I’ve been researching Google’s social Q&A sites codenamed Confucius, which are in more than 68 countries and multiple languages, but little known in the US. What I’ve seen includes some tantalizing hints about Google Plus, a description of how content submitted to Google Plus might be ranked in Google Web search, and a possible advertising model for Google Plus that was detailed in a Best Paper nominee at the World Wide Web Conference in North Carolina last year.
I started looking at Confucius a week ago, when I published the post, How Google Might Rank User Generated Web Content in Google & and Other Social Networks. My post describes a ranking signal for user-generated content in Web search results, derived from a social network user’s perceived authority on different subjects and the quality of their contributions in interactions on the network; These combined scores might be used as a ranking signal in web search results for the content that user creates. The patent filing was published at the World Intellectual Property Organization website rather than the US patent office website, and the authors of the patent were from Google China, including Edward Y. Chang, the head of Research at Google China, seen in the profile page below:
The image above is taken from keynote presentation slides by Edward Y. Chang presented at the 18th ACM Conference on Information and Knowledge Management (ACM CIKM) titled Confucius and its Intelligent Disciples. That circle and the one below might look a little familiar to anyone who has tried out Google Plus.
There’s also a paper that describes how some of the mechanisms behind Google Confucius works, and how people asking questions in Google Web search might be prompted to ask the questions at Google Confucius. The paper is Confucius and Its Intelligent Disciples: Integrating Social with Search (pdf), by Xiance Si, Edward Y. Chang, and Zoltan Gyongyi of Google Research and Maosong Sun from Tsinghua University.
As I mentioned above, Google Confucius is available in at least 68 countries at this point, including 40 countries in Africa, (Available in English and French), China, Russia, Thailand, Indonesia, and many countries with Arabic speakers.
Here’s a screenshot of the African version of the site in English:
The site provides several different ways to earn points to go up levels, win badges for different activities, and more. While the site doesn’t look too different from many of the Question and Answer (Q&A) sites you’ve probably seen on the web, some of the magic behind the site is what happens behind the scenes, and in the way that it interacts with Google. How many other Q&A sites in the world will receive visitors from Google with a prompt from the search engine to ask a question on the site?
The image below, from the Confucius and Its Intelligent Disciples paper provides a quick snapshot of some of the unique aspects of the Confucius system:
The paper provides some insights into the assumptions behind Google Confucius, as well as the major technical challenges of putting the system together:
- People seldom provide answers without incentives.
- When incentives exist, people who might abuse or spam a system may emerge, degrade the quality of the service and discourage knowledgeable users from participating.
- When a question is asked, it should be routed quickly to a domain expert to reduce answering delay.
- If an answer to a question is already available, it should be easy to find to avoid redundant work and unnecessary delay.
Here are some of the mechanisms developed for the system that makes it somewhat unique:
1. Integration into Search Results:
When someone enters a query at Google, and the search is a wh-query (when, where, why, etc.), or the search engine may not be able to return sufficiently relevant results (in instances where there might be inadequate search content with an overlap between the query terms and the potential top result pages being low), a Confucius Q&A session is recommended in the Google search results that encourage the searcher to ask their question on the Q&A site.
2.Labeling Questions:
When someone types in their questions, category labels are suggested for the questioner. Those labels can help organize questions, enable people to subscribe to the labels’ topics, and route the questions to people who might have some expertise in those topics. The authors of the paper tell us that they “employ a parallel implementation of Latent Dirichlet Allocation (PLDA) [12] to make label suggestions.”
3. Recommending Questions:
To provide faster responses and avoid people from answering the same questions repeatedly, the question recommendation part of this system looks for similar earlier questions and provides the answer to those. We’re told that PLDA is also useful in finding those previously asked questions in this recommendation task.
4. Evaluating the Quality of Answers:
The patent application I wrote about last week on ranking user-generated content provides some details on how a question or original post might be evaluated in terms of quality, as well as answers and responses. The quality of answers can rely upon how relevant they are to the question, how original they might be in comparison to other answers to that question and similar questions, how wide the coverage of the response might be (does it use a wide mix of related terms), and so on. This evaluation not only helps uncover who the top contributors might be in the system and the members who might be submitting spam. And in curbing spam.
5. User ranking:
Uses of the system are ranked based upon their contributions and their interactions with others. Those rankings are based upon different areas of expertise. They can be used to identify the top contributors in those areas, and to help in providing incentives to people answering questions, and to route questions on specific subjects to people with expertise in those subjects—route questions to domain experts.
6. NLP-based answer generation:
In addition to having people provide answers for questions, Google Confucius may also answer some questions independently, much as Google does in Q&A results on Google.com, such as, “when is the birthdate of Derek Jeter?”
