A review from the Wall Street Journal:
online.wsj.com/article/SB120648944687663981.html
One of the company sites in the review:
http://www.searchme.com/wsj
Company news:
http://www.news.com/8301-10784_3-9890911-7.html?tag=nefd.top
My initial reaction to this article was that the author must have a lot of time on her hands - I mean, who cares about a better search engine any more? Google wins, end of story.
But, because I have spent so much time with Google in the past few months, I read on.
The article chronicles the experience the author had with the products of two new companies: SearchMe and SpaceTime. I was reworded with an interesting introduction to a new take on the Graphical User Interface.
SearchMe, is a startup that is trying to innovate in a space dominated by Google and coveted by Microsoft. What can a company with 35 employees do in the land of giants?
Well, as it turns out, a lot. It can change the way you view searches. We are all very used to the list view that pumps our of searches - "enhanced" by ads on the side and lots of related links, maps, and pictures, the result is still a list. SearchMe returns a collection of screen shots of related sites that float across the page. This is a lot like the album art that Apple uses with iTunes.
Search results as floating pages? I think this is one of those things that must be tried to be understood. So I did. And, I do. It is amazing to learn how much you can grasp from seeing a page, even in a small, distorted fashion. I found myself flipping though the results and clicking away - in a fashion that I never would in my old friend Google.
In truth, the search results are not very good (hopefully this is because of the early nature of the technology, but the presentation is very refreshing. And, the refreshment has been sufficient to raise $62m in venture funding over the last couple of years.
10 April 2008
Weekly Recommender Log
This science project is over - I can say with little fear of rebuttal that iGoogle's Recommender tab is severally broken. Somebody needs to turn out the lights.
After 200 reviews, and a month to think it over, any recommender that has a CPU should be able to do something.
Alas, iGoogle's Recommender tab does practically nothing. When I login this week, I was presented with the same sad list of items from last week, including my overwhelmingly negative ratings.
In disbelief, I restarted and with some relief, found that I once again had recommendations in only 1 of 6 categories. And, in that one category - "Pages", I had another set of 50 sad recommendations.
Once again, I paged though, hoping to find some merit in this service. Alas, there is none.
What I have learned is that given a random set of 50 url's I will probably like 1 or 2.
I will need a new plan for next week.
After 200 reviews, and a month to think it over, any recommender that has a CPU should be able to do something.
Alas, iGoogle's Recommender tab does practically nothing. When I login this week, I was presented with the same sad list of items from last week, including my overwhelmingly negative ratings.
In disbelief, I restarted and with some relief, found that I once again had recommendations in only 1 of 6 categories. And, in that one category - "Pages", I had another set of 50 sad recommendations.
Once again, I paged though, hoping to find some merit in this service. Alas, there is none.
What I have learned is that given a random set of 50 url's I will probably like 1 or 2.
I will need a new plan for next week.
03 April 2008
Personalized Recommendations in the Wall Street Journal
http://online.wsj.com/article/SB120647589356663209.html
I always find it noteworthy when leading edge technology is written up in the popular press. This week the Wall Street Journal published a good article about Personalization and Recommendation technology.
The reporter focused third party suppliers that can help sites boost profits - sometimes by working together.
Four providers of recommender technology were highlighted. Here are the links:
http://www.ratepoint.com/
http://www.aggregateknowledge.com/about.html
http://www.optitex.com/index.php
http://www.webcominc.com/
I always find it noteworthy when leading edge technology is written up in the popular press. This week the Wall Street Journal published a good article about Personalization and Recommendation technology.
The reporter focused third party suppliers that can help sites boost profits - sometimes by working together.
Four providers of recommender technology were highlighted. Here are the links:
http://www.ratepoint.com/
http://www.aggregateknowledge.com/about.html
http://www.optitex.com/index.php
http://www.webcominc.com/
Weekly Recommender Log
Well, it's another week of disappointments with the iGoogle Recommendations tab. Or should I say, just more of the same.
Once again I have 50 sad "Recommended Pages" which shows no relationship to my preferences or persona. Once, again (and the last time if results are not improved) I tried and rated all 50 pages. Here's a bad sign - I have always rated new items as negative, but more than 1/3 recommended pages are news items. At the same time, the box where I am supposed to get news recommendations is always empty.
