1. Matthew O'Riordan Staff

    Founder / Director / Co-founder at easyBacklog / Aqueduct / Econsultancy

    08 July 2004 12:56pm

    Matthew O'Riordan

    Followed a link from the MSN search site to find Microsoft's stement on their Commitment to Search. The point that caught my eye was:

    "A Personalized Experience. Your Search service should learn from you. What you like, what you read, where you live. Search should deliver results that are more personal and relevant to you. "

    So, have Microsoft actually implemented this, is anyone else currently planning to do this, and how exactly does this work anyway?

    I would assume that if I regularly search on "Glasses" meaning spectacles (as opposed to drinking glasses) and I click on the spectacles related links then MSN would begin to assume that when I search on anything ambigous or related to spectacles that is what I meant.  But how do they know that is what I mean that particular time, and therefore would my results for therefore be skewed sometimes against my wishes without my control or awareness?

    Does anyone know of any search engines that are promising this new "learning and personalised" search service?

  2. Mike Grehan

    Vice President, Global Content Director at Incisive Media

    08 July 2004 17:36pm

    Avatar-blank-50x50

    Hey Matt,

    On 12:56:25 8 July 2004 matt wrote:

    > Followed a link from the MSN search site to
    > find Microsoft's stement on their Commitment
    > to Search. The point that caught my eye was:

    > "A Personalized Experience. Your Search service
    > should learn from you. What you like, what
    > you read, where you live. Search should deliver
    > results that are more personal and relevant to you."

    > So, have Microsoft actually implemented this, is
    > anyone else currently planning to do this, and how
    > exactly does this work anyway?

    Personalised search is very much the holy grail which
    all of the search engines are seeking at the moment.
    And localisation comes as a subset to personalisation.

    Just as we've always thought of the search engine as
    "the black box" - the same applies in reverse - the
    search engine thinks of the end user as a black box.

    They simply don't have enough data about the user to
    be able to fully understand the exact nature or intent
    of the query.

    So, if they could get some feedback about the user and
    their searching habits, then they could provide much
    more relevant results. But personalisation is not just
    data about an individual, it's about data on an entire
    group. This is why social network analysis plays such an
    important role in search technology.

    Think about the way that Amazon personalises responses
    for you: "Other people who bought also bought..." or
    "We have recommendations for you..."

    And then think about the search engine being able to
    provide relevant results based not just on your own
    searching habits, but that of the entire group.


    > I would assume that if I regularly search on "Glasses"
    > meaning spectacles (as opposed to drinking glasses) and
    > I click on the spectacles related links then MSN would
    > begin to assume that when I search on anything ambiguous
    > or related to spectacles that is what I meant.  But how
    > do they know that is what I mean that particular time,
    > and therefore would my results for therefore be skewed
    > sometimes against my wishes without my control or
    > awareness?

    Just to widen the analogy a little bit here. If you have
    a mail account with MSN or Yahoo! they already have certain
    data about you. Your sex for instance. This means that the
    search engine could at first distinguish a query for, say,
    golfing apparel, as coming either from a male or female and
    be able to tailor the results a bit more to suit.

    Just knowing the gender is a step towards personalisation,
    as is knowing your age. And yes, for those ambiguous searches,
    if you had been looking at travel sites first, and then did
    a search for China - the search engine could make the correlation
    of travel and the continent of China.

    Whereas, if you had been looking at jewellery sites, or wedding
    venues and then did a search for china, the search engine may
    be able to make the correlation about weddings and china dishes.

    It's a very exciting field and relies heavily on having that
    necessary input from the user. As I've mentioned before, once
    personalisation and localisation i.e. where you are when you
    key in the query, kicks in: the notion of being a number one
    for a specific keyword or phrase will depend on who types the
    query and where and...

    > Does anyone know of any search engines that are promising
    > this new "learning and personalised" search service?

    The major search engines are all heading in the same direction.
    Yahoo! and MSN were ahead of Google in the race because they
    already had a user base of subscribers. So they have people
    logging in to send mail and read news, and check stocks etc.

    This is the reason that Google rushed out the Orkut social
    network and then swiftly followed with Gmail. Even Jeeves
    has acquired an online community and "sticky mail" property
    to pull all the components together in the same way as the
    majors.

    Personalisation is most certainly on the way from all of
    the search engines. But it has a way to go... but don't start
    holding your breath just yet!

    Cheers!

    Mike.

  3. Brendon Scott

    Senior SEO at Weboptimiser

    09 July 2004 15:42pm

    Brendon Scott

    Matt,

    >> Does anyone know of any search engines that are promising this new >>"learning and personalised" search service?

    The word you are looking for is "kaltix", and Google bought them last year.

    They have a tech demo up at Google Labs (labs.google.com), and although its not as advanced as the solution Mike has described, it's interesting to see the results change before your eyes

  4. Tristam Bielecki

    Partner Management at Talus Labs

    09 July 2004 20:22pm

    Avatar-blank-50x50

    MSN has been striving to provide this personalized experience for a long time with all their products - their grand vision for msn.com homepage and its localized version in each market is that it will use behavioral prompts in combination with gathered information that users volunteer.

    The big issue has and will continue to be one of privacy - having all users signed into their passport service allows MSN to collect data on users, but is it right for them to use this data for commercial gain? Passport is designed to be just an authentication tool that users trust - the trust is built from not using any data stored in the Passport servers for marketing or to sell on to 3rd parties. However MSN could use passport to anonymously identify a unique user and the CMS (Content Management System) can deliver content that is personalized for user x without ever revealing to the CMS who that user is.

    Another question - would you really want to go through the laborious step of signing into Google each time you did a search? No - this is why MSN will likely offer the personalized services to customers that sign in elsewhere (Hotmail for example) and are then driven to Search while still signed into Passport.

    What may happen is that the MSN CMS and Search Engine will offer searches based on a set criteria - just like the friendly paper clip in MS Office does when it recognizes you are writing a letter or some other standard document and then offers help in creating the document for you. Until MSN has critical mass of loyal (i.e. signed in) users on Search it would make sense to do this in the interim.

Reply to this thread

Log in to reply to this thread or join Econsultancy for free so you can post to our forums along with other benefits.