Social Media Research Blog

Crowdsourcing PC ads

Bill Gates, Seinfeld ads barely match up to any of the user-generated ads I have seen over at YouTube. IMHO, $300 Million would be better spent on crowdsourcing Microsoft's ad campaign. There are some really cool ads out there, here are a few:

Why would you pay $300 Million? Why bother if the whole Seinfeld ad is pulled off or not! -- when you can let your users champion your cause. 

Share the Love

What makes social media so compelling is the ability to share information with others. ShareThis is a cool widget I use on my blog to enable users to share the posts with others and to easily post it on different sites like digg, StubleUpon, FriendFeed, etc or via email. Recently ShareThis commissioned a study conducted by Forrester Research, that analyzed the sharing patterns across a number of users. Following is an excerpt from their blog post:

High level findings:

• Email is still primary channel for sharing -- 69 percent of adults cite email as the primary source of receiving information
• Less than one-third of online adults said they learned about the new content from generally shared content sources such as YouTube, a wiki, or social networking sites
• 64 percent of adults and 60 percent of youth still use the traditional cut and paste  method to share a URL or information
  • However this is not trackable by publishers, making sharing by email tough to measure
• Though the primary motivation for sharing differs, 81 percent of adults claim that, when cutting and pasting content to share, they share to help others - believing that a person will benefit from the information they share

Surprisingly, email is still the primary channel for sharing and what was more interesting is that people still do it the old fashioned cut-and-paste way. The neat thing about using a tool like ShareThis is that it gives you a "one-click" way to share something with your network via email, IM or on any social media site. I somehow think that email is not quite the right format for sharing links and interesting articles. I like email to be more about actionable items: but then that is just me. I think FriendFeed, Twitter or Facebook is more suited for sharing links and other cool stuff.

There are a number of interesting nuggets on sharing demographics and differences in sharing behavior based on gender. From the post:

• The motivation for sharing amongst men differs than those of women
  • In general, men are more likely to share something to appear cool or to look like an expert
  • Men more likely to share product recommendations and video than women; 77 percent of adult males and 74 of younger males shared news and web links
• Women are more likely to share products or ideas they like via easy or direct sharing methods
  • Women strongly favor send to a friend feature; more than 60 percent of adult woman use the send to a friend feature on websites
  • Women are more likely to share directly, especially via cell phone texting
  • Adult women tend to face more obstacles when trying to share content online, with 40 percent saying they could find no ways to share the information directly from the Web page

I read through parts of the report I had received from ShareThis -- lots of nice stats there. The full report can be requested from [email protected] (Unfortunately, I cannot post it directly here.)

Invited Talk KDD 2008: Internet Advertising and Optimal Auction Design

Following are a few running notes from the KDD invited talk "Internet Advertising and Optimal Auction Design" by Dr. Michael Schwarz from Yahoo! Research:

Dr. Schwarz  explains that the original advertising models were quite simple but now new tweaks and optimizations are being added to, for example, estimate click probabilities, match ads for long tail of queries, etc. So there is a "Fudge" factor in online advertising algorithms, and the question now is how to combine these correctly?

In sponsored search options space, Overture was really the pioneer of pay-per-click. In such a scheme, the bidder would pay whatever price that was fixed during the auction of an advertising position. However, soon people realized that generalized first auction is unstable. This is because if advertisers can keep changing their bids and the minimum bid keeps fluctuating. The situation is illustrated in the following paper and as shown in the image below:

This system was also prone to manipulation by automated robots quite easily.

Next, in 2002 Google came up with the Generalized Second-Price option (GSP): here the advertiser pays for the price fixed by the next highest bidder (plus a minor amount like a cent), upon winning an auction. Later Yahoo! followed. In this mechanism the assumption is that for the ad position i, clicks are almost independent of the advertisement. A clock shows the current price that keeps increasing, a bid price is fixed only at the time of dropping out and the final payments are computed according to GSP.  Dr. Schwarz presents how this is related to "Generalized English Auctions".  The intuition is that the advertisers values the price just at the point where the advertiser drops out of the game. If we think about it, at each point the advertiser is making a decision to go closer to the point where he values the position enough but not to a point where he/she would have regrets if the next highest bidder would drop out of the auction, causing a win. The GSP approach has been shown to have a unique equilibrium. For more details you can read the paper: "Internet advertising, generalized second price options".