And, just like last week, I have yet to receive any recommendations in the boxes labeled "Recommended Videos", ""Recommended News", ""Recommended Searches", or ""Recommended Groups". Somebody is asleep at the wheel.
So, here's my new plan. In the coming week I a considerable amount of Google searches on all things automotive. Let's see if the Recommendations tab takes notice.
Once again I have 50 sad "Recommended Pages" which shows no relationship to my preferences or persona. Once, again (and the last time if results are not improved) I tried and rated all 50 pages. Here's a bad sign - I have always rated new items as negative, but more than 1/3 recommended pages are news items. At the same time, the box where I am supposed to get news recommendations is always empty.
And, just like last week, I have yet to receive any recommendations in the boxes labeled "Recommended Videos", ""Recommended News", ""Recommended Searches", or ""Recommended Groups". Somebody is asleep at the wheel.
So, here's my new plan. In the coming week I a considerable amount of Google searches on all things automotive. Let's see if the Recommendations tab takes notice.
27 March 2008
Weekly Recommender Log
iGoogle continues to mystify me. After last week's rating spree - 50 ratings in the "Recommended Pages" box of the "Recommendations" tab of iGoogle - I login today and find 50 more sad recommendations.
Sad, is a conservative descriptor. After everything that I have told iGoogle about me, they can't even come close to matching me up with web content! As evidence of my plight, I rated all 50 of the new "Recommended Pages", only 6 got a thumbs up and I was generous.
On a positive note, I finally have a "Recommended Gadget". On the down side, it is "Google News". I guess it's ok, if a bit short sighted, for Google to recommend Google. I tried to rate it but the box didn't let me register a rating. Sad.
To round out my experience, I have yet to receive any recommendations in the boxes labeled "Recommended Videos", ""Recommended News", ""Recommended Searches", or ""Recommended Groups". Somebody is asleep at the wheel.
Sad, is a conservative descriptor. After everything that I have told iGoogle about me, they can't even come close to matching me up with web content! As evidence of my plight, I rated all 50 of the new "Recommended Pages", only 6 got a thumbs up and I was generous.
On a positive note, I finally have a "Recommended Gadget". On the down side, it is "Google News". I guess it's ok, if a bit short sighted, for Google to recommend Google. I tried to rate it but the box didn't let me register a rating. Sad.
To round out my experience, I have yet to receive any recommendations in the boxes labeled "Recommended Videos", ""Recommended News", ""Recommended Searches", or ""Recommended Groups". Somebody is asleep at the wheel.
What's Missing From Pandora - Mood
This week I thought I would offer up an unconventional resource to the class - a conversation with an expert.
Last week in class we had a very interesting discussion about the music recommender, Pandora. I was motivated to contact one of my old friends who is an expert in music, radio, and computing.
Me:
Friend, Initial Response:
"Well, I played with Pandora some.
As a former DJ and musician, I think I have a pretty good idea of how this should work.
It seems to match eras more than sounds. It put Tired of Waiting by the Kinks with The Who. Its a VERY different sound and style. I don't get that. Also put 19th nervous breakdown with the Who. No again. It also put Tom Petty in here. Close but not really.
Also put Bos Skaggs with Steely Dan. NO. Steely Dan is jazz. Bos Skaggs has horns but he is not jazz. Reggae?
It did better on Miles Davis. Art Blakely and the Jazz Messengers, etc.
I'll play with it some more. This is a massive undertaking, creating formats on the fly. Certainly interesting."
Friend, Insights Begin:
"I think that I have some insight into what I think is missing from Pandora. They are arranging these songs by objective measurements (male vocal/reggae beat, etc.) but do not have anything subjective in them.
It would be possible for them to group a song with revolutionary content (Street Fighting Man, Volunteers, etc.) with a pacifists song (Bargain, Revolution (Beatles)) based on these measurements.
I think of music in more subjective terms: what feelings, thoughts, and other experiential dimensions does the song evoke? What state changes occur within the listener? What is the objective of listening to this song (calming, introspection, resolution, seduction, whatever.) and the type of artistic treatment. What is the INTENSITY? Joni Mitchell is emotionally INTENSE but she is folk/jazz/rock. This measurement is very important in this category and is probably important in others.
