Summarization

Multi-Document Guided Summarization

Repository for the project component of the LING 575 - Summarization class.

Student team:

• Anna Batra
• Sam Briggs
• Junyin Chen
• Hilly Steinmetz

Getting started

Condor or Locally

• Install conda
• conda env create -f environment.yaml to initialize conda environment
• Put the downloaded the pytorch_model.bin model file from Hugging Face in outputs/

Evaluation

• As patas does not have the necessary package installed to run ROUGE-1.5.5pl, you will have to run the tool manually on your local device.
• We have provided the config file 'rouge_run_D3.xml' in src, please change the [REPLACE PEER ROOT] and [REPLACE MODEL ROOT] with the Peer Root and Model Root on your local machine.

Conda Guide

This guide is to set up Conda on Patas/Dryas. To download and use Conda:

1. wget https://repo.anaconda.com/archive/Anaconda3-2022.10-Linux-x86_64.sh

2. sh Anaconda3-2022.10-Linux-x86_64.sh

   This will modify your .bashrc which you can find in your user directory (~). If you didn't have one before, it will create one. Your modified .bashrc will include code similar to the following:

   # >>> conda initialize >>>
   # !! Contents within this block are managed by 'conda init' !!
   __conda_setup="$('/home2/[USER]/anaconda3/bin/conda' 'shell.bash' 'hook' 2> /dev/null)"
   if [ $? -eq 0 ]; then
       eval "$__conda_setup"
   else
       if [ -f "/home2/[USER]/anaconda3/etc/profile.d/conda.sh" ]; then
           . "/home2/[USER]/anaconda3/etc/profile.d/conda.sh"
       else
           export PATH="/home2/[USER]/anaconda3/bin:$PATH"
       fi
   fi
   unset __conda_setup
   # <<< conda initialize <<<

   Note: Files starting with . are often hidden. You can list those files with ls -a

3. rm Anaconda3-2022.10-Linux-x86_64.sh

4. source ~/.bashrc

5. conda env create -f environment.yml (A simple environment!)

6. Activate the environment with conda activate summarization. You're good to go!

   Note: If you use VSCode, you can select a new python interpreter with cmd+p and using the command "Python: select interpreter".

   A helpful list of commands to manage your Conda environment can be found here!

Troubleshooting

If you're running into issues sourcing the environment, you will want to move the code from the .bashrc information to a ~/.bash_profile file.

If you do not have a .bash_profile file, then use touch ~/.bash_profile to create one. Copy the code in your .bashrc file to the .bash_profile. Delete the code you just copied from .bashrc file. Finally, add the following code to your .bashrc file.

source ~/.bash_profile