Finally, the talk discusses an optimal matching problem. If you know the probability distribution from which the values for the bidders are drawn, in theory, the highest probable match (ie optimal auction) is deterministic. According to further studies in these advertising models, Dr. Schwarz suggests that despite multiple units the value for a slot, the auction is not dependent on the number of slots! This is great from a search engine's perspective. Although I guess this hold to a certain number of slots -- or else search results would have looked rather different! Further, the decline in CTR from one slot to another also does not have any influence on the auction. This is perhaps where clever data mining techniques can be applied to help estimate the probability distribution of bidder's values for each slot and  ultimately, lead to setting optimal reserve price. I guess this is perhaps the point where search engines start guarding their trade secrets closely.

Finally, during discussions with many people at the conference, on theme that kept coming up is that of social advertising and monetization of social networks -- which has been a typically difficult problem. Personally, the way I see this (more of my 2 cents here) is three fold:

• One, are you advertising for the purpose of targeting an individual (based on the profile information)?
• Two, are you targeting an individual based on the actions of her friends and neighbors in the social graph?
• Three, are you targeting a group or a community in the network?

KDD is an exciting venue precisely because it is at the confluence of data mining, social media, large scale text and graph analysis and monetization-- all the themes that have made it a great learning experience for me.

An Empirical Study of Dynamic Pari-mutuel Markets: Evidence from the Tech Buzz Game

Live Blogging from WebKDD 2008
An Empirical Study of Dynamic Pari-mutuel Markets: Evidence from the Tech Buzz Game
Yiling Chen, David Pennock, and Tejaswi Kasturi

Presented by Dr. Dave Pennock

Tech buzz game uses the wisdom of the crowds and people can 'bet' on events like: "Bird flu outbreak in UK in 2008" etc. An example of a prediction market that David talks about is InTrade and there are tons of other clones out there: Hollywood stock exchange etc. Pretty amazing some of the prediction accuracies that are shown in some of the examples.

Parimutuel markets: All the money goes into one pool and there is a (Zero sum) wagering game in play an winnings are shared equally. In dynamic parimutuel markets (DPM) the entry price changes as more people purchase the share (for either outcomes) in the pie. There is a cost function associated with each purchase price. C(q1,q) = sqrt(q1^2+q2^2). It has a nice property that the ratio of prices is same as ratio of shares.

This idea was used behind the Yahoo! Tech buzz game which was an amazingly cool site, if you remember. It still amazes me how much folks would do for fake money or points! Some research strategies presented for DPM as a prediction mechanism:

• Baseline: Implemented using robot traders buying and selling at random. Ofcourse the robot lost a lot of 'fake money'.
• Sophisticated vs. naive traders: myopic trading -- just decide based on immediate situation and not future.
• Crime pays: people open false accounts and move money from one account to another to add up the signup money.
• For 352 stocks prediction converges to the actual value/outcome closer to the cash out event.
• Markets were pretty accurate in determining the actual buzz (as measured using query logs)

Very cool! Watchout for a new prediction game called Yoopick, a sports prediction game! Here you can predict not just the outcome but the actual score ranges.