These other attributes may get you in the neighborhood but they don't really achieve the objective for me at least, of meaningful picks. I spent many years selecting music. In broadcasting, you try to achieve a flow or transitions up and back down. Too much of one type fatigues listeners so the music moves up and down in waves.
I designed a computer program to chose music based on some of these criteria once. It was based on what I thought mattered overlaid onto what was being done in radio at top stations at the time. Radio since then has gone backwards instead of advancing, with the exception of what is going on in satellite radio......"
Me:
"Most interesting. One member of our class has used Pandora as part of her class work but, is not very impressed. She said things that showed that she was more concerned about the emotional side of music. She often used the word 'mood' in her descriptions."
Friend, The Magic Word - Mood:
"MOOD is very important in radio.
I was lucky to have gotten to work with one of the best program directors in the history of radio. The whole thing is about mood and subconscious/unconscious messages.
We were sensitive to things you wouldn't believe and have exact phrases that we followed. Some songs are allowed only in the evenings, some are banned from the mornings. It is all about matching and leading moods."
Last week in class we had a very interesting discussion about the music recommender, Pandora. I was motivated to contact one of my old friends who is an expert in music, radio, and computing.
Me:
"Some members of my class have had some happiness using a music recommender that you might have some fun with; it's called Pandora. It started out life as a project to develop a DNA for music. It is pretty well described at en.wikipedia.org/wiki/Pandora_(music_service) but the stuff I thought you might find interesting is the list of attributes that Pandora employees assign to each piece of music. The list is at en.wikipedia.org/wiki/List_of_Music_Genome_Project_attributes."
Friend, Initial Response:
"Well, I played with Pandora some.
As a former DJ and musician, I think I have a pretty good idea of how this should work.
It seems to match eras more than sounds. It put Tired of Waiting by the Kinks with The Who. Its a VERY different sound and style. I don't get that. Also put 19th nervous breakdown with the Who. No again. It also put Tom Petty in here. Close but not really.
Also put Bos Skaggs with Steely Dan. NO. Steely Dan is jazz. Bos Skaggs has horns but he is not jazz. Reggae?
It did better on Miles Davis. Art Blakely and the Jazz Messengers, etc.
I'll play with it some more. This is a massive undertaking, creating formats on the fly. Certainly interesting."
Friend, Insights Begin:
"I think that I have some insight into what I think is missing from Pandora. They are arranging these songs by objective measurements (male vocal/reggae beat, etc.) but do not have anything subjective in them.
It would be possible for them to group a song with revolutionary content (Street Fighting Man, Volunteers, etc.) with a pacifists song (Bargain, Revolution (Beatles)) based on these measurements.
I think of music in more subjective terms: what feelings, thoughts, and other experiential dimensions does the song evoke? What state changes occur within the listener? What is the objective of listening to this song (calming, introspection, resolution, seduction, whatever.) and the type of artistic treatment. What is the INTENSITY? Joni Mitchell is emotionally INTENSE but she is folk/jazz/rock. This measurement is very important in this category and is probably important in others.
These other attributes may get you in the neighborhood but they don't really achieve the objective for me at least, of meaningful picks. I spent many years selecting music. In broadcasting, you try to achieve a flow or transitions up and back down. Too much of one type fatigues listeners so the music moves up and down in waves.
I designed a computer program to chose music based on some of these criteria once. It was based on what I thought mattered overlaid onto what was being done in radio at top stations at the time. Radio since then has gone backwards instead of advancing, with the exception of what is going on in satellite radio......"
Me:
"Most interesting. One member of our class has used Pandora as part of her class work but, is not very impressed. She said things that showed that she was more concerned about the emotional side of music. She often used the word 'mood' in her descriptions."
Friend, The Magic Word - Mood:
"MOOD is very important in radio.
I was lucky to have gotten to work with one of the best program directors in the history of radio. The whole thing is about mood and subconscious/unconscious messages.
We were sensitive to things you wouldn't believe and have exact phrases that we followed. Some songs are allowed only in the evenings, some are banned from the mornings. It is all about matching and leading moods."
20 March 2008
On the Web, Signs of a Click Recession
Google has been in the new a lot recently, first of all:
http://online.wsj.com/article/SB120406885089594779.html
The Wall Street Journal reports that "Internet advertising may be showing itself more vulnerable to a consumer slowdown than many in the industry had hoped". This is based on a new report from comScore, a research company - their report was not available, at least without a fee.
