Hi Wikimedia Developers,
My name is Cheng Xing, and I'm interested in working with Wikimedia for GSoC this summer. I sent the email below to the mailing list few days ago, but it didn't seem like it went through, so here it is again.
Thank you for your time!
Sincerely, Cheng ---------- Forwarded message ---------- From: "Cheng Xing" cxing561@gmail.com Date: Apr 18, 2013 5:40 PM Subject: [Google Summer of Code '13] Project Idea - "Inspire Me" Button To: wikitech-l@lists.wikimedia.org
Hi Wikimedia Developers,
My name is Cheng Xing, and I am a freshman in Cornell University College of Engineering planning to pursue a major in Computer Science and minor in Electrical and Computer Engineering.
The gist of my idea is this: Create a magical "Inspire Me" button in the homepage of Wikimedia sites so that it directs the user to a page that he/she is most likely interested in. In other words, it's a page recommender system.
For example, if a programmer clicks the "Inspire Me" button on Wikipedia, articles such as the Whitespace programming language, Rubber Duck Debugging, etc. would show up. Things that the user probably doesn't know about, that would probably interest the user, will show up by clicking that button. Very occasionally there'd be random things like Stitches, which the user might know nothing about, but might actually be interesting.
I got this idea from three different places: Pandora, XKCD, and my own Wikiholic-ness. Pandora, for its impressive recommender system that uses user accounts and likes/dislikes to track recommendation data; XKCD, for its entertainment through their "Random" button; and lastly, my own Wikiholic-ness, for its eagerness to find random interesting things on Wikipedia.
I think the best part of Wikimedia is its ability to inspire people from all over the world, and it has achieved this by simply presenting information to the masses. In my opinion, a tool that filters and recommends information to users would be much more inspirational. Just imagine how many people all over the world can find their dreams this way.
I realize that this could become quite a big project, so if I get the chance to work on this, I will do a small part (possibly the basic infrastructure of the system) for GSoC, and I am more than willing to continue to contribute after that.
I have some ideas of how this recommender systems would work, but this email is pretty long as it is. Please send me questions and comments! I really appreciate it.
Sincerely, Cheng
Hi Cheng,
As you say there are many recommendation systems, some of them already work or used to work with wikipedia (like StumbleThru) or just to find interesting articles (like reddit.com/r/wikipedia ). In my opinion those systems are better developed externally because if not done right, they can get easily annoying... for instance, if you use previous contribution or watchlist data, I might have edited or be watching articles about themes I am not longer interested in. And if done right, then you need to put many more hours than a Gsoc would allow you. Anyway, maybe a mentor is interested, if not take a look to the proposed projects with mentor. There are some cool ideas there too :)
Micru
On Mon, Apr 22, 2013 at 12:54 PM, Cheng Xing cxing561@gmail.com wrote:
Hi Wikimedia Developers,
My name is Cheng Xing, and I'm interested in working with Wikimedia for GSoC this summer. I sent the email below to the mailing list few days ago, but it didn't seem like it went through, so here it is again.
Thank you for your time!
Sincerely, Cheng ---------- Forwarded message ---------- From: "Cheng Xing" cxing561@gmail.com Date: Apr 18, 2013 5:40 PM Subject: [Google Summer of Code '13] Project Idea - "Inspire Me" Button To: wikitech-l@lists.wikimedia.org
Hi Wikimedia Developers,
My name is Cheng Xing, and I am a freshman in Cornell University College of Engineering planning to pursue a major in Computer Science and minor in Electrical and Computer Engineering.
The gist of my idea is this: Create a magical "Inspire Me" button in the homepage of Wikimedia sites so that it directs the user to a page that he/she is most likely interested in. In other words, it's a page recommender system.
For example, if a programmer clicks the "Inspire Me" button on Wikipedia, articles such as the Whitespace programming language, Rubber Duck Debugging, etc. would show up. Things that the user probably doesn't know about, that would probably interest the user, will show up by clicking that button. Very occasionally there'd be random things like Stitches, which the user might know nothing about, but might actually be interesting.
I got this idea from three different places: Pandora, XKCD, and my own Wikiholic-ness. Pandora, for its impressive recommender system that uses user accounts and likes/dislikes to track recommendation data; XKCD, for its entertainment through their "Random" button; and lastly, my own Wikiholic-ness, for its eagerness to find random interesting things on Wikipedia.
I think the best part of Wikimedia is its ability to inspire people from all over the world, and it has achieved this by simply presenting information to the masses. In my opinion, a tool that filters and recommends information to users would be much more inspirational. Just imagine how many people all over the world can find their dreams this way.
I realize that this could become quite a big project, so if I get the chance to work on this, I will do a small part (possibly the basic infrastructure of the system) for GSoC, and I am more than willing to continue to contribute after that.
I have some ideas of how this recommender systems would work, but this email is pretty long as it is. Please send me questions and comments! I really appreciate it.
Sincerely, Cheng _______________________________________________ Wikitech-l mailing list Wikitech-l@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/wikitech-l
How do you intend to get the data you need to pick a page?
For something like this to work properly traditionally I'd expect it would involve storing personal information that currently isn't stored. Something that would require controversial changes to Wikimedia's privacy policy. Have you talked to anyone about this?
On Mon, 22 Apr 2013 09:54:15 -0700, Cheng Xing cxing561@gmail.com wrote:
Hi Wikimedia Developers,
My name is Cheng Xing, and I'm interested in working with Wikimedia for GSoC this summer. I sent the email below to the mailing list few days ago, but it didn't seem like it went through, so here it is again.
