 16th, 9:30am – 12:30pm
Google highlighted the latest news on search innovations. Featured speakers included Marissa Mayer, VP of Search Products & User Experience; Udi Manber, VP of Engineering; and Craig Silverstein, Technology Director.
Topics covered include:
Ghost of Google Past
From 0:8:21 to 0:35:36
A talk by Craig Silverstein (Google’s first employee - Currently technology director)
Covering the early start-up days at Stanford which paints a picture of enthusiasm and resourcefulness of the founders - “We had three computers in the dorm room” • Finding their first premises • Key tenets that guided their priorities in the early days - these also hold-true today - comprehensiveness - relevance - speed and user experience • The problems faced in the early days e.g launch of GigaGoogle day in 2000 - 1 billion pages within the index (16:45ish in the video)
Ghost of Search Engine Present
From 0:35:36 to 0:57:00
A talk by Kerry Rodden (Google’s Search Usability Expert ;0) Ben Gomes (their Search Quality Czar) highlighting the various features in Google search and how users interact with them.
Including: Page Rank (0:37:48ish) • Relevance algorithms - page structure (0:38:50ish) • Spamming (0:39:15ish) • Usability- presenting the results (with a nifty video of an eye-tracker’s view of Google’s result page used to build an F shaped ‘heat map’ that shows the most popular portions of the screen) (0:40:50ish) • ‘Working out what a user wants’ e.g the difference between a user on the web versus on a mobile (0:46:50 ish) • Spelling correction and ‘concept correction’ (0:48:10ish) • Information seeking behaviours (0:51:18ish) • Query refinements (0:54:19ish) • Site links (small links below the main result directing users to portions of major sites -0:55:17 ish)
This ends with a final joke about finding tangible objects such as your car keys (0:56:47ish). This may not be so far fetched. I don’t know about you, but after a heavy research session on Google I have often felt foolish when I have subconsciously reached for an imaginary keyboard to find a letter sitting in my, real life, hard-as-metal, filing cabinet! With a little RFID magic, I could soon be using Google to actually find my car keys!
Interval
from 0:57:38 until 01:19:01
Some light music, which is later accompanied by a stream of search queries (kicking in at 01:08:40ish)
Ghost of Google Future - The Future of Search
From 01:21:30 until 02:20:19
This part of the talk is divided into two sections;
(i) A guarded peek ‘under the bonnet’ looking at the issues that the Google techies have been dealing with.
(ii) A User’s View of the new-look Google - Where most of the key announcements were made.
(i) A guarded peek ‘under the bonnet’ looking at the issues that the Google techies have been dealing with.
From 01:21:30 until 01:50:40
Udi Manbe (or is it Mander??), (vice president of Engineering and Head of Search at Google) chooses his words carefully as he pulls the curtain aside and shines a light into the back-stage area of Google. This behind-the-scenes view does not give much away, but it is still interesting.
Topics / Themes :
20-25% of searches that Google deals with have never been seen before. You realise, when you are dealing with ‘once in a lifetime queries’, that Search is Very Hard • Search is about ‘people’ not just logic • Its about understanding; understanding users, understanding queries, understanding languages (1:40:40ish), understanding content (01:47:19), understanding locales, understanding tasks.
1 - ‘Cross-Language Information Retrieval’ or CLIR (01:43:21 ish).
A very big shake up is on its way!
We will, soon, be able to search in one language and see results that were previously ‘not relevant’ due to language barriers. E.g A user submits a query in their own language, say Arabic –> Google translates that query, to English –> Search the English web –> Translates the results to Arabic and presents them, translating the destination pages.
I guess there are few things of special note regarding this:
- Rankings will change dramatically as a previously high ranking website will now have competition from more sites - and a website that was previously not ranked in a specific locale will now attract more foriegn language visitors
- The ’symantic web’ and your web data structures becomes more important - the more help you can give to enable Google to translate your site properly the more traffic you will receive
It opens the web, universally to the whole World
Regarding ‘understanding content’ (01:47:19). Udi explains the problems of comparing say, Apples and Oranges. e.g in a search for “fruit” - which comes higher? Are apples more important because more fresh apples are eaten than oranges? etc.
Udi then explained that the problem doesn’t end there - they not only have to compare different ‘concepts’, but also different media types (01:49:43)
In short - its hard - don’t even think about competing with Google - they are way ahead!
“This is a very large, multi-dimensional, not-well-defined problem - That’s our strength” - He smiles and then goes on to introduce Marissa…
(ii) A User’s View of the new-look Google - Where most of the key announcements were made.
From 01:50:59 until 02:20:19
Here come the biggies!
First, Marissa talks of her first day at Google where she had her first ‘tour of the source code’ (01:51:40).
She builds up to the first big announcement (01:53:55) highlighting the number of genres of information /media types that are online these days. She doesn’t want people to have ‘the room mate problem’ (01:54:20) e.g Users should not ask themselves “If I was Google, where would I have put the images?”
