presentation: adds last 2 slides and finalizes everything

diff --git a/img/model-overview.png b/img/model-overview.png
new file mode 100644 (file)
index 0000000..d471a38
Binary files /dev/null and b/img/model-overview.png differ

diff --git a/img/styleguide-example.png b/img/styleguide-example.png
new file mode 100644 (file)
index 0000000..c521acf
Binary files /dev/null and b/img/styleguide-example.png differ

diff --git a/presentation.tex b/presentation.tex
index c7c1a0a34a40f101728a80529511680250a4ef2d..98323436ac0b31a5080dd22a38aaf94828660741 100755 (executable)
--- a/presentation.tex
+++ b/presentation.tex
@@ -3,8 +3,8 @@
 % Meta informations:
 \newcommand{\trauthor}{Vincent Dahmen}
 \newcommand{\trtype}{Spottalk} %{Proseminar} %{Seminar} %{Workshop}
-\newcommand{\trcourse}{Evaluating domain control mechanism on NMT using  real life data sets}
-\newcommand{\trtitle}{Evaluating domain control mechanism on NMT using  real life data sets}
+\newcommand{\trcourse}{Evaluating a domain control mechanism on NMT using real life data sets}
+\newcommand{\trtitle}{Evaluating a domain control mechanism on NMT using real life data sets}
 \newcommand{\trmatrikelnummer}{6689845}
 \newcommand{\tremail}{4dahmen@informatik.uni-hamburg.de}
 \newcommand{\trinstitute}{Dept. Informatik -- Knowledge Technology, WTM}
@@ -39,7 +39,7 @@
 %\setbeamertemplate{sections/subsections in toc}[sections numbered]
 %\beamertemplatenavigationsymbolsempty
 \setbeamertemplate{headline}[default]  % deaktiviert die Kopfzeile
-%\setbeamertemplate{navigation symbols}{}% deaktiviert Navigationssymbole
+\setbeamertemplate{navigation symbols}{}% deaktiviert Navigationssymbole
 %\useinnertheme{rounded}
 
 \usepackage{acronym}                    % Acronyms

diff --git a/slides/01-motivations.md b/slides/01-motivations.md
index 60b2854084274e9d6b56b3bc8e3220981c9e2bf8..13bc6bd2dc7cd8d3918897555c96c3ff3db12b55 100644 (file)
--- a/slides/01-motivations.md
+++ b/slides/01-motivations.md
@@ -10,11 +10,11 @@
   2) multi source translation systems
   3) complex attention models
 
-## Why does it matter?
-- supportive facts are complex to learn 
-  - even for humans
-- evaluating a working approach with new data is necessary to prove it
-- collecting new datasets allows more precise evaluation of new models
+## Example Styleguide
+\begin{figure}[ht]
+       \centering
+       \includegraphics[scale=0.5]{img/styleguide-example.png}
+\end{figure}
 
 ## Research Questions
 - How can we transform real life/industry data sets to use in deep learning?

diff --git a/slides/03-Experiments.md b/slides/03-Experiments.md
index 6ef9535fb7948451954bd45cb474b20b1bfc3bf0..f6336802042ae33d136f706c733737208b17e0e6 100644 (file)
--- a/slides/03-Experiments.md
+++ b/slides/03-Experiments.md
@@ -1,5 +1,5 @@
 # Experiments and Evaluation
-
+## Experiments and Evaluation
 ### Experiment 
 - Train a neuronal net on multiple data sets using BPE and a fixed vocabulary
 - Reduced dataset
@@ -8,4 +8,9 @@
 ### Evaluation
 - BLEU
 - automated annotaion system by Sennrich
-- for industry data: distance to same translation of other category (if avaialble)
+- for industry data: distance to same translation of other category 
+
+## Model: FSMT
+\begin{figure}[ht]
+       \includegraphics[scale=0.8]{img/model-overview.png}
+\end{figure}

diff --git a/slides/04-Schedule.md b/slides/04-Schedule.md
index 441ca928fccc576c4e74cc1113e872c34c99db22..52323af1b390051a2bd517ffcc71d290d4c8e094 100644 (file)
--- a/slides/04-Schedule.md
+++ b/slides/04-Schedule.md
@@ -1,4 +1,7 @@
 # Time Plan
 ## Time Plan
-- Provide a sketch of your thesis schedule             
-- Identify reasonable time-slots for writing, experiments and implementation
+- Month 1: Prepare the data
+- Month 2: Implement the RNN
+- Month 3: Train the models
+- Month 4: Experiment with the models
+- Month 5: Write the thesis

diff --git a/slides/05-Contribution.md b/slides/05-Contribution.md
index 07e4299cc1d6a89ac10e1618d710c973b91bf203..acc932d8430b6523cedbbdc0cabcd7f55fbcb23b 100644 (file)
--- a/slides/05-Contribution.md
+++ b/slides/05-Contribution.md
@@ -1,4 +1,6 @@
 # Contribution
 ## Contribution
-- What would be the novelty in your approach           
-- What would be your contribution to the group?
+- supportive facts are complex to learn 
+  - even for humans
+- evaluating a working approach with new data is necessary to prove it
+- collecting new datasets allows more precise evaluation of new models