X-Git-Url: https://www.fi.muni.cz/~kas/git//home/kas/public_html/git/?a=blobdiff_plain;ds=sidebyside;f=pan13-poster%2Fposter.tex;h=bfd9762556c9faa8c86de36a7761b656d1fab521;hb=fa03e2ee0827e56f549a32c1782cde605b062e60;hp=ce9f7a324da40f830b3ffbf36fe21c3f284e6839;hpb=c091d54abf1aa848d425ee061f66004bfaab69c5;p=pan13-paper.git

diff --git a/pan13-poster/poster.tex b/pan13-poster/poster.tex
old mode 100644
new mode 100755
index ce9f7a3..bfd9762
--- a/pan13-poster/poster.tex
+++ b/pan13-poster/poster.tex
@@ -4,9 +4,15 @@
 \usepackage{amsmath}
 \usepackage{amssymb}
 \usepackage{multicol}
-\usepackage{bera}
 \usepackage[utf8]{inputenc}
 %\usepackage{fancybullets}
+%\usepackage{floatflt}
+%\usepackage{graphics}
+\usepackage{fontspec}
+\usepackage{xunicode}
+\setmainfont[Mapping=tex-text]{DejaVu Sans}
+\setsansfont[Mapping=tex-text]{DejaVu Sans}
+\setmonofont[Mapping=tex-text]{DejaVu Sans Mono}
 
 \definecolor{BoxCol}{rgb}{0.9,0.9,1}
 % uncomment for light blue background to \section boxes 
@@ -18,7 +24,7 @@
 \definecolor{ReallyEmph}{rgb}{0.7,0,0}
 
 \renewcommand{\titlesize}{\Huge}
-\title{Distributed System \\ for Discovering Similar Documents}
+\title{Diverse Queries and Feature Type Selection for Plagiarism Discovery}
 
 % Note: only give author names, not institute
 \author{Å imon Suchomel, Jan Kasprzak, and Michal Brandejs}
@@ -45,10 +51,17 @@
   }
   { \end{itemize} }
 
-\conference{{\bf ICEIS 2008}, 12--16 June 2008, Barcelona, Spain}
+\conference{{\bf CLEF 2013}, 23--27 September 2013, Valencia, Spain}
 
 \setlength{\figbotskip}{\smallskipamount}
 
+\renewcommand{\SubSection}[2][?]{
+  \vspace{0.5\secskip}
+  \refstepcounter{subsection}
+  {\bf \subsectionsize \textcolor{SectionCol}{\arabic{section}.\arabic{subsection}~#2}}
+  \par\vspace{0.375\secskip}
+}
+
 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
 %%% Begin of Document
 
