AMR-Parsing-Summarization: A Graph-to-Text Framework

This repository contains a modular, configurable implementation of the abstractive text summarization framework presented in the paper: "Text summarization based on semantic graphs: an abstract meaning representation graph-to-text deep learning approach" (Kouris et al., 2024).

The project provides a pipeline for generating abstractive summaries by first parsing text into Abstract Meaning Representation (AMR) graphs, and then using a deep learning model to generate a summary directly from this semantic representation.

(Example of a styled AMR graph generated by this project)

---

Project Structure

The framework is organized into a modular structure for clarity and extensibility:

AMR-Parsing-Summarization/
│
├── config/
│   └── config.py           # Central configuration for paths, models, and hyperparameters
│
├── data/
│   ├── data_loader.py      # Handles data loading, AMR parsing, and graph transformations
│   └── train.csv           # (Example) Training data
│
├── models/
│   ├── as2sp.py            # Attentive Sequence-to-Sequence with Pointer-Generator
│   ├── trce.py             # Transformer with (frozen) