\section{Introduction}
\label{sec:intro}

\begin{motivation}
The World Wide Web represents an essential factor in the dissemination of scientific work in many fields. At the same time, its exponential growth is reflected in the substantial increase of the amount of scientific research being published. As an example, in the biomedical domain, the well-known MedLine~\footnote{http://medline.cos.com/} now hosts over 18 million articles, having a growth rate of 0.5 million articles / year, which represents around 1300 articles / day~\cite{tsujii2009}. In addition, we can also mention the lack of uniformity and integration of access to information. Each event has its own online publishing means, and there is no central hub for such information, even within communities in the same domain. Consequently, this makes the process of finding and linking relevant work in a particular field a cumbersome task.
\end{motivation}

On the desktop, we can find a somewhat similar problem, though on a smaller scale. A typical researcher acquires (and stores) an significant number of publications over time. Generally, the files representing these publications have a non-intuitive name (often the same cryptic name assigned by the system publishing them), and may, in the best case scenario, be structured in intuitive folder hierarchies. Thus, finding a particular publication or links between the existing ones represents quite a challenge, even with the help of tools like Google Desktop~\footnote{http://desktop.google.com/}.

\begin{background}
Semantic Web technologies have been proved to help at alleviating, at least partially, the above mentioned issues. And at the foundation of the Semantic Web we find semantic metadata. Used in particular contexts, semantic metadata enables a more fertile search experience, complementing full text search with search based on different facets (e.g., one can search for a publication by a specific author and with some specific keywords in its title). In addition, subject to its richness, it can also leverage links between publications, e.g. citation networks.
\end{background}

Looking at the status of semantic metadata for scientific publications in the two directions, i.e. the Web and the Desktop, we observe the following. With the emergence of the Linked Open Data (LOD)~\footnote{http://linkeddata.org/} initiative, an increasing number of data sets were published as linked metadata. Regarding scientific publications, efforts like the Semantic Web Dog Food Server started by M\"{o}ller et al.~\cite{moeller2007} represent pioneering examples. The repository they initiated acts as a linked data hub, for metadata extracted from different sources, such as the International or European Semantic Web conferences, and now hosts metadata describing over 1000 publications and over 2800 people. The manual creation of metadata is their main drawback, as well as of the other similar approaches. Within the second direction, i.e. on the desktop, different Semantic Desktop efforts improve the situation, by extracting shallow metadata, either file-related (e.g. creator, date of creation), or even publication-related, such as title or authors. In conclusion, we currently have two directions targeting similar goals and having the same foundation:
\begin{inparaenum}[(i)]
  \item the $<$LOD --- semantic metadata$>$ bridge, linking publications on the web, and
  \item the $<$Semantic Desktop --- semantic metadata$>$ bridge, linking publications and personal information, on the desktop.
\end{inparaenum}

\claim{In this paper, we propose a solution for bridging the two directions, with the goal of enabling a more meaningful searching and linking experience on the desktop, having the linked data cloud as the primary source.} \claim{Our method consists of a three step process and starts from a publication with no metadata, each step carried out incrementally to enrich the semantic metadata describing the publication.} The three steps are:
\begin{inparaenum}[(i)]
  \item \emph{extraction} -- we extract automatically metadata from the publication based on a document-engineering oriented approach;
  \item \emph{expansion} -- we use the extracted raw metadata to search the linked data cloud, the result being a set of clean and linked metadata; and 
  \item \emph{integration} -- the metadata is further enriched by embedding it within the semantic desktop environment, where it is automatically linked with the already existing personal metadata.
\end{inparaenum}
\claim{The main results of our process are: a simple and straightforward way of finding related publications based on the metadata extracted and linked automatically, and the opportunity of weaving the linked publication data on the desktop, by means of usual desktop applications (e.g. file and Web browser).}

The remainder of the paper is structured as follows: Sect.~\ref{sec:scenario} introduces the scenario used for exemplifying our approach, while in Sect.~\ref{sec:approach} we detail the technical elements of each of the three steps in the process. Sect.~\ref{sec:evaluation} describes the preliminary evaluation we have performed, and before concluding in Sect.~\ref{sec:conclusion}, we have a look at related efforts in Sect.~\ref{sec:relatedwork}.