ComScore reports a "7% decline in the number of times U.S. consumers clicked on ads appearing alongside Google's search results in January compared with December; clicks were 0.3% lower compared with January 2007. That follows a 7% decline from November to December." Others, including Google, see these numbers as being very way to negative. Regardless, the Google stock price was reported in the same article as "trading 38% lower than its 52-week intraday high".
Other recent articles:
Google Closes DoubleClick Deal After EU Approves
online.wsj.com/article/SB120524277833327159.html
Google to Unveil A New Ad Service For Web Publishers
online.wsj.com/article/SB120537498814032575.html
Pleasing Google's Tech-Savvy Staff
online.wsj.com/article/SB120578961450043169.html
http://online.wsj.com/article/SB120406885089594779.html
The Wall Street Journal reports that "Internet advertising may be showing itself more vulnerable to a consumer slowdown than many in the industry had hoped". This is based on a new report from comScore, a research company - their report was not available, at least without a fee.
ComScore reports a "7% decline in the number of times U.S. consumers clicked on ads appearing alongside Google's search results in January compared with December; clicks were 0.3% lower compared with January 2007. That follows a 7% decline from November to December." Others, including Google, see these numbers as being very way to negative. Regardless, the Google stock price was reported in the same article as "trading 38% lower than its 52-week intraday high".
Other recent articles:
Google Closes DoubleClick Deal After EU Approves
online.wsj.com/article/SB120524277833327159.html
Google to Unveil A New Ad Service For Web Publishers
online.wsj.com/article/SB120537498814032575.html
Pleasing Google's Tech-Savvy Staff
online.wsj.com/article/SB120578961450043169.html
Weekly Recommender Log
This week I continue my exploration of the iGoogle Recommendations Tab. And still it puzzles me.
I am told that the 6 categories of recommendations on this page - all in their own little content boxes - are updated every day, that's right, every day. This sounds strange to me. Google has built a huge company by tossing out recommendations right and left. What could they possibly be doing that takes a day? I am not all that impatient, it's just that this seems so much different than their normal business practice.
And then there is the almost total lack of recommendations. That's a little strong but I only have recommendations in one of the six boxes. And, one of the boxes continually shows an error message.
So, I am left with one of six to work with. It is the Recommended Pages box. In it I find a link and heading for 50 recommended pages, each with a opportunity to vote the link up or down. As a whole, it is a pretty lame list. I don't see how more than a handful could have been based on my search history or my iGoogle profile.
But, in the interest of forwarding the technology, I dutifully click though all 50 recommendations and give them a thumbs up or down (mostly down unless it matches some part of my online persona).
I can't wait to see what the next round brings my way. Let's hope that it won't recommend anymore weather sites for Dayton Ohio.
I am told that the 6 categories of recommendations on this page - all in their own little content boxes - are updated every day, that's right, every day. This sounds strange to me. Google has built a huge company by tossing out recommendations right and left. What could they possibly be doing that takes a day? I am not all that impatient, it's just that this seems so much different than their normal business practice.
And then there is the almost total lack of recommendations. That's a little strong but I only have recommendations in one of the six boxes. And, one of the boxes continually shows an error message.
So, I am left with one of six to work with. It is the Recommended Pages box. In it I find a link and heading for 50 recommended pages, each with a opportunity to vote the link up or down. As a whole, it is a pretty lame list. I don't see how more than a handful could have been based on my search history or my iGoogle profile.
But, in the interest of forwarding the technology, I dutifully click though all 50 recommendations and give them a thumbs up or down (mostly down unless it matches some part of my online persona).
I can't wait to see what the next round brings my way. Let's hope that it won't recommend anymore weather sites for Dayton Ohio.
21 February 2008
Weekly Recomender Log
I start the this week exploring the world through the eyes of Google.
Last week I added a new Interesting Items For You gadget to my Humor tab of iGoogle. When I logged in this week, I still didn't have anything recommended but it suggested that I click to setup a new tab just for recommendations - just for me. I felt honored, so I clicked.
I now am the proud owner of a new iGoogle tab called Recommendations. It plans on finding items of interest for me in the following areas:
I start off the session with only 37 searches and the goal of seeing when the recommender will kick in. I start searching for skin care information. I click a few adds. I refine my search using Google's recommended searches. I Click an ad that takes me to Science Daily's skin care news, which happens to feature several Google ads prominently displayed in the middle of news stories. I click on one of these ads a learn about the science of skin care.
