ConvoSumm

Data, code, and model checkpoints for the ACL 2021 paper ConvoSumm: Conversation Summarization Benchmark and Improved Abstractive Summarization with Argument Mining!

Data

The data can be accessed from this Google Drive link.

The data-non-processed contains the original, non-processed data and is 27MB, while data-processed contains the data for vanilla, -arg-filtered, and -arg-graph experiments, as well as model outputs, and is 611 MB.

Using the gdrive cli, download the folders with the following command

gdrive download --recursive 1HfyCMa1fQ5DkzME9RQZkytZQfyDjE1EK

The data can also be downloaded from this S3 bucket.

aws s3 cp --recursive s3://convosumm/data/ ./data

Code and Model Checkpoints

Please see this README for code details.

Model checkpoints can be downloaded from the S3 bucket (~80GB) or from this Google Drive link:

aws s3 cp --recursive s3://convosumm/checkpoints/ ./checkpoints