Advertising Models: From Contextual to Conceptual/Semantic to Social

Contextual advertising relies on matching an advertisement with a page based on its content. Most often advertisers bid on keywords and the ad platform finds the appropriate pages on which these ads can be displayed by matching keywords and phrases with the content. There have been a number of situations where such an approach may fail. A few real world examples are discussed in the paper by Broder et al. [2]:

a page about a famous golfer named “John Maytag” might trigger an ad for “Maytag dishwashers” since Maytag is a popular brand. Another example could be a page describing the Chevy Tahoe truck (a popular vehicle in US) triggering an ad about “Lake Tahoe vacations”. Polysemy is not the only culprit: there is a (maybe apocryphal) story about a lurid news item about a headless body found in a suitcase triggering an ad for Samsonite luggage! In all these examples the mismatch arises from the fact that the ads are not appropriate for the context.

These examples highlight the need for moving from a contextual to conceptual/semantic advertising models. The paper by Broder suggest mapping the pages as well as ads into a common ontology/taxonomy and thereby finding the appropriate higher level concept (Like politics/sports etc) that relates the two. There are very few papers on advertising models since this is such a closely guarded secret and a search companies' substantial revenue is tied to their ad platform's performance.

I believe that advertising is in it's infancy and more interesting approaches would soon replace the current state of the art. One problem is that due to the lack of datasets it becomes quite difficult to be in academia and make a significant contribution towards this area.

The next avenue for advertising seems to be Social advertising. While Facebook has its own approach to social advertising. I think the general idea of Social Advertising is to utilize not just the context of the page but also the social information to better place the advertisement. One question here is whether the ad placement is done to target the user or his/her audience. These might require slightly different models. For example, if my friends are all clikcing on the iphone ad on a social network, the platform might decide to also target ME personally for the marketing the iPhone. On the other hand if it can identify that a lot of users come to visit my profile due to the social media posts I write -- then perhaps the advertising could target them instead.

One potential market for advertising that I think is completely untapped is the referral. Companies are ready to pay huge sums of money to get new clients. Often cell phone companies, stock trading sites and banks launch promotions where they pay upwards of $50 for referring a friend. But when was the last time you actually did that? I think that is the best way to alienate your friends -- by hoarding corporate America's products and services or spamming their inboxes with unwanted referrals. But still, this is a huge market and worth billions -- if we can crack it! One approach I am thinking of is to build a referral platform (perhaps there are some out there -- I just dont know?) -- one which would benefit publishers and advertisers alike. I as a publisher have a (sort of) general sense of what my audience would like. I can for example even decide that I might be willing to share the $50 I receive from the advertiser and pass on the benefit to my readers (since my payoff is in having the readers come to my blog!) -- thus subsidizing that iPhone you wanted to buy. In the current model, there is'nt much incentive for me to share (a few cents???) /pass on the benefit with the final consumer. But for higher valued products, my guess is that It might just as well work right. Moreover, the referral platform manages the entire process thereby making it easier on the advertiser to launch new schemes and manage their inventory of referral programs.

I had intended to write a brief note on some of the recent papers [1-4] on this topic but turned out sharing my thoughts on the advertising instead -- which is perhaps more fun anyway :-).

[1] http://www.cs.cmu.edu/~deepay/mywww/papers/www08-interaction.pdf
[2] http://portal.acm.org/citation.cfm?id=1277837
[3] www.csulb.edu/web/journals/jecr/issues/20081/Paper1.pdf
[4] http://www2008.org/papers/pp231.html

Evri: Search Less, Understand More

I just received the beta invite to Evri.com (Yaaayy!). It is a really cool site that aims to help people find information. Right now they just have a browse interface. You can see what are the top concepts and named entities (primarily from News sources) and navigate through semantically related terms. The main idea behind their approach is that you can construct the graph of all the concepts and entities by analyzing the text. Here is an example of the top names in the news. Clicking the terms (from the graph) "Barack Obama" and "Ralph Nader" for example, would pull up all the stories related to recent controversies.

One can browse through the graph or the popular terms. I checked out what they found on Obama. Here is a snapshot on the left. I think that a really neat trick that Evri is using is the idea that working on sentence level semantics can provide sufficient meaning to help organize information. Constructing a complete parse tree that is both syntactically and semantically accurate is a difficult problem. There are many vagaries of natural language text that make this challenging. Evri, at least for now, bypasses some of these problems by organizing information around simple questions like "what is Obama doing?" which can have easy to identify clues directly accessible from the text (critisizing, leading, denying, facing....). Similarly for other entities like organizations one can ask "What is happening with Yahoo?" (bidding, reject, acquire, etc.). 

This is a fascinating approach to organizing information and I think that Evri has a great potential. Lets think about it for a minute. One of my favorite passtime is to go to Wikipedia, pull up a random article and then browse through related articles. It is this serendipity and the feeling of chance discovery of something interesting that is so compelling about Evri.

Evri also reminded me about the way I had hoped to implement SemNews, a semantic search engine, that analyzed RSS snippets of News articles and processed it through OntoSem, an ontological semantics based Natural Langugage Processing system. Once the semantics/meaning representations were extracted, I would store the meanings in an OWL store so that RDQL queries could be performed to find relevant news items. I believe that the way we can accomplish Dr. Tim Berners-Lee's vision of Semantic Web is by advancing both information extraction (web scraping, entity annotation etc) and NLP techniques that would automatically annotate text and make it available in machine readable format.

Although, the founder claims that they are not a search engine, they surely join the group of NLP-based startups like Powerset and Hakia. Another powerful tool is Freebase which uses primarily Wikipedia as its source of information. Finally, it is also worth mentioning that Kosmix is yet another startup that aims to "Organize the Web so that you can explore, learn and discover".

The next obvious question that comes up is regarding the monetization and business model of these startups. Ofcourse, the story goes... the information is more focussed so ads would be more relevant... and no surprise that is indeed so TRUE. Just check out some of the advertisements on Evri. On the left, is a screenshot of an advertisement on Barack Obama's info page. But I think there is an opportunity beyond simply relevant advertising!

Many companies have huge websites with lots of information -- some organized and most not quite as much. If you wanted to ensure that your customers are able to get to the exact information they need -- Evri like approach can be ideal to help them browse through the various facets to get to what they really need. The applications to Enterprises and Enterprise search can be another monetization platform for Evri. 

Finally, IMHO, some hurdles that Evri faces could be dealing with noisy text, especially with Social Media. Many approaches that rely on linguistic or gramatical correctness of sentences simply fail miserably when dealing with social media content. The second problem might be esuring coverage of information. Right now, it seems to me like the News soruces Evri relies on are primarily US centric. As they aim to capture more audience outside US as well, they would have to concentrate on foreign languages, disambiguating named entities and location names. These are all interesting research problems and fun stuff to work on!

These Tweets are from Mars

I am really enjoying the tweets from @MarsPhoenix. Ofcourse this isn't the actual robot sending Twitter updates from millions of miles away, the researchers tweeting on it's behalf are definitely engaging in some interesting conversations. This is one fantastic example of how large organizations can engage in Social Media.

The thought that we are having a conversation with a tiny bot makes the whole experience rather exciting. It wouldn't have been half as much fun if it were for a human persona at the other end. This "bot persona" is more lovable, in part due to our collective imagination and desire of being able to have an intelligent conversation with machines -- our R2-D2s and WALL*Es.

This has been a fantastic experiment in social psychology as well as a superb publicity approach.  MarsPhoenix has about 20K followers  making it one of the most popular Twitter users. Accolades to JPL researcher Veronica McGregor, for this terrific idea and posting interesting updates.

I think that there is a lot more to this story. I imagine that soon we will have more devices that we can talk to directly on Twitter and IM. One idea I had recently was to rig up our lab's coffee machine, Mr. Capresso, with a temperature sensor so that he can automatically inform us when fresh coffee is brewed in the lab.

And at the cost of sounding much like Eliza, I think that for a limited domain, we might even have the capabilities to build Natural Language Generation tools that could automatically post Tweets. I am aware that there are many bots on Twitter. But the tools I would like to see are the ones that can do more than just post a message (like a new video on qik, etc) -- true interaction would come only from conversations. A really wacky (but simple) example would be a poetic bot (yep! people have researched on that too! ;-) )that would send intelligible rhymes in response to @ messages. Might be quite hilarious to follow it!!

