trad stat in english

macros.tex

@@ -40,6 +40,9 @@
 \newcommand{\liumcyan}[1]{{\color{liumcyan} #1}}
 
 
+\newcommand{\bos}{\textnormal{<s>}}
+\newcommand{\eos}{\textnormal{</s>}}
+\newcommand{\unk}{\textnormal{<unk>}}
 
 \newcommand{\hs}{\hspace{1cm}}
 
@@ -97,7 +100,10 @@
 
 \newcommand{\ra}{$\rightarrow$}
 \newcommand{\Ra}{$\Rightarrow$}
+\newcommand{\la}{$\leftarrow$}
+\newcommand{\La}{$\Leftarrow$}
 
+\newcommand{\eg}{e.g.}
 
 %\DeclareMathOperator*{\argmax}{argmax}
 \newcommand{\argmaxx}{\operatornamewithlimits{argmax}}

macros_en.tex

@@ -37,6 +37,9 @@
 \newcommand{\liumcyan}[1]{{\color{liumcyan} #1}}
 
 
+\newcommand{\bos}{\textnormal{$<$s$>$}}
+\newcommand{\eos}{\textnormal{$<$/s$>$}}
+\newcommand{\unk}{\textnormal{$<$unk$>$}}
 
 \newcommand{\hs}{\hspace{1cm}}
 
@@ -107,6 +110,9 @@
 
 \newcommand{\ith}{$i^{\mbox{\tiny{th}}}$}
 \newcommand{\jth}{$j^{\mbox{\tiny{th}}}$}
+\newcommand{\ieme}{$i^{\mbox{\tiny{th}}}$}
+\newcommand{\jeme}{$j^{\mbox{\tiny{th}}}$}
+
 
 \graphicspath{{../}{../figures/}{./figures/}}
 %{../figures/figures.pivot}{../figures/figures.dataselection}}

trad_statistique/m2_trad_statistique.tex

@@ -52,28 +52,16 @@
 
 \usefonttheme[onlymath]{serif}
 
-\usepackage[english]{babel}
-\usepackage[utf8]{inputenc}
-\usepackage{times}
-\usepackage{epsfig}
-\usepackage{comment}
-\usepackage{url}
-\usepackage{multirow}
-\usepackage[T1]{fontenc}
-%\usepackage{natbib}
-%\usepackage{multimedia}
-\usepackage{booktabs}
 \usepackage{amsmath}
 \usepackage{amssymb}
 \usepackage{array}
 \setlength{\extrarowheight}{3pt}
-%\usepackage{xspace}
 
 \usepackage{algorithm}
 \usepackage[noend]{algpseudocode}
 \usepackage{multicol}
 
-\input ../macros.tex
+\input ../macros_en.tex
 \input ../macros_beamer.tex
 
 
@@ -89,12 +77,12 @@
 \subject{Statistical Machine Translation}
 \title[]{Statistical Machine Translation}
 
-\author[]{Loïc Barrault}
+\author[]{Lo\"ic Barrault}
 
-\institute[LIUM, Le Mans Université]
+\institute[LIUM, Le Mans Universit\'e]
 {
-  Loic.Barrault@univ-lemans.fr \\
-  Laboratoire d'Informatique de l'Université du Maine \\
+  loic.barrault@univ-lemans.fr \\
+  Laboratoire d'Informatique de l'Universit\'e du Maine \\
 }
 
 %\date{09 janvier 2017}

trad_statistique/mt_statistique.tex

 %!TEX root = m2_trad_statistique.tex
 
-\section{Statistical Machine Translation}
+%\section{Statistical Machine Translation}
 %\subsection{Introduction}
 
 
@@ -49,7 +49,7 @@
 \frametitle{Statistical MT}
 
 \centerline{
-  \includegraphics[width=0.50\textwidth]{figures/smt}
+  \includegraphics[width=0.50\textwidth]{figures/smt_en}
   }
 
 \end{frame}
@@ -66,6 +66,14 @@
 \item Statistical MT
 \end{itemize}
 \end{block}
+\vspace{\stretch{1}}
+\centerline{
+  \includegraphics[width=0.50\textwidth]{figures/triangle_vauquois}
+  }
+
+\centerline{
+ \tiny{Fig. Vauquois triangle.}
+ }
 \end{frame}
 
 
@@ -169,9 +177,9 @@
 \item Translate "He buys a book on machine translation"
 \item Examples in data base:
 	\begin{itemize}
-	\item \orange{He buys} an apple : \cyan{Il achète} une pomme
+	\item \orange{He buys} an apple : \cyan{Il ach\`ete} une pomme
 	\item I read \orange{a book on} statistics : Je lis \cyan{un livre sur} les statistiques
-	\item \orange{machine translation} is great! : \cyan{la traduction automatique}, c'est génial !
+	\item \orange{machine translation} is great! : \cyan{la traduction automatique}, c'est g\'enial !
 	\end{itemize}
 \end{itemize}
 \end{block}
@@ -226,7 +234,7 @@
 \begin{columns}
 	\begin{column}[T]{.45\textwidth}
 	\begin{itemize}
-	\item When I look at an article in Russian, I say: “This is really written in English, but it has been coded in some strange symbols. I will now proceed to decode.”
+	\item When I look at an article in Russian, I say: "This is really written in English, but it has been coded in some strange symbols. I will now proceed to decode."
 	\item Warren Weaver (1949)
 	\end{itemize}
 	\end{column}%
@@ -527,9 +535,9 @@ Total		& 1000	& 1
 \begin{frame}
 \frametitle{IBM model 1: example}
 
