Danny Sullivan is moderating this session. Jonthan Lelang from A9.com starts off the session, by giving some examples of how A9.com works. He shows a screenshot of A9, and describes that they have toolbar. He says one of the unique things about A9 is that they have multiple columns and you can check off which choices you would like to view. A9 includes a search history, enabling visitors to see when they visited the result such �visited 9 months ago�. Users have the ability to enable search history or turn it off. It appears you need the A9 toolbar to have this option available to you. Unique. So what is history useful for? How do you find new things based on where you have been? It can provide you with options for other sites based on what you searched for last. They are building on a lot of technology with Amazon.com, that providing recommendations and suggestions goes over well. They also wanted to enable the user the ability to make their interface and personalize their results. He gives the example of searching for the space program. By personalizing the search, you can eliminate things you are not interested in such as astrology when you really want astronomy. A9 lets you build in custom searches into the toolbar and take it anywhere. Another interesting feature they have is something called the A9 diary. You can add comments about the website. Anytime you go back to visit these sticky notes are added to the top or tac it to the top for later retrieval.
Tim Mayer was up next from Yahoo. He says personalization has always been the holy grail for search engines. They have failed in some aspects in doing this. One the obstacles with personalization is when a session begins and when a session ends. When do you want to search for something different. Other issues are transient interest and genuine interest. Such as buying a country CD for a Christmas gift as opposed to you liking country.
He says they want to enable people to find, use, share, and expand all human knowledge. They wan they to use the search not for the sake of searching byt to acheieve a purpose. Sharing knowledge with people you connect with and connecting to people who you share knowledge with. He says they look at the web in three acts. Act 1 is the public web search, Act 2 is personal such as desktop search. Finally, Act 3 is the shared web such as search communities. He mentioned Yahoo My Web 2.0 in that people can share their favorites with others. He gives an example of My Web 2.0, you can search on your web or search the general web. They don�t track you unless you opt for it. He gives more examples of searches using My Web. My Web makes it rather easy to get back to info you have saved. In the default web experience its pretty basic, in the future you will get some good My Web results. The other good thing about using the product is that you don�t need to use the search to do this, you can use the bookmark features.
His next example is about Yahoo Mindset! Which came from Yahoo research labs. It will determine between whether you are shopping or researching. You can use a sliding bar to determine what type of results you can see. He ends by saying to visit the Yahoo blog for more information.
Ask Jeeves was up next with Jim Rainey. He explains background info about Ask Jeeves with lots of numbers. In my experience they like to do this before getting into the subject. He asks to start, What is personalized search? It is using various personal data to improve the quality of keyword driven search results or searchable repositories of personally relevant data. A breakdown as follows. Keywords are already pretty specific. Word polysemy is a minority of queries, such as the difference between jaguar the car, and jaguar the cat. There is also the amount of choices in the search results. Getting it wrong can make results worse as well, which is not what they want.
The challenges for personalized search is getting data from users (registration? long profiles?). Doing the right things with data and keeping the user interface simple. Like do people really want sliders? He says that search has a long way to go. One of the big areas that personalization will be big is in personal data stores. Such as email, chat text, files and documents, and so on. He explains they launched a desktop search, my jeeves which keeps a search history, search results, and bookmarks. Bloglines and using RSS feeds. He concludes with talking about how to save a search or result on Ask Jeeves.
Eurekster was up next with Grant Ryan speaking about what they are doing for personalized results. The starts with a discussion about the evolution of media and using organization and personalization. He talks about targeted search engines and channels. There are many different types of targeted search engines. Search focused on shopping, travel, jobs, and other verticals for example. He shows an example from a gaming website, in that they setup personalized search for. There are natural extensions to web personal publishing. Eureksters Search Publisher is powered by the publishers users. The technology harnesses the collective wisdom of the publishers community. Communities help each other and finding a community of like-minded individuals is a valuable tool.
Marissa Mayer from Google is the last speaker. She says that she will explain where Google is today on using personalized search. Her first slide is a picture of a sample searcher who is using Google�s personalized search. They want it to really easy for the user. Personalized search is built into Google and it is on by default. Users have a lot of control over the search history and how they navigate to standard results. There is very little work required for people to use personalized search such as history, clustered related history over time. They also provide an activity counter letting you know when you where most active. They also recently rolled out the personalized homepage. The first set of the homepage only allowed 12 modules. Last couple weeks they changed this to allow RSS feeds and bookmarks. You can also customize news such as adding your favorite paper to the search. She ends with how Google has gone about unifying all the personalized possibilities at Google.
Q&A
Q: What is the adoption rate of this personalized search? A: We don�t give exact numbers, but they have had been satisfied. The people that seems to use the personalized search are the most loyal users.
Q: So when Google first launched personalized results, it used checkboxes, and then it switched to a implicit method. A: Its nice to empower the user in what they are interested in. She gives an example of Tivo, it will record what you want, but also guess at what else you might be interested in.
Q: Is there effect on search rankings with people using bookmarks? A: It�s possible, and its going to be useful. Will use balances and checks, and compared with link analysis it provides usefulness about seeing those that come back and visit often. Yahoo says that they are possibly exploring this further. A9 was the only one to admit they are actively using bookmarks to affect the natural results. Danny says it�s worth clarifying that its difficult to determine if someone has bookmarked the site. Its goes further on knowing what people have clicked on. He says it�s prone to spamming betting on clicks. The mechanisms today are better at determining what you may be doing to mess with it. Eurekster says they have come up with interesting ways to determine spammers, such as looking for patterns and how long they stay at a result. Overall the panelists like bookmarking as an indicator of quality.
Q: Jim mentioned sliders? In that earlier cars that you needed to be a mechanic to use them. He draws a comparison between a stick shift and an automatic. He asks whether people want sliders and do they have a future? A: Yahoo is letting people play with it. They aren�t planning to launch it yet because you can�t be 30% shopping and 70% researching. It doesn�t make sense. Google seconds that, it doesn�t make sense to implement that yet. Marissa says people didn�t like looking half personalized half regular results. Its an either or issue.
Q: Are they using demographics? A: Google hasn�t used a lot of demographics. They found the users needs vary a lot. They do some zip code stuff such as weather. Danny explains that country of origin is better demographic targeting, such that if you search in the UK it will be a little different than the US.
Q: Different personas or personalities, how do you handle the different interests or areas people may switch back and forth with interest in. A: It�s been difficult to turn one off and one on. They are looking to blend them. People are good at managing this. Google says they would rather you be just one person because logging in and out between different id�s is not fun. She says that maybe 3-5% of results may change based on your preferences. They have also looked at changing things at times of the day.
