Align then Summarize: Automatic Alignment Methods for Summarization Corpus Creation

About

This repository contain code of the paper Align then Summarize: Automatic Alignment Methods for Summarization Corpus Creation with instructions for reproduction.

We also provide public_meetings a novel corpus of meetings (with pairs of transcriptions and reports).

Cite the paper:

@inproceedings{Tardy2020,
author = {Tardy, Paul and Janiszek, David and Est{\`{e}}ve, Yannick and Nguyen, Vincent},
booktitle = {Proceedings of the 12th Conference on Language Resources and Evaluation (LREC 2020)}, 
keywords = {Alignment,Corpus Annotation,Summarization},
pages = {6718--6724},
title = {{Align then Summarize: Automatic Alignment Methods for Summarization Corpus Creation}},
year = {2020}
}

Getting Started

1) Clone the repo (along w/ sub-repositories for external models):

git clone https://github.com/pltrdy/autoalign --recursive

2) Install dependencies:

python setup.py install

Automatic Alignment

Using parameter sets

The ./experiments directory contains parameter sets of our experiments.
We can use specific parameter set (based on their ids) using ./align_mapping.py (to just align) or ./validate_exp.py (to align and compare with reference):

So-called mapping are just json formatted definitions of meetings file paths, e.g.

[
    {
        "dir": "/abs/path/to/dir",
        "mapping": [
            {
                "doc": "doc_name.docx"
                "ctm": [
                    "first_ctm_name.ctm",
                    "second_ctm_name.ctm",
                ]
            }, {
                ...
            }
        ]
    }, {
        ...
    }
]

You can either specify a mapping.json path, or use preset mapping by name e.g. public_meetings.

out_dir="./exp_embeddings"
mkdir -p "$out_dir"

./align_mapping.py \
    -param_path ./experiments/exp_embeddings_decay.json \
    -mapping_name "public_meetings" \
    -out_dir $out_dir \
    -prefix "public_decay" \
    -n_thread 2 \
    -id 146

To evaluate against a set of reference alignment, use the similar validate_exp.py:

out_dir="./exp_embeddings"
mkdir -p "$out_dir"

./validate_exp.py \
    -path ./experiments/exp_embeddings_decay.json \
    -mapping_name "public_meetings" \
    -root $out_dir \
    -prefix validate_public_decay \
    -aligned_name "public_meetings" \
    -n_thread 2 \
    -ids 146 140

New experiment

In short, we define grid-search parameters in exps.py and run it by specifying the experiment id to e.g. ./multi_exps.py $id

External models

External models, namely TextTiling, C99 (+ preliminary work with Alemi's Segmentation) can be found under autoalign/external