Conclusion
While Google Confucius is much more of a straightforward Q&A site than Google Plus, many ideas being developed at Google Confucius may become part of Google Plus. The circles I showed above look like they may have been part of Confucius for a while, though I did a little looking around and didn’t see them on some of the different versions of Confucius now. I signed up for the English language version of the African Q&A site to see the profile pages for the site. I didn’t see any circles, but I also didn’t add any contacts.
Those ideas that might carry over to Google Plus might include the authority/contribution rankings for individuals to determine whether or not relevant Google Plus content might appear in search results.
The paper AdHeat: An Influence-based Diffusion Model for Propagating Hints to Match Ads pdf (Presentation: AdHeat, An Influence-based Social Ads Model& its Tera-scale Algorithms) was nominated as a best paper at the WWW2010 conference in North Carolina, and it’s worth spending some time with. As the paper tells us, AdHeat is a social ad model that considers user influence in addition to relevance for matching ads with content:
We performed three experiments on Google Confucius, an online Q&A service available in China, Russia, Thailand, and 17 Arab-speaking countries. The results show that Ad-Heat to be more effective over content-relevance and user targeting ad models, outperforming them in CTR by significant margins.
Wil AdHeat appear on Google Plus? Maybe.
I find these comments of particular interest with respect to Q & A sites:
-People seldom provide answers without incentives.
-When incentives exist, people who might abuse or spam a system may emerge, degrade the quality of the service and discourage knowledgeable users from participating.
[RANT]I honestly do not see how this could be changed.
We all have to make a living. Why not reward contributors? It’s all UGC, and Google would make money off of it. It just seems like they want to make the money without anybody else making any. [/END RANT]
Wow Bill – that’s some pretty awesome research and a very compelling argument for where google could be heading.
Brilliant as always!
On a slightly different note – some of the Q & A’s on the Google Confucius system in the above screen grab are… interesting…
This service will definately make searches easier. But as someone said… google searches makes our memory weaker. True or false?
Great stuff man, this brings up so many questions… When will Google’s growth terror end?!?! They seem to be coming up with the best web applications on the web. First a new social button (+1), then a social network (Google+), they’ve started looking into Google Coupons more, and now this Confucius Q&A deal… All I can say about Confucius is this – this will undoubtedly become the #1 Q&A system, because they can establish experts in certain areas and can smartly redirect searches automatically to Q&A posts. Can’t wait to see it in action!!
For Google’s sake I don’t think they should be going down this path. They are going to end up like Yahoo circa 2003, with about 10,000 different apps, and no clear direction. Just looking at the Confucius system. It looks quite good, but is it going to add anything to Google’s core business?
Bill,
Thanks for your awesome in-depth research on these things. Confucius say, “Study the past if you would define the future.” We’ll see how they monetize Confucius+.
Dave
I would imagine the type of semantic data Google gets from this Q&A service is invaluable. Essentially they’re learning exact answers to questions, and overtime could use this data to score the accuracy of website content in the normal search results.
Great Research Bill, thanks!
Hey Bill,
Thanks for an interesting in-depth look at Confucius. Just the other day while searching Google for something, the two top results returned were Yahoo Q&A from over 3 years ago. This is a good indicator of how Google values UGC for direct questions on the Web.
I find the User profiling described in your sketch a bit disturbing, it will be interesting to see how this is implemented and what the entry level for “new authorities” will be.
Duru
Confucius seems ominous to yahoo answers, quora, stackoverflow and the like.. though SO is more focussed on programming, the first thing i do is to search google and then go to Stackoverflow for anything.. Imagine how they can completely get rid of all competition and make users not leave google property for anything..
Bill, this is a fantastic article, and you truly have a great grasp on the world of Global Social Media here.
What I am thinking is how many networks do we need? Further, Confucius is similar to Google+, or vice versa, so is this perhaps an example of taking a great idea and then twisting it enough just to escape criticisms? Or are the ideas just eerily similar.
It is getting tough to manage my accounts across so many channels…. how do you decide what channels to choose with so many options? I am mainly a LinkedIn person, but I would love to hear how others choose. Cheers!
Google’s behavior reflects its quintessential marketing strategy – to swoop down on established markets and grab chunks from the reigning giants. Along with Google’s social ambitions, Confucius is again an attempt to foray into the Q&A sphere. But given the history of Google’s (social) products and their shelf-lives, I hope it has some real intense innovations to boast of.
Fabulous article, as always.
Hi Tom,
Thank you. It was interesting digging into how Confucius works and the presentations and papers behind it.
Some of the questions on my screenshot are unusual. I grabbed some screenshots of the Q&A system in Chinese and Arabic, and from a few other countries as well, but decided that the English language version would provide a better sense of the different sections of the system, even with those questions.
