\documentclass[a4paper]{article}
\usepackage[english]{babel}
\usepackage[utf8x]{inputenc}
\usepackage{amsmath}
\usepackage{graphicx}
\usepackage[colorinlistoftodos]{todonotes}
\textheight 8.5in \topmargin -0.5in
\title{Evaluation}
\author{Adam Poliak — apoliak1 : David Russell - drusse19}
\begin{document}
\maketitle
\section{METEOR}
\subsection{METEOR Implementation}
As instructed in the assignment, we implemented the METEOR in the script \texttt{./evaluation} by calculating and using precision and recall as described in the assignment. The implementation is contained in the method \texttt{meteor}
\subsection{Running METEOR}
To run our implementation of METEOR, \texttt{./evaluation} must be called with the $\alpha$ flag. To add an $\alpha$ parameter, use the \textbf{-a} flag followed by a decimal number to represent the $\alpha$. If the \textbf{-a} flag is not provided, then the baseline implementation of \texttt{./evaluation} provided will be ran instead.
\subsection{$\alpha$ tuning}
After running \texttt{./evaluation} $100$ times, we determined that the best alpha on the dev data is $\alpha$ = $.87$ resulted in an accuracy of $0.5045$
\newline
\newline The graph below shows the results from tuning alpha. The x-axis represents the alphas and the y-axis represents the accuracy for corresponding alphas used in METEOR.
\newline
\newline
\includegraphics[width=\textwidth]{alpha_tuning.jpg}%\textwidth]
\newline
\newline The data in the graph was generated by running the script \texttt{test\textunderscore alpha} to determine the best $\alpha$. This script takes about $5$ minutes to run.
\section{WordNet Synonyms}
\end{document}