2 - Universal Search (1:55:03)
In essence the results page will change to integrate all the different media types into the new Universal Search results:
- Good ol fashioned web URLs
- Plus, Books (02:00:24)
- Plus , Local Search (01:57:54)
- Plus, images
- Plus, News
- Plus Video
- in-page full length movies (01:58:52),
- highly integrated YouTube i.e Martin Luther King speech - (02:01:30) or even How-to clips (02:02:22) or a far more personal type of query, which has an effect on the ‘long tail’, such as a clip of your friend’s baby’s first burp. (02:04:28)
She then (02:06:12) gives a brief outline of the key hurdles that Google faced to make this possible:
- Gathering Results (02:06:53)
The old infrastructure had to be completely re-designed (over a few years) to make it more computationally feasable - Ranking Results (02:07:38)
Apples verse Oranges - they had to design a completely new ’scoring function’. - Displaying Results (02:08:06)
A lot of research went into answering questions posed about displaying the results. Should they clump the results by media type? A new vernacular had to be created to describe these prototypes: interleaved, fixed placement, segmented etc. Resulting in the tricky job of offering users what they expect (02:08:44) - An ordered list of results
…and in the future… more corpora will be added, different user interfaces will be tried, and as always, things will keep on improving.
3 - Contextual Navigation Links at the top of the page (02:09:20)
If the wealth of media types becomes too much, and you know that you just want to search news, or videos, or web pages etc. you will be able to focus on a particular corpus and drill down through the results. Google will present the most relevant ‘corpus links’ based on the search query - with links to all the other corpi close by.
4 - Universal Navigation Bar (02:11:15)
This gives easy access to the multitude of services that google now offers. Instead of have to click 8 times to get to gmail from the home page - the universal navigation bar will present links that are most closely related to the task in hand. It will, now, show the nearest neighbours to your current activities.
5 - Google Experimental (02:13:17)
The idea is to open up a new channel for Google to interact with the users. You can now sign-up, opt-in and start using Googles experimental search ‘properties’ as your regular search tool.
Example experiments include:
- Views (02:14:30)
Resulting from the New York team’s work on ‘Data Extraction’ - looking at ways of getting the most out of a webpage. E.g See your results in Timeline or Map View - KeyBoard shortcuts (02:17:45) & ‘Left-Hand-Right-Hand Navigation’ - let them know which one you prefer
A quick final summary (02:18:18)
“We think these features, ultimately, bring us one step closer to realising Google’s mission - which is to organise the World’s information and make it universally accessible and useful”
Questions and Answer Session
From 02:20:19 until the end
(A rough - gist of the questions…)
Q1a (02:20:19): There seems to be a lot of scrolling - discuss
Q2:(02:21:36): Re: Experimental - do you think the user’s awarenes that they are taking part in an experiment will affect the integrity of the results?
Q3 (02:22:19) : How muchdo these feature rely on personalisation and has google thought about the ethics of this - e.g with regard to China / Privacy
Q4 (02:23:44)
(a) What’s the difference between Google Experimental and Google Labs?
(b) How does the innovation process work at Google - is it still chaos?
Q5 (02:26:15): With so much emphasis on speed, how much quality is being sacrificed when a search result is returned?
Q6 (02:28:26)
(a) How did you develop your new infrastructure?
(b) At what point did you realise that you needed to completely restructure it?
(c) By how much of a factor did you manage to improve (fine-tune) the infrastructure [Google -could answer this but were not willing to say ;o) ]
Q7 (02:30:51): Are you thinking about incorporating new forms of advertising / rich media advertising into the Universal Search Results
Q8 (02:31:44)
(a) Copyright issues - discuss
(b) How are the videos going to work - streaming from their site?
Q9 (02:33:45):
(a) When is it going live?
(b) Is this one of the first big pay-offs from the acquisition of YouTube?
Q10 (02:33:08):
(a) How broadly are you going to include video in terms of other video sites like Media Cafe?
(b) its unclear whether its opt-in or if you’re going to search any site that has publicly available video - discuss
Q11 (02:37:14)
People used to spend a couple of seconds on Google before going off to another website. this seems like a move to draw people in to spend more time ‘inside Google’, how affec do you think tihs will have on search results, search qualty and advertising revenues.
Q12 (02:39:18)
When will you be applying abstraction to local search, e.g a search for “Seattle hotel with swimming pool“?
Q13 (02:40:10):
Re: Gmail being 8 clicks from the home page. In the past you have launched many ‘personalised home page’ aspects of Google. What kind of take up have you had of these or is this an alternative strategy?
THEN, YOU WILL NOT BELIEVE THE NEXT ANNOUNCEMENT!!….
(02:41:10) Lunch was announced.
;o)
Related pages:
Alternative links to the video are here, split into 3 parts: part 1, part 2, part 3