| title | author | date | header-includes | classoption | |
|---|---|---|---|---|---|
Title of the paper \
venue, year
|
Originally by A, B, C \
Presented by X, Y, Z \
for NLP class at Iowa State University
|
\today |
\usepackage{amssymb,mathtools,blkarray,bm}
\usefonttheme[onlymath]{serif}
\usepackage[vlined,algoruled,titlenotnumbered,linesnumbered]{algorithm2e}
\usepackage{algorithmic}
\setbeamercolor{math text}{fg=green!50!black}
\setbeamercolor{normal text in math text}{parent=math text}
\newcommand*{\vertbar}{\rule[-1ex]{0.5pt}{2ex}}
\newcommand*{\horzbar}{\rule[.5ex]{2ex}{0.5pt}}
\setlength\labelsep {.5pt}
\setbeamersize{text margin left=3mm,text margin right=4mm}
\setlength{\leftmargini}{15pt}
\usepackage{hyperref}
\hypersetup{
colorlinks=true,
linkcolor=blue,
filecolor=magenta,
urlcolor=cyan,
}
\setlength{\abovedisplayskip}{1pt}
\setlength{\belowdisplayskip}{1pt}
\usepackage{graphicx}
|
|
- Provide information for an average CS graduate student to know the problem solved by the paper
- Use the language that a documentary would use to explain a scientific concept to general audiences.
- Example: Motivation section of https://github.com/tensorflow/docs/blob/r1.15/site/en/tutorials/representation/word2vec.md
- A paper usually goes over related work. If the paper you are assigned to does not, then you can skip this part.
- How other researchers approached the problem before the authors.
- Explain their main ideas in simple English and short equations.
- Pros and cons of their approaches.
- List the authors' new ideas in bullet points in a language that an average CS graduate student can understand
- If needed, use short equations and illurations (like Figure 1 of Mokolov et al.'s word2vec paper) to give people a high level idea.
- Math can be your friend because it can be concise and clear than English in many cases, e.g.,
$E=mc^2$ . - Use examples if they help explain the authors' idea.
- Example: "Scaling up with Noise-Contrastive Training" section abd "The Skipgram model" section of https://github.com/tensorflow/docs/blob/r1.15/site/en/tutorials/representation/word2vec.md
- No need to list all results reported in the paper.
- List those most important or surprising.
- Include qualitative results, which do not take your audiences much time to digest, like Figure 3 in NEURAL MACHINE TRANSLATION BY JOINTLY LEARNING TO ALIGN AND TRANSLATE at ICLR 2015.
{width=400px}
- Future work listed by the authors at the end of the paper
- Future work according to papers citing the work
- Future work according to you
- We have 35 teams in total.
- Each team has 15 minutes to present and 5 minutes for Q&A.
- Remember: let your presentation be a trailer of the paper.
- The slides can be generated via Pandoc in the command
pandoc -t beamer presentation_template.md -o presentation_template.pdf