Looking for illumio in Google Reader
My old company, Tacit Software, has ceased to be: its intellectual property assets were acquired by Oracle. Two weeks ago Oracle shut down the illumio service, bringing the illumio experiment to a close. illumio was a feed reader with matching based filtering, dynamic interest discovery and adaptive matching. With illumio I was able to subscribe to more than 200 RSS feeds and keep up, as the system automatically showed me items most relevant to my interests. illumio automatically discovered what I was interested in and got better with use through my feedback. It presented articles using a newspaper metaphor, with a front page and sections grouping RSS feeds by topic area. It was addictive, but now it is gone, and I have switched to Google Reader.
Google Reader works well, updates quickly and is easy to read. I like the sharing feature and the ability to expose my shared items as an RSS feed. I use this feed to populate the box at the bottom right of this blog. I also like the incredibly easy search and subscribe user interface that makes it trivial to discover RSS feeds.
But there were some features of illumio that I sorely miss.
Matching Based Filtering
illumio matched articles to the interests of the user, assigning a matching strength to each article and then allowing filtering of the articles by strength.
Each article was distilled into a search expression capturing the article's main focus which was then used to match the article against the interests of the user. Articles were assigned a matching strength giving default priority to the best matching articles. The illumio home page showed the best, current matching articles, selected from a variety of sources. It kept changing during the day, much like an online newspaper. It provided a personalized view of the world, customized just for you. Each of the newspaper sections could also be viewed by strength, as well as recency and other criteria.
Matching was done against a user profile composed of explicit interests and dynamic interests automatically discovered from user activity.
Dynamic interests were discovered by looking at sent email, documents on the user's machine and web history. As the user talked about a certain subject in email or browsed for it on the web, articles on that subject would be brought to their attention automatically. Since illumio had a client running on the user's machine, none of this information left the machine without user consent, ensuring total privacy.
Explicit interests could be declared by the user, globally and by individual RSS feed. The user could ask illumio to see all articles about "Barack Obama" in the Huffington Post feed or all articles about "Pro Evolution Soccer" in any feed. Explicit matching resulted in a special kind of strength, letting people easily zone in on explicit matches.
With matching and filtering I could subscribe to more feeds than I would ever read. I used illumio to track around 200 feeds feeling confident that I would not miss articles in areas of real interest.
Without it, I'm already drowning. I've subscribed to about 20 feeds in Google Reader and I already feel overwhelmed. I need to scan every individual item to make sure I don't miss something I'm interested in. Without filtering, I cannot use Google Reader to track more than a handful of feeds.
I've searched for alternatives which come in a few different flavors. One set is to use a third party to filter my feeds with keywords and then subscribe the results with Google Reader, using Yahoo Pipes, for example. Two problems with this approach: it's a pain to set up and the removal of the articles before arriving in Google Reader is too extreme. Another approach is to use a Firefox extension to highlight articles based on regex matching on the titles. This approach is also limited: it only matches the titles and it only works in the list view, which I'm not using at all.
Adaptive Matching
Using explicit interests and ignores (the opposite of interests), illumio could be trained to get better. Anytime you saw an article you liked you could click "more like this" and tell illumio what aspect of the article you'd like to match more of in the future (or "less like this" to prevent matching on inane subjects).
With adaptive matching illumio got better and better after some use, in small incremental steps. After a while I did not feel the need to train it any further.
Get Contents From Web
This felt like a tiny feature at a time, but now that I use Google Reader, I miss it a great deal. RSS feeds are not all equal. Each site can choose how much of the content to include with each article. Some sites choose to include the full text of the article, even with pretty formatting. For these sites, like the Huffington Post or Engadget for example, Google Reader is great. Other sites include a tiny snippet of the article as a teaser, expecting you to click on to their site to enjoy the full contents together with their ads, branding and other noise content I don't care for. For these sites I find Google Reader useless. I almost want to immediately unsubscribe from these feeds. I don't like clicking through to open the target site on another tab. I'm in the reader to read the article, but the article is not there. I feel cheated, mired in a poor user experience.
illumio dealt with this problem by following the link in the article and heuristically scrubbing the HTML of the article page for the contents of the article. This resulted in almost all feeds being able to display contents within the client without needing to jump to the original article site. It was not perfect and formatting was poor, but I could read the article without leaving the reader.
So... why not in Google Reader?
Google Reader is very nice in a number of ways and it could be even greater with some of these features, matching based filteringbeing the key one to add. And Google is in a unique position to make it work for us. If the user has GMail, then Google already has the emails to automatically derive interests from. If we use Google for searching the web, and who doesn't, Google knows what the user is searching for...
Making sure there's always text to show for articles is another obvious improvement. Neither of these features can be done efficiently on the client side using a Firefox extension. If makes much more sense to do it once on the server as the articles are being processed. Of course, evaluating matching of each article for each user subscribed to its feed increases the resources required to process feeds for a large population very significantly. If anyone out there is up to it, it's Google.