@@ -106,46 +119,216 @@
 \vspace{-.02\textwidth}
 
 %%% Begin of Multicols-Enviroment
+%\begin{abstract}
+%{\sffamily\itshape
+%Nějaký abstrakt.
+%}
+%\end{abstract}
+
+
+\begin{multicols}{2}\setlength{\columnseprule}{0pt}
+\section{Introduction}
+%
+A program for helping detering real-world plagiarism needs to accomplish many tasks.
+Original documents which served for creation of plagiarism must be retrieved and also suspicious passages according to
+input document must be highlighted. This poster presents methodology used during PAN2013 competition on uncovering plagiarism.
+
+The whole process is depicted at picture~\ref{fig:process}. The source retrieval task is divided into
+2 subtasks: Quering and Selecting, during which the software utilizes a given search engine. The retrieved
+sources must be examined in detail in order to highlight as many plagiarism cases as possible. This process is depicted
+as Text Alignment. Results of this process are called {\em detections}, i.e.~passages of {\em source document} and {\em suspicious document}, which are similar enough to each other, and can serve as a basis for further manual examination for possible plagiarism.
+%
+\vfill
+\columnbreak
+%
+\begin{figure}
+ \centering
+  \includegraphics[width=0.8\textwidth]{img/source_retrieval_process.pdf}
+  \caption{Plagiarism discovery process.}
+  \label{fig:process}
+\end{figure} 
+\end{multicols}
 \begin{multicols}{2}
+%\rm
+%%% Introduction
+\section{Querying}
+Querying means to effectively utilize a search engine in order to retrieve as many relevant
+documents as possible with the minimum amount of queries.
+%We consider the resulting document relevantif it shares some of text characteristics with the suspicious document.
+In real-world, queries as such represent appreciable cost, therefore their quantity minimization should be one of the top priorities. 
+%\subsection{Types of Queries}
+During initial phase, there were three diverse types of queries extracted from each suspicious document.\\
+\begin{minipage}{0.55\linewidth}
+\subsection{Keywords Based Queries}
+\begin{ytemize}
+\item TF--IDF base automated keywords extraction;
+\item 5-token long; 
+\item Deterministic;
+\item Non-positional;
+\item Non-phrasal.
+\end{ytemize}
+\end{minipage}
+\begin{minipage}{0.45\linewidth}
+\begin{figure}[h]
+ %\centering
+  \includegraphics[width=1\linewidth]{img/document_keywords.pdf}
+\end{figure}
+\end{minipage}
+\begin{minipage}{0.55\linewidth}
+\subsection{Intrinsic Plagiarism Based Queries}
+\begin{ytemize}
+\item Averaged Word Frequency Class based chunking~\cite{awfc};
+\item Random sentence selection from the chunk;
+\item Non-deterministic;
+\item Positional;
+\item Phrasal.
+\end{ytemize}
+\end{minipage}
+\begin{minipage}{0.45\linewidth}
+\begin{figure}[h]
+ %\centering
+  \includegraphics[width=1\linewidth]{img/document_awfc.pdf}
+\end{figure}
+\end{minipage}
+\begin{minipage}{0.55\linewidth}
+\subsection{Paragraph Based Queries}
+\begin{ytemize}
+\item Longest sentences from miscellaneous paragraphs;
+\item Deterministic;
+\item Positional;
+\item Phrasal.
+\end{ytemize}
+\end{minipage}
+\begin{minipage}{0.45\linewidth}
+\begin{figure}[h]
+ %\centering
+  \includegraphics[width=1\linewidth]{img/document_paragraphs.pdf}
+\end{figure}
+\end{minipage}
 
-\rm
-%%% Abstract
-\begin{abstract}
+\begin{figure}[h]
+ \centering
+  \includegraphics[width=0.8\linewidth]{img/queryprocess.pdf}
+   \caption{Stepwise queries execution process.}
+\end{figure}
+
+\section{Selecting}
+Document snippets were used for deciding whether to download the document for the text alignment.
+We used 2-tuples measurement, which indicates how many neighbouring word pairs coexist in the snippet and in the suspicious document.
+Performance of this measure is depicted at Figure~\ref{fig:snippet_graph}.
+Having this measure, a threshold for download decision needs to be set in order to maximize all discovered similarities
+and minimize total downloads.
+A profitable threshold is such that matches with the largest distance between those two curves.
+
+\begin{figure}
+  \centering
+  \includegraphics[width=0.8\textwidth]{img/snippets_graph.pdf}
+  \caption{Downloads and similarities performance.}
+  \label{fig:snippet_graph}
+\end{figure}
+%
+% Yenyova cast
+%
+\section{Text Alignment}
+
+The system uses the same basic principles as in \cite{suchomel_kas_12}:
+
+\begin{ytemize}
+\item{\cemph{common features} between source and suspicious documents}
+\begin{ytemize}
+\item{word 5-grams}
+\item{stop-word 8-grams \cite{stamatatos2011plagiarism}}
+\end{ytemize}
+\item{\cemph{valid intervals} of characters covered by common features
+	``densely enough''}
+\item{\cemph{postprocessing}---remove overlapping detections,
+	join neighbouring detections}
+\end{ytemize}
+
+\subsection{Alternative Features}
+
+\begin{ytemize}
+\item{\cemph{contextual n-grams} \cite{torrejondetailed}}
+\begin{ytemize}
+\item{\cemph{The quick} brown \cemph{fox jumped} over the lazy dogs.}
+\item{The \cemph{quick brown} fox \cemph{jumped over} the lazy dogs.}
+\end{ytemize}
+\item{plain word 4-grams}
+\begin{ytemize}
+\item{\cemph{The quick brown fox} jumped over the lazy dogs.}
+\item{The \cemph{quick brown fox jumped} over the lazy dogs.}
+\end{ytemize}
+\end{ytemize}
+
+\begin{table}
 
