MAIN-ictir2017.tex

\documentclass{sig-alternate-05-2015}

\usepackage{mathabx} % for the math fonts in the stat table
\usepackage{times}
\usepackage{url}
\usepackage{epstopdf}
\usepackage{bbm}
\usepackage{dsfont}
\usepackage{tabularx}
\usepackage{multirow}  % to create muti-row tables
\usepackage{graphics}  % to resize large tables
\usepackage{mathabx} 	% for the math fonts in the stat table
\usepackage{url} % to embed clickable urls into the text
\usepackage{booktabs} % for formating the tables
\usepackage{caption} % for subfigure
\usepackage{subcaption} % for subfigure
\usepackage{algorithm,algorithmic} % this is to render the ILP formulation as a box
\usepackage[dvipsnames]{xcolor} % this is to add colour to the example
\usepackage{mathtools} % for intertext spacing
\usepackage[sort,square,compress,comma,numbers]{natbib}% for citations


%These commands are used in the statistical analysis table in the Experiment section.
\newcommand{\vvvinc}{$\blacktriangleup$}
\newcommand{\vvinc}{$\smalltriangleup$}
\newcommand{\vinc}{{\fontsize{2}{1}$\wedge$}}
\newcommand{\vvvdec}{$\blacktriangledown$}
\newcommand{\vvdec}{$\smalltriangledown$}
\newcommand{\vdec}{{\fontsize{2}{1}$\vee$}}

%change the lable of the algorithm to ILP
\makeatletter  % Add this!!!!
%\renewcommand{\ALG@name}{ILP} %doesnt work
\makeatother

\begin{document}
	\setcopyright{acmcopyright}

	\conferenceinfo{ICTIR '17}{}
	\title{Estimating Query Focus Time with Neural Word Embeddings}

	\numberofauthors{4}

	\author{
	    \alignauthor
		Supratim Das\\
		\affaddr{Max Planck Institute for Informatics\\ Saarbr{\"u}cken, Germany}\\
		\email{sudass@mpi-inf.mpg.de}
		\alignauthor
		Arunav Mishra\\
		\affaddr{Max Planck Institute for Informatics\\ Saarbr{\"u}cken, Germany}\\
		\email{amishra@mpi-inf.mpg.de}
		\alignauthor
		Klaus Berberich\\
		\affaddr{Max Planck Institute for Informatics\\ Saarbr{\"u}cken, Germany}\\
		\email{kberberi@mpi-inf.mpg.de}	
		\alignauthor
        Vinay Setty\\
        \affaddr{Aalborg Universitet\\ Aalborg, Denmark}\\
        \email{vinay@cs.aau.dk}
        \alignauthor
		}
		
	\maketitle

	\begin{abstract}
	A news event is primarily composed of three key components, a group of actors, one or more locations and a particular time-frame. These components are interwoven and they retain their complete semantic meaning only within their entire context. Certain event queries may miss one of the three key features, and still, make sense. Certain queries may also contain these three features and still be ambiguous.
	
    Focus time of an event-query is defined as the point of time at which a particular event has taken place, it is not the same as the time of creation or publication of the query.
    
    Temporal information is an essential characteristic of a news event. So, making use of temporal dimensionality in the text-based analysis is an interesting idea to explore. Events may be similar in temporal space, e.g. \textit{``The death of Carrie Fisher"} and \textit{``Trump wins the election"}. Events may be ambiguous in temporal space e.g. \textit{``Micheal Phelps wins gold"} and \textit{``Civil wars in Egypt"}. We try to solve this phenomenon by training a model on a huge news corpus and then evaluate on a set of historical event-queries.
	\end{abstract}

	% Keywords
	\keywords{Word2Vec, Temporal Embedding, Recursive Neural Networks, Temporal Information Extraction, Focus Time Extraction, Deep Learning}

	\input{Introduction}
	\input{RelatedWork}
	\input{Approach}
%	\input{Experiments}
	\input{Results}
	\input{Discussion}

    \bibliographystyle{abbrv}
	\bibliography{ICTIR2017Ref} 


	\balancecolumns

\end{document}