-\includegraphics[height=0.3\textheight]{figures/lex_1} \hspace{0.1 cm}
-\includegraphics[height=0.3\textheight]{figures/lex_2} \hspace{0.1 cm}
-\includegraphics[height=0.3\textheight]{figures/lex_3} \hspace{0.1 cm}
+\includegraphics[height=0.3\textheight]{figures/lex_1} \hspace{0.05 cm}
+\includegraphics[height=0.3\textheight]{figures/lex_2} \hspace{0.05 cm}
+\includegraphics[height=0.3\textheight]{figures/lex_3} \hspace{0.05 cm}
 \includegraphics[height=0.3\textheight]{figures/lex_4} 
 
 \begin{itemize}
@@ -1092,7 +1100,7 @@ $p(W) = \ds \prod_{i=1}^{T} p(w_i|h_i)$
 \end{itemize}
 \item Probabilities are independent from the position in the sentence
 \begin{itemize}
-\item add begin (<s>) and end (</s>) of sentence tokens
+\item add begin (\bos) and end (\eos) of sentence tokens
 \end{itemize}
 \item Probabilities are estimated using a large quantity of data (corpus), which are supposed to be {\bf well written}
 \end{itemize}
@@ -1116,7 +1124,7 @@ $p(W) = \ds \prod_{i=1}^{T} p(w_i|h_i)$
 \begin{itemize}
 \item Sequences that are not allowed by the language
 \begin{itemize}
-\item Ex.: "ils part tôt", "elle est beau" 
+\item Ex.: "ils part t\^ot", "elle est beau" 
 \end{itemize}
 \item Sequences that are not seen in the training corpus
 \end{itemize}
@@ -1226,7 +1234,7 @@ PPL & = & 2^H \mbox{~~~~~{\small [$H$ is the cross-entropy}]}\\
 \item[] use the outputs of the following command:  {\bf ngram -debug 2 -ppl ...}
 \item Compute perplexity, interpolated model: ngram
 \item[] {\bf -ppl <dev corpus>}: compute perplexity on development corpus
-\item[] {\bf -mix-lmK <lmK> -mix-lambdaK <coeffK>} : interpolate several models <lmK> with weights <coeffK> (K ranging from 0 to 9)
+\item[] {\bf -mix-lmK $<$lmK$>$ -mix-lambdaK $<$coeffK$>$} : interpolate several models $<$lmK$>$ with weights $<$coeffK$>$ (K ranging from 0 to 9)
 \end{itemize}
 \end{frame}
 
@@ -1272,7 +1280,7 @@ PPL & = & 2^H \mbox{~~~~~{\small [$H$ is the cross-entropy}]}\\
 \item 2st iteration: cover 2 source words
 \item etc.
 \end{itemize}
-\item Ex: \emph{la voiture de course rouge a été volée il y a trois jours}
+\item Ex: \emph{la voiture de course rouge a \'et\'e vol\'ee il y a trois jours}
 \item[]
 \centerline{  \includegraphics[width=0.8\textwidth]{figures/decoder_stacks_en}  }
 \item Scores are propagated along with the hypotheses
@@ -1293,9 +1301,9 @@ PPL & = & 2^H \mbox{~~~~~{\small [$H$ is the cross-entropy}]}\\
 \only<1>{ \begin{itemize}
 \item Hypotheses leading to the same state but with a lower score can be stopped
 \begin{itemize}
-\item Ex.: \emph{la voiture de course rouge a été volée il y a trois jours}
-\item 1st hypo: "la voiture de course rouge" -> "the red race car", using only 1 phrase pair (seen in phrase-table)
-\item 2nd hypo: "la voiture de course rouge" -> "the red race car", using 3 phrase pairs  "la"->"the", "voiture de course"->"race car", "rouge"->"red"
+\item Ex.: \emph{la voiture de course rouge a \'et\'e vol\'ee il y a trois jours}
+\item 1st hypo: "la voiture de course rouge"\ra"the red race car", using only 1 phrase pair (seen in phrase-table)
+\item 2nd hypo: "la voiture de course rouge"\ra"the red race car", using 3 phrase pairs  "la"\ra"the", "voiture de course"\ra"race car", "rouge"\ra"red"
 \item[\ra] Source word coverage and partial hypotheses identical \Ra\ keep only the one with best score
 \end{itemize}
 \end{itemize} }