When I make it back to Google and, once again, search for skin care, I find that the top site on the non-advertisement list is my old friend Science Daily - kind of curious behavior.
I went on to search for an elusive vacuum sweeper filter - not available in stores.
I ended the day with 50 searches but still no new recommendations on my Recommendations tab. Then I noticed that the text in the boxes said: "You have no recommended searches for today."
It looks like gratification must be deferred. They update these recommendations daily.
Last week I added a new Interesting Items For You gadget to my Humor tab of iGoogle. When I logged in this week, I still didn't have anything recommended but it suggested that I click to setup a new tab just for recommendations - just for me. I felt honored, so I clicked.
I now am the proud owner of a new iGoogle tab called Recommendations. It plans on finding items of interest for me in the following areas:
- Videos
- Pages
- News
- Groups
- Gadgets
- Searches
I start off the session with only 37 searches and the goal of seeing when the recommender will kick in. I start searching for skin care information. I click a few adds. I refine my search using Google's recommended searches. I Click an ad that takes me to Science Daily's skin care news, which happens to feature several Google ads prominently displayed in the middle of news stories. I click on one of these ads a learn about the science of skin care.
When I make it back to Google and, once again, search for skin care, I find that the top site on the non-advertisement list is my old friend Science Daily - kind of curious behavior.
I went on to search for an elusive vacuum sweeper filter - not available in stores.
I ended the day with 50 searches but still no new recommendations on my Recommendations tab. Then I noticed that the text in the boxes said: "You have no recommended searches for today."
It looks like gratification must be deferred. They update these recommendations daily.
Personalized Conversational Case-Based Recommendation
http://www.cs.utah.edu/~cindi/papers/ewcbr.pdf
This is a foundation document that the authors (Goker & Thompson) published back in 2000 and has been referenced by them, and others, multiple times over the years.
The team built a recommender system that featured a conversational approach to a personalized recommendation - in this early case, a restaurant choice. The system was know as the Adaptive Place Advisor.
This research was funded by an automotive company, so we shouldn't be surprised that the target user is busy driving a car. The interactive user interface uses voice prompts and voice responses. The system gets to know the user over time and has the goal of not only providing agreeable restaurant choices but an improved user experience over time.
An examples of how the system adapts to the user: The system will delay the repeated recommendation of the same restaurant, even if the user likes the recommendation - the thought being that even if you like cheap Mexican food, you probably don't want to hear about Poncho's every day.
It's a good reference that addresses many of the issues of building this kind of recommender.
This is a foundation document that the authors (Goker & Thompson) published back in 2000 and has been referenced by them, and others, multiple times over the years.
The team built a recommender system that featured a conversational approach to a personalized recommendation - in this early case, a restaurant choice. The system was know as the Adaptive Place Advisor.
This research was funded by an automotive company, so we shouldn't be surprised that the target user is busy driving a car. The interactive user interface uses voice prompts and voice responses. The system gets to know the user over time and has the goal of not only providing agreeable restaurant choices but an improved user experience over time.
An examples of how the system adapts to the user: The system will delay the repeated recommendation of the same restaurant, even if the user likes the recommendation - the thought being that even if you like cheap Mexican food, you probably don't want to hear about Poncho's every day.
It's a good reference that addresses many of the issues of building this kind of recommender.
14 February 2008
I Cann't Recomment This Service - SeenThis?
http://seenthis.loomia.com/
http://www.facebook.com/apps/application.php?id=10066213622
Here we have a new facebook application that allows you to "take your social network with you".
Loomia has just released a new application - SeenThis? - that allows users on media sites (Wall Street Journal, NBC), with just a click, send their facebook network their latest recommendation. This recommendation gets combined with the other recommendations of your friends / groups / networks and presented to them while on facebook.
Loomia gets all excited about the flip side of the application. While on a partner site, say, the Wall Street Journal, you get to see what your facebook contacts think about the current articles.
It seems that you are never alone with SeenThis? and facebook.