Hi Mark,
I know that Google tried to provide incentives in the form of payments for people who provided answers when they ran Google Q&A in the US. I never used the service, but it was unsuccessful enough so that they discontinued it.
Hi Drew,
It depends on how you use Google. If you use it as a crutch, so that you don’t have to remember some things than it could potentially help make your memory weaker, much in the same way that speed dial on your phone might cause you not to bother to memorize telephone numbers of people whom you call frequently.
If you use Google as a tool to help you learn new things that you then memorize, then it might help strengthen your memory.
Hi Jon,
I’m not sure if we will see Confucius in the US. Google tried out Google Answers, and it failed here. In places like China and Russia, it seems like it’s working fine, and it has been for at least a couple of years.
I suspect that rather than bring Confucius to many other countries, it may be more likely tht Google will build something involving Q&A into Google Plus, and that Google Plus may fill that role as it is now.
Hi Robert,
I’m guessing that Google is feeling that way a little themselves. For example, the recent announcement from them that they will be closing down Google Labs, to “streamline” what they offer. Confucius seems to be working in the countries where it’s located now, but I think it makes more sense to take what they’ve learned from it and apply things like User Rankings to Google Plus.
Hi Dave,
Take a look at the AdHeat link in the conclusion section of my post. It looks like it was fairly successful when they experimented with it in Confucius. I suspect we will see something like it in Google Plus sometime in the future.
Hi Donnie,
Good points. They are getting sample answers from actual users to real questions, and they are also using information extraction approaches to come up with their own automated answers, and for both types of answers, they are getting user feedback.
Hi Duru,
Q&A answers and user generated content is a great way for a search engine to come up with content that people sometimes might have tried to find answers to questions through the search engine, and decided that they might get better answers if they asked someone.
In the right settings, I’m not sure if most people might be too bothered by being acknowledged as having some level of expertise on a specific subject, but I understand your apprehension.
Hi Kristina,
I’m not sure that Google really wants people to stay on Google. It makes as much sense for them to send people to other places as well. Chances are that if Google can help people find answers to questions, whether on other websites, or on a Google Q&A site, people will keep on coming back, and keep on using Google as their search engine.
Some types of questions are better asked on a Q&A site, and services like Stack Overflow, Quora, and so on provide people with an opportunity to get answers that may just not be available on the Web. So, it makes sense for Google to introduce Q&A where it might be helpful. I don’t think that in the countries where Confucius is available that they suggest that people ask a question on one of the Q&A sites that they run if there seems to be adequate pages responding to their queries in the search results.
Hi Chris,
Thank you.
Confucius has been around for a few years now, but Google’s Q&A service didn’t survive. I’m not sure that we will see another Q&A from Google, but a number of the ideas from Confucius might make their way into Google Plus.
I suspect that once Google comes out with business profiles for Google Plus, that it may become something of a rival for LinkedIn.
It’s possible that one of the incentives behind Google Plus is to replace a number of the many networks out there so that we don’t have to make as many choices. 🙂
Hi Tommy
Thanks. If Google does introduce a social Q&A site in other countries, I suspect that it will be through Google Plus rather than Confucius.
I give Google props for not only rising to the “challenge” of social networking & media but seeing it as an opportunity to actually increase their online footprint by still offering the most relevant and most refined ordering of data. They’re still bringing order to chaos!
Hi Matt,
I hope so. Getting rid of some services like Google Labs seems to point towards them trying to show some restraint. They’ve also made a few changes to Google Places, such as incorporating Hotpot into it, and focusing more upon their own reviews rather than those from other review sites. It also seems like plans for services like Google Plus have been in place for a while.
I think Google Confucius is great.But Google Plus has taken away its role and now Google Confucius is starting to lose its popularity. Hope Google would end up thinking of something more effective Q&A ideas to connect to the viewers.
Google acquired a start-up that does almost the same trick
Hello,
You might find it interesting to know that for America and Europe there is a similar Google project running, although not under the Google-flag of Confucius but under the name Aardvark, the name of a start-up company they acquired some time ago.
The name of the start-up is Aardvark and there website is : vark.com
To sign in you need either a Google or a Facebook account.
Hi negativeangel,
I agree that there’s a very real chance that the Google Confucius sites will lose users to Google Plus. I suspect that we might see Q&A capabilities built into Google Plus at some point in the future.
Hi Jan,
I was tempted to mention Aardvark in this post, and I probably should have. Thanks for bringing it up.
I have written a couple of posts about Aardvark in the past:
Google As a Social Search Engine: Aardvark Answers & Circle Posts in Google Search Results?
Yahoo’s Social Search Answer to Google’s Aardvark?
I’ve also tried it out a little as well. I suspect that it could find itself more fully integrated into Google Plus in the future as well.