My First FireEagle App: Pizza Coupons Search

Yesterday, I discussed an idea around the FireEagle geolocation API. I was envisioning an app where you could have a mobile phone and as you walk down the Mall or any location, it would pre-fetch relevant coupons and offers from the local restaurants. Being a grad student, we always learn to find good Pizza deals online. So I decided to use the FireEagle API to develop a Pizza  coupon finder. The way it works is that it authenticates with FireEagle to access your current location and then fetches the coupons from Google Maps and then parses the output to display on your mobile phone or a browser. You can try it at the following URL  http://wikimatix.com/coupon/pizza.php
if you have a FireEagle account already. First the application will try to authenticate with FireEagle and request the appropriate permission to access the exact or approximate location information and then passes this to the Google Coupon Finder.

Finally you have all the coupons you need to order your fresh pizza. The Documentation and example walkthrough code on FireEagle's developer area is excellent. It took hardly any time to put together this demo!

I think that the possibilities that this opens up for mobile advertising are exciting. We should also keep an eye on Android -- this space is gonna be fun to watch. [Update: Fixed the broken link. Sorry]

Yahoo FireEagle: Geolocation made simple

This service is currently in alpha but thanks to Pranam Kolari I was able to get an invitation to Yahoo!'s FireEagle platform. FireEagle is an easy way to manage and share location information across many applications. Currently, I publish my location information across many different sites and applications and it is rare that I put in the actual effort to update it everywhere. For example I use Dopplr to publish my travel plans, twitter and Brightkite to update my current location and Facebook to indicate my home address and other details. I was impressed with how easy it was (using OAuth) to allow Dopplr and others to share and access information with FireEagle. If you have a GPS enabled phone you can even update the geolocation on the go! Damn! Thats is neat!

One really compelling application is Wikinear.com -- it shows you the nearest places of interest by matching the location information obtained from FireEagle with Wikipedia entries. This is great especially if you are traveling to a new location or a tourist spot and would like to know the places of interest nearby.

Another very cool application is Metosphere. (PS: I wish I had an iPhone!). With this app, you can leave a digital message for a given location, see places and events of interest and even report Graffiti and City Repair! This gives me a reason to believe that the next big thing is going to be mobile advertising. The advantage of easy availability of geolocation information specific to a user is immense. This reminds me of a project at eBiquity research group a few years back, called Agents2go,  that talked about a very similar concept. Imagine that you were walking down the during lunch and the agent on your iPhone would automatically collect coupons or find deals at the nearest restaurants as you walk by. The idea that we can have a query free, geographically relevant search is really exciting. Yahoo! is innovating and pushing hard on the open initiative. With the availability of an API it would be fun to integrate Google Coupons! (OK here is one more fascinating idea and little time at hand!)

Location is a very sensitive piece of information and the best part of FireEagle is that you can manage permissions and privacy settings or even temporarily stop sharing your location. You can allow a specific application to only access location information at a certain granularity: exact, zip, neighborhood, state or even country.  More at Techcrunch.

Samsung Viral Video Based Advertising

Check this out! It is a really cool video -- a two minute clip of 10 really cool optical illusions.

It is apparently sponsored by Samsung to promote their new soul phone model. Perhaps, they are trying to suggest "look at the kind of cool things you can do if you could take videos with our phone". In any case it is a really neat idea and definitely getting a lot of attention on the Blogosphere. If this is the new approach to advertising then I think it is something we will see more of. User generated advertising trend is on the rise. Check out this other crazy/funky video that could just as well be a levis ad

One thing is for sure - the Samsung video did grab by eyeballs and it was entertaining enough that I watched it a couple of times! Whoever said ads had to be boring 15 second "hooks" (avg length/increments in which TV advertising is sold) ?