-{\sffamily\itshape
+\begin{center}
+\begin{tabular}{|l|r|r|r|r|}
+\hline
+\bf feature & \bf recall & \bf precision & \bf granularity & plagdet \\
+\hline
+plain      5-grams & 0.6306 & 0.8484 & 1.0000 & \cemph{0.7235} \\
+contextual 4-grams & 0.6721 & \cemph{0.8282} & 1.0000 & \cemph{0.7421} \\
+plain      4-grams & \cemph{0.7556} & 0.7340 & 1.0000 & \cemph{0.7447} \\
+\hline
+\end{tabular}
+\end{center}
 
-Nějaký abstrakt.
+\caption{Comparison of contextual 4-grams and plain word 4-grams}
+\end{table}
 
-}
-\end{abstract}
+\subsection{Global Postprocessing}
 
-%%% Introduction
-\section{Šimonova část}
-
-\subsection{Kdovíco}
+\begin{ytemize}
+\item{Similar to PAN 2010 \cite{Kasprzak2010}}
+\item{Overlapping detections removal}
+\item{\cemph{Result:} improvement, but not as significant as in 2010}
+\end{ytemize}
 
-\section{Yenyova část}
+%
+% Spolecna cast
+%
 
 \section{Conclusion}
 
-Nějaký závěr
+\subsection{Candidate retrieval}
 
-%%% References
+\begin{ytemize}
+\item{Second best ratio of recall to the number of queries}
+\item{Missing support for phrasal search in ChatNoir is a big stumbling block}
+\end{ytemize}
 
-%% Note: use of BibTeX als works!!
+\subsection{Text alignment}
 
-\bibliographystyle{plain}
-\begin{thebibliography}{1}
+\begin{ytemize}
+\item{Significant improvement against PAN 2013}
+\item{Word 4-grams are better than contextual 4-grams}
+\item{We need a better ranking system than plagdet!}
+\end{ytemize}
 
-\bibitem{ISMU}
-\cemph{Masaryk University Information System}\\
-{\tt http://is.muni.cz/}, contact: {\tt iscor@fi.muni.cz}.
+%%% References
 
-\bibitem{Theses}
-\cemph{Czech National Archive of Graduate Theses}\\
-{\tt http://theses.cz/}, contact: {\tt theses@fi.muni.cz}.
+%% Note: use of BibTeX als works!!
 
-\end{thebibliography}
+\bibliographystyle{plain}
+\bibliography{pan13-notebook}
+\nocite{awfc}
+
+%\begin{thebibliography}{1}
+%
+%\bibitem{ISMU}
+%\cemph{Masaryk University Information System}\\
+%{\tt http://is.muni.cz/}, contact: {\tt iscor@fi.muni.cz}.
+%
+%\bibitem{Theses}
+%\cemph{Czech National Archive of Graduate Theses}\\
+%{\tt http://theses.cz/}, contact: {\tt theses@fi.muni.cz}.
+%
+%\bibitem{AWFC}
+%\cemph{Sven Meyer Zu Eissen and Benno Stein: Intrinsic Plagiarism Detection}\\
+%{\tt Proceedings of the European Conference on Information Retrieval (ECIR-06)}, {\tt 2006}
+%
+%\end{thebibliography}
 
 \smallskip
 \hrule height .1em
@@ -153,14 +336,20 @@ Nějaký závěr
 
 % \sffamily
 
-QR kód?
 
+\hbox to \hsize{
+	{\hsize=0.5\hsize\vbox{
 \cemph{Contact information:}\\
-	Å imon Suchomel {\tt suchomel@fi.muni.cz},\\
-	Jan Kasprzak, {\tt kas@fi.muni.cz}.
-
+	Å imon Suchomel {\tt suchomel@fi.muni.cz}\\
+	Jan Kasprzak {\tt kas@fi.muni.cz}\\
+	{\cemph{\tt http://www.fi.muni.cz/\~{}kas/pan13/}}
+}
+	\hfill
+	{\hsize=0.4\hsize\vbox{
+	\includegraphics[width=\hsize]{qrcode.png}
+}}}}
+	
 
 \end{multicols}
 
 \end{document}
-