I don't much like this whole idea. I am not interested in sharing this kind of information. I am also not interested in seeing what all of my buddies are reading. And, just when I thought I was all alone with this kind of antisocial networking bias, I looked at the second link that shows that not very many people are using this service. Somehow this makes me feel better about life on our planet.
http://www.facebook.com/apps/application.php?id=10066213622
Here we have a new facebook application that allows you to "take your social network with you".
Loomia has just released a new application - SeenThis? - that allows users on media sites (Wall Street Journal, NBC), with just a click, send their facebook network their latest recommendation. This recommendation gets combined with the other recommendations of your friends / groups / networks and presented to them while on facebook.
Loomia gets all excited about the flip side of the application. While on a partner site, say, the Wall Street Journal, you get to see what your facebook contacts think about the current articles.
It seems that you are never alone with SeenThis? and facebook.
I don't much like this whole idea. I am not interested in sharing this kind of information. I am also not interested in seeing what all of my buddies are reading. And, just when I thought I was all alone with this kind of antisocial networking bias, I looked at the second link that shows that not very many people are using this service. Somehow this makes me feel better about life on our planet.
Weekly Recomender Log
Last week I reported that I had moved to Charleston, SC. I am sad to report that Google doesn't believe me. I logged on to my Google account and noted that the logo was sporting a Valentine's Day theme. So, I thought that I would do some shopping. I searched and clicked away on all sorts of romantic ads.
Then I noticed that I was getting some ads for romantic getaways in Charlotte & Asheville - something from my past must be at play.
I found a couple of additional places in Google to plug in my zip. And after a few more shopping sessions, I no longer was invited to North Carolina locations. Apparently, South Carolina doesn't have have any romantic spots - at least not ones that need to advertise.
I found something that I hadn't seen before on Google - a gadget from Google called "Interesting Things For You". It seems that they can keep an eye on my activities and find things that I would find, well, "interesting". All of this must take some time because the tool currently reports:
There aren't enough searches for interesting items yet. Check back after searching for a couple days and be sure you're signed in when you search.
I will try harder.
Then I noticed that I was getting some ads for romantic getaways in Charlotte & Asheville - something from my past must be at play.
I found a couple of additional places in Google to plug in my zip. And after a few more shopping sessions, I no longer was invited to North Carolina locations. Apparently, South Carolina doesn't have have any romantic spots - at least not ones that need to advertise.
I found something that I hadn't seen before on Google - a gadget from Google called "Interesting Things For You". It seems that they can keep an eye on my activities and find things that I would find, well, "interesting". All of this must take some time because the tool currently reports:
There aren't enough searches for interesting items yet. Check back after searching for a couple days and be sure you're signed in when you search.
I will try harder.
07 February 2008
Weekly Recomender Log
This week I moved to Charleston, SC - at least my Google persona is in Charleston! I searched the real estate market and settled into the James Island neighborhood. My new zip code is 29412. I am beginning to feel right at home.
This is a very strange process for me. Over the last several years, I have been concerned about privacy and have only given out demographic data if I had a very good reason. Now I am clicking like mad!
Another strange notion for me is all of the ads I am seeing. I tend to bock ads. But not now, everybody in Charleston want to sell be something. I must be important, everybody want to get to know me.
I noticed that when I search on "James Island, SC" that I got a lot of adds from Realtors that were promoting them selfs as James Island but when I click, I was taken to their generic Charleston site where I was invited to search again for James Island.
This is a very strange process for me. Over the last several years, I have been concerned about privacy and have only given out demographic data if I had a very good reason. Now I am clicking like mad!
Another strange notion for me is all of the ads I am seeing. I tend to bock ads. But not now, everybody in Charleston want to sell be something. I must be important, everybody want to get to know me.
I noticed that when I search on "James Island, SC" that I got a lot of adds from Realtors that were promoting them selfs as James Island but when I click, I was taken to their generic Charleston site where I was invited to search again for James Island.
The BellKor solution to the Netflix Prize
http://www.netflixprize.com/assets/ProgressPrize2007_KorBell.pdf
Here's the tell all story of how the leaders in the NetFlix Prize did it - A step by step guide to all the 107 steps to glory. Glory in this context is defined as a RMSE score of 0.8712.
I don't begin to understand more than one underlying point of this document. And, that point is that none of their individual optimizations worked near as well as a blending of the methods.