As we found in developing illumio, adding matching and filtering creates a number of non trivial challenges for the user interface. The UI gets more complicated when you need to show matching indicators as well as all the machinery for filtering by strength, ordering and adaptive matching. Once you show a matching strength people need to see an intelligible justification and act on it to modify behavior.
The bottom line question: does Google want to turn Google Reader into the best RSS reader or is it content with the current, adequate effort?
Google Reader works well, updates quickly and is easy to read. I like the sharing feature and the ability to expose my shared items as an RSS feed. I use this feed to populate the box at the bottom right of this blog. I also like the incredibly easy search and subscribe user interface that makes it trivial to discover RSS feeds.
But there were some features of illumio that I sorely miss.
Matching Based Filtering
illumio matched articles to the interests of the user, assigning a matching strength to each article and then allowing filtering of the articles by strength.
Each article was distilled into a search expression capturing the article's main focus which was then used to match the article against the interests of the user. Articles were assigned a matching strength giving default priority to the best matching articles. The illumio home page showed the best, current matching articles, selected from a variety of sources. It kept changing during the day, much like an online newspaper. It provided a personalized view of the world, customized just for you. Each of the newspaper sections could also be viewed by strength, as well as recency and other criteria.
Matching was done against a user profile composed of explicit interests and dynamic interests automatically discovered from user activity.
Dynamic interests were discovered by looking at sent email, documents on the user's machine and web history. As the user talked about a certain subject in email or browsed for it on the web, articles on that subject would be brought to their attention automatically. Since illumio had a client running on the user's machine, none of this information left the machine without user consent, ensuring total privacy.
Explicit interests could be declared by the user, globally and by individual RSS feed. The user could ask illumio to see all articles about "Barack Obama" in the Huffington Post feed or all articles about "Pro Evolution Soccer" in any feed. Explicit matching resulted in a special kind of strength, letting people easily zone in on explicit matches.
With matching and filtering I could subscribe to more feeds than I would ever read. I used illumio to track around 200 feeds feeling confident that I would not miss articles in areas of real interest.
Without it, I'm already drowning. I've subscribed to about 20 feeds in Google Reader and I already feel overwhelmed. I need to scan every individual item to make sure I don't miss something I'm interested in. Without filtering, I cannot use Google Reader to track more than a handful of feeds.
I've searched for alternatives which come in a few different flavors. One set is to use a third party to filter my feeds with keywords and then subscribe the results with Google Reader, using Yahoo Pipes, for example. Two problems with this approach: it's a pain to set up and the removal of the articles before arriving in Google Reader is too extreme. Another approach is to use a Firefox extension to highlight articles based on regex matching on the titles. This approach is also limited: it only matches the titles and it only works in the list view, which I'm not using at all.
Adaptive Matching
Using explicit interests and ignores (the opposite of interests), illumio could be trained to get better. Anytime you saw an article you liked you could click "more like this" and tell illumio what aspect of the article you'd like to match more of in the future (or "less like this" to prevent matching on inane subjects).
With adaptive matching illumio got better and better after some use, in small incremental steps. After a while I did not feel the need to train it any further.
Get Contents From Web
This felt like a tiny feature at a time, but now that I use Google Reader, I miss it a great deal. RSS feeds are not all equal. Each site can choose how much of the content to include with each article. Some sites choose to include the full text of the article, even with pretty formatting. For these sites, like the Huffington Post or Engadget for example, Google Reader is great. Other sites include a tiny snippet of the article as a teaser, expecting you to click on to their site to enjoy the full contents together with their ads, branding and other noise content I don't care for. For these sites I find Google Reader useless. I almost want to immediately unsubscribe from these feeds. I don't like clicking through to open the target site on another tab. I'm in the reader to read the article, but the article is not there. I feel cheated, mired in a poor user experience.
illumio dealt with this problem by following the link in the article and heuristically scrubbing the HTML of the article page for the contents of the article. This resulted in almost all feeds being able to display contents within the client without needing to jump to the original article site. It was not perfect and formatting was poor, but I could read the article without leaving the reader.
So... why not in Google Reader?
Google Reader is very nice in a number of ways and it could be even greater with some of these features, matching based filteringbeing the key one to add. And Google is in a unique position to make it work for us. If the user has GMail, then Google already has the emails to automatically derive interests from. If we use Google for searching the web, and who doesn't, Google knows what the user is searching for...
Making sure there's always text to show for articles is another obvious improvement. Neither of these features can be done efficiently on the client side using a Firefox extension. If makes much more sense to do it once on the server as the articles are being processed. Of course, evaluating matching of each article for each user subscribed to its feed increases the resources required to process feeds for a large population very significantly. If anyone out there is up to it, it's Google.
As we found in developing illumio, adding matching and filtering creates a number of non trivial challenges for the user interface. The UI gets more complicated when you need to show matching indicators as well as all the machinery for filtering by strength, ordering and adaptive matching. Once you show a matching strength people need to see an intelligible justification and act on it to modify behavior.
The bottom line question: does Google want to turn Google Reader into the best RSS reader or is it content with the current, adequate effort?
Comments
Post a Comment