Slashdot Mirror

As a dabbler in collaborative filtering (memigo predated Findory), I got to agree: item-to-item collaborative filtering was neither that obvious nor that easy to implement. I hope that Findory doesn't get in to trouble over this, as they are doing some really neat stuff...

Well, this is (half) a shameless plug but my newsbot has been ranking news sources and referring meta-news sources according to "trust" for over 3 yrs now. Findory (no affiliation) does something similar by ranking each individual news story.

The problem is technical, and solvable: my newsbot for example offers a personalized list of top news articles formatted for PDA/mobiles. I am sure there are other services that go beyond news...

The problem is technical, and solvable: my newsbot for example offers a personalized list of top news articles formatted for PDA/mobiles. I am sure there are other services that go beyond news...

Shameless plug ahead: my newsbot (which predates Google News BTW), has a text version which is not only geared for thin devices (PDAs/phones) but also links to news article versions that are also lightweight. Try it out.

Shameless plug ahead: my newsbot (which predates Google News BTW), has a text version which is not only geared for thin devices (PDAs/phones) but also links to news article versions that are also lightweight. Try it out.

I admit it, I am biased, but memigo (my newsbot) let's you share news recommendations with a network of peers, or keep them to yourself --and it's been doing so for over 3 years... Either way, you reap the benefits of collaborative filtering and aggregation.

Well, I am biased, but I think the future of news delivery is towards personalized news delivery, tailored to the interests of each reader. Memigo and Findory (no relation) are two examples of personalized news agents. MSN Newsbot is also going into that direction and I am betting Google will follow soon.

Well, I am biased, but I think the future of news delivery is towards personalized news delivery, tailored to the interests of each reader. Memigo and Findory (no relation) are two examples of personalized news agents. MSN Newsbot is also going into that direction and I am betting Google will follow soon.

Is this the same Henk Poley from Memigo?

I run a similar, albeit personalized service (which predates Google News actually) and I'll have to pipe in and say that I doubt that the real reason for the absense of ads on GN is that Google is afraid: first of all, GN drives traffic to news sites, and more traffic means more money for the originating site. Excluding yourself from GN is basically handing money to your competition.

I think the real problem with GN, is that context sensitive advertising does not work for news. I've been running AdSense ads on memigo.com for a while now and Google never managed to keep up: by the time they spidered the site, the content had changed. Now, let's assume that they can solve this problem since GN is their own site, and they can update immediately: which advertisers are going to rely on context ads for news items? Imagine a story popping up on the US feed about say a Ford Explorer flipping over, with nice big Ford ads next to it: a waste of money and space. And if you try to go the other way, showing ads only for positive pieces of news (hard, but let's say it's doable) you'll be accused of bias and selling out.

So, the only reasonable choice is to sell non-context ads on GN. It could happen, but I think Google likes a challenge; they'll mine GN clicks and probably do personalized ads before they go back to plain-old ads...

Well, in that case you may ne interested in memigo; granted, it's not as customized as a home-brew solution, but it offers a boat-load of features including custom RSS feeds, RSS feeds of your browsing history, keyword searching of read articles, related news articles, PDA views, etc, etc, etc... check it out (end of shameless plug).

Well, at least as far as Google News goes, you can try my newsbot, which actually pre-dates GN...

You may wanna try the PDA version of my newsbot: not only it's light-weight itself, but you can customize it to link to light-weight versions of news articles only --plus it's personalized to your interests. Give it a shot...

Well, a login/password on memigo is needed only if you want the more advanced features. At any rate, memigo and sites like it are essentially serendipity agents, they are not meant to be your only source of web news.

But you mentioned the Zaurus, so you will probably like the PDA version of memigo, which was actually the original reason I built the site (not just PDA-formatted, but linking to PDA-formatted articles as well, and customized for you)...

Memigo is very cool with a ton of power user features. It's a fantastic site.

If you're looking for something simple and easy to use, Findory News is another good choice.

MSN Newsbot does look a lot like Google News, but it does have something unusual, personalized news. The site watches which articles you read and attempts to find other interesting articles. Microsoft certainly isn't the first here, but they are the biggest.

If done right, personalized news can work very well. Of course, trust is a big issue. If you don't trust Microsoft, give Findory News or Memigo a try.

