Information for publishing at *ACL conferences

Generating proceedings for the ACL Anthology

The ACL Anthology requires a particular file layout in order to ingest material (described in Step 3 below). This README describes how to produce this layout from files exported from STARTv2 or EasyChair, or manually.

1. Installation and setup

First, clone the official repository:

git clone

2. Assembling the data

If you are using:

2a. Instructions for users of Softconf’s STARTv2

Create a file called start_urls.txt that contains a list of the START urls for all volumes associated with the conference. As an example, the start_urls.txt file in the current directory contains many of the NAACL 2015 volumes:

Next, download all the tarballs. You can use the provided script:

bin/ start_urls.txt

This automatic downloading is provided as a convenience; you could also do it manually (and may need to do so, if there are workshops that assemble their proceedings outside of START).

This downloads each track/workshop’s proceedings. The result should be something like the file structure in Step 2(c).

2b. Instructions for users of EasyChair

If you are using EasyChair, please see the instructions in the easy2acl repository. The documentation there describes how to assemble the Easychair output and run the script in order to generate a layout similar to what START exports. That code helps you produce a file format similar to Step 2(c).

2c. The file format

The ACLPUB scripts work from a conference organized in the following file format. A conference is one or more main conference volumes, along with zero or more workshop volumes. Each volume gets its own top-level directory underneath data/. For example:


Looking within each of these proceedings directories, we see the following format, which is the STARTv2 export format for a single volume:

  meta                               Information about the conference
    semeval-2018.bib                 BibTeX file containing entries for all papers
    semeval-2018.pdf                 PDF of whole proceedings
      2018.semeval-1.1_Software.tgz  Software attached to paper 1   Dataset attached to paper 3
      2018.semeval-1.3_Note.pdf      Note attached to paper 1003
      2018.semeval-1.0.bib           BibTeX entry for the whole proceedings volume
      2018.semeval-1.1.bib           BibTeX entry for paper 1
      2018.semeval-1.2.bib           etc.
      2018.semeval-1.0.pdf           PDF of frontmatter
      2018.semeval-1.1.pdf           PDF for paper 1
      2018.semeval-1.2.pdf           etc.

The “meta” file

Each meta file is just a collection of key/value pairs, one per line, with the key and value separated by whitespace. The lines of interest are (using SemEval as an example):

abbrev SemEval
volume 1
title 12th International Workshop on Semantic Evaluation
booktitle Proceedings of the 12th International Workshop on Semantic Evaluation
shortbooktitle Proceedings of SemEval
month January
year 2018
sig siglex
chairs Marianna Apidianaki
chairs Mohammad, Saif M.
chairs Jonathan May
chairs Ekaterina Shutova
chairs Steven Bethard
chairs Marine Carpuat
location Berlin, Germany
publisher Association for Computational Lingustics

The following fields are of special importance:

For those using START, these should have been set by publications and book chairs (Publication Console -> ACLPUB -> CDROM).

In addition, please ensure that the chairs lines are BibTeX-formatted names as above. If there are just two names, we can easily format it into BibTeX’s “family name, given name” format. If there are more than three names (e.g., “Mohammad, Saif M.”), please format this way yourself. Do not add affiliations as these are not used and they complicate parsing.

NOTE: If you are wondering what happened to the bib_url field, it is no longer used. If it’s present in START, you can ignore it.

3. Sanity check

After assembling your data to look like Step 2(c), please run the following script, which runs some basic sanity checks that may save you and the Anthology Director some grief.

./bin/ /path/to/data

4. Package and send the data/ directory

Package up the anthology directory:

tar czvf <conference>_data.tgz data

Upload the resulting file (<conference>_data.tgz) to a file server or cloud storage (e.g. Google Drive) and email the link to it to the Anthology Director. Please do not send the file as an email attachment.

Notes on supported attachment formats

Current attachment types supported by the Anthology are: