deletions | additions       diff --git a/cc-intro.tex b/cc-intro.tex  index 0df8e6c..f0e66b4 100644  --- a/cc-intro.tex  +++ b/cc-intro.tex 
...
\subsection{Prior partial examples}  \textbf{[Jazz, recommender systems, HR.]} \paragraph{{[}To add: Jazz.{]}}  Research in recommender systems focusses on stimulating serendipity on the user side, by suggesting items that might more likely be unexpected, potentially novel, and necessarily of value to the user. Several components of the proposed model were implemented in related work, and art and music recommendation connects to the area of computational creativity. \paragraph{{[}To add: HR.{]}}  A recommender system must come with a prepared mind \paragraph{Recommender systems.} Research  in terms of full knowledge of recommender systems focusses on stimulating serendipity on  the user side, by suggesting  items in the search space, that might more be unexpected, novel,  and of valuable to  the user's partial knowledge user.   % Several components  oftheir existence and properties. Predictions are based on existing knowledge of the items known to  the user, proposed model were implemented in related work,  and his or her preferences. The system's goal is to recommend an item as result which is unexpected, valuable art  and potentially novel music recommendation connects  toa specific user, while  the user's goal is to find items that are valuable in different respects. area of computational creativity.  Related work tries to structure the search space and exploit patterns as serendipity triggers. For example, \cite{Herlocker2004} as well as \cite{Lu2012} associate less popular items A recommender system must come  with a higher unexpectedness. Clustering was also frequently used to discover latent structures \textbf{prepared mind} in terms of full knowledge of the items  in the search space. For example, \cite{Kamahara2005} partition users into clusters space, and  of common interest, while \cite{Onuma2009} as well as \cite{Zhang2011} perform clustering on both users the user's partial knowledge of their existence  and items. In properties. Predictions are based on existing knowledge of  the work by \cite{Oku2011}, items known to  the user user, and his or her preferences. The system's goal  isallowed  to select two items in order recommend an item as \textbf{result} which is unexpected, novel, and valuable  to mix their features a specific user. % while the user's goal is to find items that are valuable  in some sort of conceptual blending. different respects.  If a pattern is found, it is used Related work tries  to bridge between items that are known structure the search space  and valuable exploit patterns as \textbf{serendipity triggers}. For example, \cite{Herlocker2004} as well as \cite{Lu2012} associate less popular items with a higher unexpectedness. Clustering was also frequently used  to discover latent structures in  the user, and those that are potentially unexpected. As an search space. For  example, \cite{Sugiyama2011} connects \cite{Kamahara2005} partition  users with divergent interests, into clusters of common interest,  while \cite{Onuma2009} weight as well as \cite{Zhang2011} perform clustering on both users and items. In the work by \cite{Oku2011}, the user is allowed to select two  items stronger that bridge between topical clusters. in order to mix their features in some sort of conceptual blending.  If a pattern is found, it is used to \textbf{bridge} between items that are known and valuable to the user, and those that are potentially unexpected. As an example, \cite{Sugiyama2011} connects users with divergent interests, while \cite{Onuma2009} weight items stronger that bridge between topical clusters.  Recommender systems have to cope with a dynamic world \textbf{dynamic world}  of user preferences and new items that are introduced to the system. The imperfect knowledge about the user's preferences and interests represents perhaps the strongest dimension of chance. \textbf{chance}.  Determining the value \textbf{value}  of an item, both in terms of value for the user and unexpectedness, is of paramount importance. While standardized measures such as the $F_1$-score or the (R)MSE )are are  used to determine the accuracy of an evaluation in terms of preferred items in the user's history, there is no common agreement on a measure for serendipity yet \cite{ Murakami2008, Adamopoulos2011, McCay-Peet2011}.