While in the early years of the Web, searching for information and keeping in touch used to be the two main reasons for 'going online', today we turn to the Web in many di erent situations, including when we look for entertainment to pass the time or relax. A popular tool to facilitate the users' desire for entertainment is StumbleUpon, which allows users to \stumble" through the Web one (semi-random) page at a time. Interestingly to us, many StumbleUpon users appreciate being served Wikipedia articles, which are informative pieces of text that educate the reader about a particular concept. The leisure activity of stumbling can thus also incorporate a learning experience. Since life-long learning is an important characteristic of knowledge economies, it is crucial to understand the interplay between these two - at rst sight - opposing forces. We hypothesize that a greater understanding of what makes certain Wikipedia articles more attractive to the serendipitously browsing user than others, will enable us to develop adaptations that expose a greater amount of Wikipedia articles to the leisure seeking user.
In the early years of the Web, searching for information and keeping in touch used to be the two main reasons for 'going online'. Today, we rely on the Web in increasingly diverse situations including shopping, consultations and learning. While these examples are all directed towards a particular goal the user has, we also turn to the Web at times when we simply want to be entertained to pass the time or relax. The possibilities for entertaining yourself on the Web are manifold, one can play games, listen to music, watch movies or simply browse through the Web in the hope of nding entertaining pages. Due to the sheer size of the Web though, random browsing is not e ective for discovering pages that may b interesting to the individual user. For this reason, a number of services have become popular that recommend web pages to users based on their interests. One popular tool to facilitate the users' desire for entertainment by serendipPresented at Searching4Fun workshop at ECIR2012. Copyright c 2012 for the individual papers by the papers’ authors. Copying permitted only for private and academic purposes. This volume is published and copyrighted by its editors. itous browsing is StumbleUpon1 (SU), which allows users to \stumble" through the Web one (semi-random) page at a time. Interestingly to us, many SU users appreciate being shown Wikipedia2 articles, which are informative pieces of text that educate the reader about a particular concept. The leisure activity of stumbling thus can also incorporate a learning experience, which might contribute to the development of novel ideas and lead to creative insights. Since life-long learning is an important characteristic of knowledge economies, it is crucial to understand the interplay between these two seemingly opposing forces (entertainment vs. learning). We hypothesize that a greater understanding of what makes certain Wikipedia articles more attractive to the serendipitously browsing user than others, will enable us to develop adaptations that expose a greater amount of Wikipedia articles to the leisure seeking user.
In this position paper we make an argument for the
importance of this task. We draw from a number of insights
gained in museum studies [
A rst experiment on the stumbled Wikipedia pages revealed that, just as in museums not all objects are equally attractive to visitors, not all articles are interesting to the average StumbleUpon user. In fact, only a very small number of Wikipedia articles gather a large number of views by SU users, most articles are rarely viewed. While we have no answer yet to the question of how to automatically classify articles according to their attractiveness to the serendipitously browsing user, we have developed a number of hypotheses which are outlined in Section 3.2.
If we assume for a moment that we are indeed able to develop such an approach, a number of application scenarios can be envisioned:
A qualitative study of the features that play a role in to trickling the interest of users who do not have an information need, will enable Wikipedia contributors to write their articles in a way that is more accessible to such users.
Wikipedia is available in many di erent languages and such a prediction method would allow us to bootstrap a recommender like StumbleUpon in di erent languages by adding an initial set of interesting, high quality pages before the critical mass of users is reached. 1http://www.stumbleupon.com/ 2http://www.wikipedia.org/ Outliers (articles with many 'Likes' but a low probability of being attractive) can be manually investigated to reduce spam. Or conversely, undiscovered articles are obtained and can be injected into the index. The passages that trigger the surprise or the attractiveness of an article can be identi ed and highlighted to the browsing user. This may help to keep those serendipitously browsing users engaged that initially only quickly scan the article.
E-learning applications can also bene t, as articles which are interesting to the casual reader can be found this way.
The rest of the paper is organized as follows: related work is presented in Section 2, followed by a preliminary analysis of stumbled Wikipedia pages (Section 3) and the conclusiosn (Section 4).
For this work, we draw inspirations from two areas. On the one hand we consider research into so-called educational leisure settings and free-choice learning which is a multidisciplinary eld that includes aspects from sociology, psychology and education. On the other hand, our work is also strongly related to serendipity.
Education leisure settings can be found in a wide range
of institutions including museums [
Serendipity, the act of encountering information nuggets
unexpectedly, has mostly been investigated in the context
of education [
Finally we note that di erent aspects of Wikipedia
articles have also been investigated in the past, though not
from a perspective of serendipitously browsing users. For
instance, in [
The usage of StumbleUpon is depicted in Figure 1. A user
\stumbles" pages with a simple click of the 'Stumble!' button
in his browser toolbar. In response, the user is presented
with a random page from the Web, biased according to his
user pro le or his friends' 'Likes'. The simplicity of the
system protects the user from information overload [
In all experiments we report here, we utilize the English Wikipedia dump enwiki-20111007 from October 2011. In a pre-processing step, we selected all Wikipedia articles that are neither redirects to other articles, nor new articles or explicit disambiguation pages and have a length of at least 500 characters (to remove stubs). In total, 3; 552; 059 articles remained.
In order to determine the popularity of Wikipedia articles in StumbleUpon, we randomly selected half of these Wikipedia articles and queried the StumbleUpon API for their number of views by SU users. Since SU is a recommendation engine, we can safely assume that the highly viewed pages are also highly popular and liked. We note, that the number of 'Likes' a page has received is not accessible through the StumbleUpon API. The information is accessible though at the SU meta-data page, which we manually checked for the results reported in Table 1.
Among the evaluated 1; 776; 029 articles, we found 267; 958 (15:13%) of them to be contained in the SU index. In our initial investigation, we also considered French and German Wikipedia which are two of the largest non-English Wikipedia repositories. However, we only found a very limited number of their articles in the SU index (in both cases less than 1%) and thus did not consider them further. Thus, an application scenario as proposed in the introduction (to bootstrap a recommender for a new language) is highly desirable.
Let us now focus on those articles that were submitted by Stumblers to the index. Figure 2 shows a scatter plot of the number of views versus the number of Wikipedia articles in the index. As can be expected, most articles have very few views (the median number of views is 10) while a small number of articles have gathered more than half a million views.
To give an impression of the type of articles that have gathered few or many views, Table 1 contains the ten most viewed Wikipedia articles in our data set as well as ten random examples of articles that were viewed one hundred times. We chose these two settings as they represent two extremes: on the one hand, articles that were viewed and also liked by a large number of people and on the other hand articles, that were shown a number of times but less well received by the SU users.
It should also be noted that the SU category Bizarre &
Oddities, which dominates the list of the ten most viewed
articles is not as prevalent when considering a larger set of
articles. In fact, the top 100 viewed articles in our data set
belong to 59 di erent SU categories: Bizarre & Oddities occurs
12 times, followed by the Writing category (5 times) and a
number of categories with three occurrences, including Arts,
Science and Linguistics. Only one of the top 100 articles was
a so-called featured article (indicating that previous work on
featured article prediction, e.g. [
While for a human reader it is usually not di cult to quickly judge whether an article is potentially interesting to him or not, it is a challenge to derive a method that automatically classi es articles accordingly. What exactly makes one article more interesting to the general public than another? In order to get get a rst understanding of what users think about the most viewed articles and possibly also why they like them, we analysed the comments that were posted on the SU info page for each of the ten most viewed Wikipedia articles. This analysis is very cursory, as compared to the number of views, very few users actually comment on an article, as commenting distracts from the 'stumbling' experience. For example, the article Wrap rage with 0.86 million views and forty-thousand likes has a 41 comments. In total, we analysed 479 comments and identi ed four broad categories: (A) Comments expressing surprise \There's a name for this?" \I'd never heard of this before (go StumbleUpon!).
Very cool." (B) Comments expressing admiration, sadness, sorrow, etc. \That's so sad" \No one should go through life afraid to take a walk." \don't know what to say actually.." (C) Comments about the usefulness of the knowledge \Simple, but helpful for designers." \An exceptional list of colours and their code, invaluable to graphic designers, webmasters etc." (D) Comments expressing negative sentiments towards the article \Fake." \Why stumble everyday wikipedia articles?" 3.2
Based on the preliminary qualitative insights gained, we developed three intuitions that we believe will enable us to predict to what a Wikipedia article is likely to be bene cial to the average SU user.
Intuition A. Articles that contain unexpected nuggets of information can be identi ed by considering how semantically related the article is to the other articles it contains links to. For instance, the List of unusual deaths Wikipedia article has, among others, outgoing links to the following diverse articles: Common g, Malvasia (wine), Eddystone Lighthouse, Hawaii, and Chimney. We hypothesize that nding such seemingly unrelated articles can be used as a measure of the likelihood of the article being of interest.
Intuition B. Articles that evoke emotional feelings can be discovered through a form of sentiment analysis. Although Wikipedia articles are written in a neutral style, some topics are bound to evoke emotions and those emotional topics can be identi ed.
Intuition C. Articles that contain useful knowledge may be identi ed indirectly, when considering their Talk pages, the amount of discussions that are ongoing and the style of the discussions. Articles about practically useful information are not likely to be emotionally charged, unlike discussions for instance about politicians, religious topics, etc.
We emphasize, that these are hypotheses that need to be veri ed in future work. 4.
In this position paper we have proposed to investigate what makes certain Wikipedia articles interesting to users who are browsing the Web without a goal in order to pass the time or relax. Since such articles are education to some degree, the leisure activity of browsing (stumbling) can thus also incorporate a learning experience. Since life-long learning is an important characteristic of knowledge economies, it is crucial to understand the interplay between these two Most viewed articles List of unusual deaths Flying Spaghetti Monster Wrap rage Shigeru Miyamoto Benjaman Kyle One red paperclip List of colors Do not stand at my grave and weep Fuel cell Raymond Robinson (Green Man)
SU Category Bizarre/Oddities
Satire Bizarre/Oddities
Video Games Bizarre/Oddities Bizarre/Oddities
Arts Poetry
Science Bizarre/Oddities
Example articles viewed 100 times Biblioscape Edge of chaos Gottfried Wilhelm Leibniz Prize Mario Buda Proto-Indo-European language Cisco Adler Biofeedback Ovipositor Concealer Winklepickers forces. We argue that a greater understanding of features are indicative of an article's attractiveness to the average user (stumbler) will enable us to develop adaptations that expose a greater amount of Wikipedia articles to the leisure seeking user.