Well, at least one thing that MSN Newsbot dows that Google News doesn't, is that MSN attempts to personalize the news page based on passed clicks. Now, my newsbot has been doing that for almost 3 years now (plus a lot more, like customized XML and PDA feeds, peer networks, etc) with a more varied selection of sources (end of shameless plug :-)

As a provider of personalized, customized RSS feeds, I really do like your idea. Essentially a "Next-Modification" HTTP header that would tie together with the "Last-Modified" header would help greatly. Right now, the best you can do is hope that the RSS client respects 304 error codes, and goes away when you tell it there is no updated content. That doesn't stop it though from coming back 15' later... RFC? :-)

Well, then check out memigo (my own newsbot) or findory (not affiliated) for interesting experiments on news delivery without Google's infrastructure (both sites are one-man operations, and I only work on memigo on my spare time).

Because I always hated /.'s RSS myself, I ended up building this newsbot to get my news fix: memigo will scrape HTML and read RSS and it will rank articles based on user ratings and the "reputation" that each source builds up over time. More interestingly (and on-topic), memigo will produce custom RSS feeds with just your recommended news articles --you basically get a special URL to get custom RSS from, or even custom PDA-optimized feeds if you want.

As far as I understand things, besides personality issues, the Atom folks were looking for more i18n and for a more-specific standard --there are tags in RSS who are being (mis)used differently by different content-producers exactly because the spec was not very clear from the beginning.

As an RSS producer/consumer myself, the one thing I've always hated about RSS was the encoding of the description tag: some feeds escape any HTML included in description, some make the whole tag a big CDATA entity, and in any case there is no information provided as to the encoding of the included HTML. One of the side effects has been that if you are parsing RSS, you have to assume that description includes HTML. So, if you happen to have > or < or any other HTML-looking entities within description, your content will be mangled by the RSS-consuming code.

You maybe interested in memigo (shameless plug). Memigo attempts to complete automate the slashdot process: anyone can submit an article, but a) the site code does the QA of the article itself automatically, and b) the users rate each and every article, in effect moderating the front page. As an added bonus, memigo is context-sensitive (so you can monitor topics you're interested in) and of course Amazon-like collaborative filtering. Check it out...

Let me just point out there are other ways to make more interactive web apps: JSRS is a free JavaScript Remote Scripting library that lets client-side JS communicate with the server: think listboxes or menus that get populated based on button clicks or check-boxes in the web page, without re-loading the page. It works here and now, and on IE, Gecko, KHTML and Opera (not a plug, just a happy developer).

You can see JSRS in action on my newsbot, where it lets you rate articles dynamically without re-loading the web page or submitting forms (in my example the server-side solution is Python Webware, but JSRS is simple enough to get to work with anything, and in fact there are already libraries for PHP, Perl, ASP, etc).

Agreed. I hate this place when the geeks think they're being funny. Slashdot became much more readable when I started assigning a -3 moderation to anything moderated as "Funny" (check the prefs -- great feature). Unfortunately the same trick doesn't work for the front page.

In the meantime, here are some other good tech/news sites to check out while Slashdot is useless:

• Ars Technica
• Kuro5hin
• Plastic
• MetaFilter
• MemePool
• Memigo

Enjoy!

Bayesian needs pre-determined "bins" of data to assign a new piece of information to --that's a limited approach that will break down for news articles or generic Web pages. A combination of context- and collaborative-filtering is a much better approach IMSHO (that's my newsbot, BTW).

Well, my newsbot does this already for news articles you've read through it: you can search everything, you can search articles you've read, or articles you've read *and* rated highly. You can also set up "search alerts" that search any new articles and then stick them to your front page (or your personalized RSS feed, or your personalized PDA-optimized page). Check it out.

GN is good, but does it let you program personal searches every time you visit? does it learn the subject you're more interested in and suggest new articles on the front page every day? does it let you add to its news feeds directly? give you personal XML and PDA feeds to take home? let you share your favorite stories with friends (automatically)? No, not yet anyway... until then there is memigo [end of shameless plug]

Another self-plug: my newsbot extends the same ideas (peer recommendations, collaborative filtering) for news articles: it's a much more powerful way to filter the news than say, /. :-)

Check out memigo Now! (plug). Memigo can provide you with a customized light-HTML page that links to recommended articles in light formats only (i.e. PDA- or printer-friendly). Much nicer for PDAs or cell phones than a plain RSS feed.

The standards wars: RSS 0.9 vs RSS 1.0 vs RSS 2.0 vs Atom. As a provider if I want to reach as many people as possible I will have to provide 4 different formats! (RSS 2.0 should be readable by RSS 0.9 readers but you never know).

The bad client implementations: repeat after me: 304 Modified. If you consume XML/RSS, make sure your client supports 304 Modified responses, and provides Last Modified and ETags. Otherwise, you're wasting my bandwidth, and I'll have to ban your customers (which I don't want to do!).

RSS is less two-way than HTML: a lot (not all definitely) of the RSS clients make it hard to interact with the authoring site, much more so than plain HTML and a browser. Fortunately, this is changing.

IMHO, RSS is a good first attempt at a truly automated, interactive Web experience. But the killer apps will have to wait for better technology and infrastructure...

• Plus, they HAVE done new things, such as google news.

I had the same thought some years ago and implemented a similar concept in memigo. Memigo is a newsbot/news aggregator where users can rate articles; memigo is also aware of the context of each article, so implicitly users rate contexts as well. Based on that data, some neat things are doable: Amazon-like collaborative filtering, automated formation of "alike" users, "Interest Alerts", where the newsbot searches for articles with context that you have rated highly in the past, etc. Memigo is pretty experimental (and a one-man, spare-time operation) but it's pretty good, if I may say so and has a few more innovative features I won't get to here; check it out.

End of blatant plug :-)

There are other alternatives, such as cutting the budget of Cold War-era military projects that offer no particular value.

IBM and Philips are also getting into the action.

OK, this is a plug, but I think memigo does a mobile news portal genuinely better: you get a personalized URL to access news over a mobile phone or PDA, and what's more the headline view can link you to only lightly-formatted content of each article (i.e. a PDA- or printer-formatted version of each article). Much better than linking to whatever version the site's RSS feed links to. Check it out.

I would be really interested to see a study seeking to find a link between internet usage and awareness/involvement in current events.

Many of my friends who aren't on the internet very much are always asking me what's going on in the world. Though I am not sure if it is internet users or memigo users.

-Jackson

Well, for an attempt at a better newsbot than Google news, you can check out newsbot here. It does a few things that GN leaves out (XML feeds, PDA version, peer recommendations, etc, etc) and I believe it has a better S/N ratio. End of shameless plug.

Well, my newsbot does a lot of what you describe, at least for news articles. Give it a shot.

This is my newsbot.

Shameless plug ahead: my home-brew newsbot predates both MSN Newsbot and Google News and has way more features (incl. collaborative filtering, news alerts, customized XML and Palm feeds and more).

The meta-news site Memigo.com was doing this story relationship analysis two years ago. Too bad the creator didn't patent the idea. :-)

Don't we have Memigo for this?

For the unknowing, Memigo is an intelligent news agent. It allows registered users to rate articles. High rated news items will come up on the regular frontpage. When you create a login yourself Memigo will 'learn' what news you like, and via collaborative filtering, others with similar tastes will recommend news items to you.

Another interesting news experiement is memigo.com. It's a meta-news site, like Google News, except it uses an Amazon.com-like algorithm to predict the news stories that you will want to read.

It's a clever idea, but the stories get a bit repetitive after it learns your preferred news topics. I think the algorithm should include a few more random and underrated links.

[Plug] My newsbot does News Search alerts plus context alerts (news articles it thinks you may like based on your past read/rated articles) and they are all available via customized RSS and PDA feeds.

[Plug] My newsbot does News Search alerts plus context alerts (news articles it thinks you may like based on your past read/rated articles) and they are all available via customized RSS and PDA feeds.

[Plug] My newsbot does News Search alerts plus context alerts (news articles it thinks you may like based on your past read/rated articles) and they are all available via customized RSS and PDA feeds.

Clean site, presents a lot of content at once, adapts to your experience-level (new users get contextual help items).

http://memigo.com

In my view, personal recommendations from a search engine are mostly valuable for topical content --i.e. news items. However, the optimizations from these papers don't sound to me like they can do much for this case --news items pop up in a news site, and re-indexing the news source itself (say, the front page of CNN) won't tell you much about a particular CNN story.

At any rate, personal news recommendations is a favorite topic of mine: this is why I built Memigo: to create a bot that finds news I am more likely to like. Memigo learns from its users collectively and each user individually --and BTW, it predates Google News by a good 6 months, IIRC. The memigo codebase (all in Python) is now up to the point where it can start learning what content each user likes... If you like Google News you'll love Memigo.

And BTW, I did RTFA when it was on memigo's front page this morning :-)...

There are easy ways around this: you can age songs (guaranteeing that new songs get priority) or you can use web-of-trust ratings to see what high-trusted peers recommend currently.

BTW, I know this as my newsbot does the same sortof filtering for news articles, and although there is some feedback reinforcement for recent news, it does work suprisingly well.