Thank you for your time!
Sincerely, Cheng ---------- Forwarded message ---------- From: "Cheng Xing" cxing561@gmail.com Date: Apr 18, 2013 5:40 PM Subject: [Google Summer of Code '13] Project Idea - "Inspire Me" Button To: wikitech-l@lists.wikimedia.org
Hi Wikimedia Developers,
My name is Cheng Xing, and I am a freshman in Cornell University College of Engineering planning to pursue a major in Computer Science and minor in Electrical and Computer Engineering.
The gist of my idea is this: Create a magical "Inspire Me" button in the homepage of Wikimedia sites so that it directs the user to a page that he/she is most likely interested in. In other words, it's a page recommender system.
For example, if a programmer clicks the "Inspire Me" button on Wikipedia, articles such as the Whitespace programming language, Rubber Duck Debugging, etc. would show up. Things that the user probably doesn't know about, that would probably interest the user, will show up by clicking that button. Very occasionally there'd be random things like Stitches, which the user might know nothing about, but might actually be interesting.
I got this idea from three different places: Pandora, XKCD, and my own Wikiholic-ness. Pandora, for its impressive recommender system that uses user accounts and likes/dislikes to track recommendation data; XKCD, for its entertainment through their "Random" button; and lastly, my own Wikiholic-ness, for its eagerness to find random interesting things on Wikipedia.
I think the best part of Wikimedia is its ability to inspire people from all over the world, and it has achieved this by simply presenting information to the masses. In my opinion, a tool that filters and recommends information to users would be much more inspirational. Just imagine how many people all over the world can find their dreams this way.
I realize that this could become quite a big project, so if I get the chance to work on this, I will do a small part (possibly the basic infrastructure of the system) for GSoC, and I am more than willing to continue to contribute after that.
I have some ideas of how this recommender systems would work, but this email is pretty long as it is. Please send me questions and comments! I really appreciate it.
Sincerely, Cheng _______________________________________________ Wikitech-l mailing list Wikitech-l@lists.wikimedia.org https://lists.wikimedia.org/mailman/listinfo/wikitech-l
I believe E:Getting_Started is already working on something like this.
On 04/23/2013 12:37 AM, K. Peachey wrote:
I believe E:Getting_Started is already working on something like this.
It's similar, though "Inspire Me" seems targeted towards interesting things to *read*, whereas GettingStarted is meant for interesting things to *edit.
Also, GettingStarted is not currently personalized, but we are considering possible approaches to doing so.
Matt Flaschen
Hello Cheng, thank you for sharing your proposal.
On 04/22/2013 09:54 AM, Cheng Xing wrote:
I realize that this could become quite a big project, so if I get the chance to work on this, I will do a small part (possibly the basic infrastructure of the system) for GSoC, and I am more than willing to continue to contribute after that.
GSoC projects must be self-contained. If there is no time for a full-fledged and solid solution, then at least a prototype should be delivered to prove the concept.
I have some ideas of how this recommender systems would work, but this email is pretty long as it is.
Ok, we are now ready to see the sauce of this project. :)
Thank you for your interest in contributing to Wikimedia!
On Mon, Apr 22, 2013 at 9:54 AM, Cheng Xing cxing561@gmail.com wrote:
For example, if a programmer clicks the "Inspire Me" button on Wikipedia, articles such as the Whitespace programming language, Rubber Duck Debugging, etc. would show up. Things that the user probably doesn't know about, that would probably interest the user, will show up by clicking that button. Very occasionally there'd be random things like Stitches, which the user might know nothing about, but might actually be interesting.
Matt is right about how the Getting Started work relates to this. I think it would be complementary rather than duplicative.
The big question for implementing a reader recommender system is where you get your data from. Tracking reader activity on Wikipedia is a _very_ touchy subject for a whole host of legitimate reasons, and to be totally honest I don't think you're going to be able to implement any recommender system based on people's reading habits.
You might be able to do this if you limit yourself to recommending solely based on the current page, or based on user input (e.g. a recommender search with "What's your favorite article?" and some matching based on various attributes you could tinker with.)
If you're interested in recommender systems, as Matt says we are nearly ready to explore personalizing the list at Special:GettingStarted based on your past editing history. This is a slightly smaller scope, so if you want to take a whack at this, we might be able to help you along the way with advice, code review, etc. (assuming the engineers on the team don't think I'm crazy).
Steven
On 23 April 2013 11:00, Steven Walling steven.walling@gmail.com wrote:
Tracking reader activity on Wikipedia is a _very_ touchy subject for a whole host of legitimate reasons, and to be totally honest I don't think you're going to be able to implement any recommender system based on people's reading habits.
Can't he use *broad* categories for tracking? like instead of registering an activity as by the article or the immediate category in which it fall, we can traverse up the categories until we get a category under which there are atleast 1000 articles. This will ensure that the user is not tracked explicitly, like more or less he might not even notice, that he is being tracked (Note: tracking like this doesnot make us evil :P)
For example: if a reader visits the page [Burmese rupee], then we can traverse up the categories like, Burmese rupee-> Rupee-> Currency denominations-> Currency-> International finance-> International economics-> Economics. And then save the visit as an activity in category of "International economics" or "Economics" rather than in "Rupee" or "Currency".
Though this will make the suggestions less efficient, but it will surely generate insights into the active interests of the user (and thats the purpose). Also, there can always be an option to manually register the article by the user (kindof bookmark/watchlist), and if he does so, the suggestions will improve for that user. . Regards Gaurav
wikitech-l@lists.wikimedia.org