These people don't mind telling you how they do their work. A background document can be found at:
ftp://ftp.computer.org/press/outgoing/proceedings/icdm07/Data/3018a043.pdf
Here's the tell all story of how the leaders in the NetFlix Prize did it - A step by step guide to all the 107 steps to glory. Glory in this context is defined as a RMSE score of 0.8712.
I don't begin to understand more than one underlying point of this document. And, that point is that none of their individual optimizations worked near as well as a blending of the methods.
These people don't mind telling you how they do their work. A background document can be found at:
ftp://ftp.computer.org/press/outgoing/proceedings/icdm07/Data/3018a043.pdf
31 January 2008
It's Hard to Hide From Your 'Friends'
Here's the link - http://online.wsj.com/article/SB120164900173426771.html
This article from the 30-Jan-08 Wall Street Journal, relates directly to last week's discussion about Double Click and the unintended use of one's Internet activity for unintended use.
The article starts by giving three examples of questionable sharing of Internet activities by Facebook, Google, and Sears.
Facebook has a system called Beacon that likes to tell your "Friends" what you have purchased at sites like Overstock.com - so everybody on your 'friends' list automatically know about your new, little black dress.
Google Reader can automatically let your 'friends' know about what you are reading. But, it also has a curious feature where Google determines that you have should be 'friends' with other people based on your Google Talk address book. So even if you don't know them, they are still friends and should also know about your reading list.
And, perhaps the craziest of all, Sears offered a purchase tracking service at ManageMyHome.com that, given a name, phone number, and address would report on purchases at Sears. The company has dropped the service but claimed that it was only providing "easy access to useful information". Indeed.
This rush to personalize our ever experience online, once again is banging into privacy issues.
This article from the 30-Jan-08 Wall Street Journal, relates directly to last week's discussion about Double Click and the unintended use of one's Internet activity for unintended use.
The article starts by giving three examples of questionable sharing of Internet activities by Facebook, Google, and Sears.
Facebook has a system called Beacon that likes to tell your "Friends" what you have purchased at sites like Overstock.com - so everybody on your 'friends' list automatically know about your new, little black dress.
Google Reader can automatically let your 'friends' know about what you are reading. But, it also has a curious feature where Google determines that you have should be 'friends' with other people based on your Google Talk address book. So even if you don't know them, they are still friends and should also know about your reading list.
And, perhaps the craziest of all, Sears offered a purchase tracking service at ManageMyHome.com that, given a name, phone number, and address would report on purchases at Sears. The company has dropped the service but claimed that it was only providing "easy access to useful information". Indeed.
This rush to personalize our ever experience online, once again is banging into privacy issues.
15 January 2008
Recommender Watch
For this project, I intend on monitoring everybody's favorite searcher - Google.
If I knew more about recommender sites, I might have branched out a bit. But still, I have tried to have tried to have some fun with this even if it is a familiar site. I have setup a new computer and a brand new user ID in order to not let all of my years of history run over the engine. And, the characteristics of this user ID are quite different than my "real" ID.
As we go forward, I should be able to quickly try things on both machines and check for any variations.
If I knew more about recommender sites, I might have branched out a bit. But still, I have tried to have tried to have some fun with this even if it is a familiar site. I have setup a new computer and a brand new user ID in order to not let all of my years of history run over the engine. And, the characteristics of this user ID are quite different than my "real" ID.
As we go forward, I should be able to quickly try things on both machines and check for any variations.
Week 1 Resource
'Reinforcing the Blockbuster Nature of Media': The Impact of Online Recommenders
http://knowledge.wharton.upenn.edu/article.cfm?articleid=1818Underlying research paper:
http://papers.ssrn.com/sol3/papers.cfm?abstract_id=955984
This week's news / resource is an article from the Warton Business School that reviews and comments on some recent research by members of the Warton staff.
The article, and supporting research, suggests that online retailers need to be careful when implementing a recommender system so that they don't continually recommend a short list of items that are supposed to be part of "the Long Tail" of their inventory. Instead of milking more profits from the "Tail" they will simple create more "Blockbuster" sales.
One little tangent in the article was of particular interest to me. Apparently, the whole recommender thing at Amazon was a real after thought. The developer of the original feature was contacted and claims that he was told to stop development of the feature because it would get in the way of a speedy checkout.
Subscribe to:
Posts (Atom)