CALAMR: Component ALignment for Abstract Meaning Representation

This repository contains code and data the paper CALAMR: Component ALignment for Abstract Meaning Representation. This code is used to align the components of a bipartite source and summary AMR graph. The results are useful as a semantic graph similarity score (like SMATCH) or to find the summarized portion (as AMR nodes, edges and subgraphs) of a document or the portion of the source that represents the summary.

Inclusion in Your Projects

The purpose of this repository is to reproduce the results in the paper. If want to align AMR graphs for your own work, please refer to the zensols.calamr repository, which has reusable code and examples. If you use this library or the PropBank API or PropBank curated database, please cite our paper.

Documentation

The recommended reading order for this project:

1. The conference slides
2. The abstract and introduction of the paper CALAMR: Component ALignment for Abstract Meaning Representation
3. Overview and implementation guide
4. Full documentation
5. API reference

Reproducing the Results

To reproduce the results from the paper, first process the corpus. These next steps create the document summarization and parser metrics.

Preprocessing the corpus with the following steps:

1. Install a Python 3.10.8 virtual environment on Linux. Note this version of the code assumes Linux, but new version does not.
2. Clone this repository: git clone https://github.com/uic-nlp-lab/calamr
3. Enter the repository and create release directory that corpora to be installed: cd calamr && mkdir download
4. Download the AMR Release 3.0: cp .../path/to/download/amr_annotation_3.0_LDC2020T02.tgz download
5. For reproducing the results that compare with earlier work on the AMR Release 1.0 corpus, place that corpus file in download directory as well.
6. Install the environment: ./bin/install.sh <path to Python home directory>. If you use conda, create a new conda 3.10.8 environment and set it to the Python home directory it creates (not including the bin/python3 directory)
7. Check the previous step to make sure it successfully creates new Python environment in directory pyenv. Also make sure it clones the amr_coref repository, and applies the patch successfully.
8. Create the sentence type/align merged corpus file: ./bin/prep.sh mergeanons
9. Create the mismatch corpus (please contact the authors for the original corpus file used in the experiments as the random seed was not set): ./bin/prep.sh mismatchcorp
10. Create the parser output of the corpora: ./bin/prep.sh parsecorp
11. Create the JAMR output for the corpora. This is a manual process, which includes downloading and installing the JAMR parser. We created this file manually, but will provide it for requests that include proof of purchase of the AMR Release 3.0 corpus.
12. Score documents and pairs (document table): ./bin/prep.sh score
13. Align documents: ./bin/prep.sh align
14. Output alignment statistics: ./bin/prep.sh alignstats

To recreate the example diagrams from the paper

The micro corpus are short examples for illustrating the alignment algorithm via component diagram. You can add your own sentences to the AMR parser input and rerun the micro corpus create an align steps below.

Follow the steps to creating the virtual environment in result reproduction section, and then:

1. Create the AMR micro corpus: ./bin/micro.sh createcorp
2. Align the micro corpus graphs: ./bin/micro.sh align

Attribution

This project, or reference model code, uses:

• Python 3.10
• amrlib for AMR parsing.
• amr_coref for AMR co-reference
• zensols.amr for AMR features and summarization data structures.
• Sentence-BERT embeddings
• zensols.propbankdb and zensols.deepnlp for PropBank embeddings
• zensols.nlparse for natural language features and NLP scoring
• Smatch (Cai and Knight. 2013) and WLK (Opitz et. al. 2021) for scoring.

Citation

If you use this project in your research please use the following BibTeX entry:

@inproceedings{landes-di-eugenio-2024-calamr-component,
    title = "{CALAMR}: Component {AL}ignment for {A}bstract {M}eaning {R}epresentation",
    author = "Landes, Paul  and
      Di Eugenio, Barbara",
    editor = "Calzolari, Nicoletta  and
      Kan, Min-Yen  and
      Hoste, Veronique  and
      Lenci, Alessandro  and
      Sakti, Sakriani  and
      Xue, Nianwen",
    booktitle = "Proceedings of the 2024 Joint International Conference on Computational Linguistics, Language Resources and Evaluation (LREC-COLING 2024)",
    month = may,
    year = "2024",
    address = "Torino, Italy",
    publisher = "ELRA and ICCL",
    url = "https://aclanthology.org/2024.lrec-main.236",
    pages = "2622--2637"
}

License

MIT License

Copyright (c) 2023 - 2024 Paul Landes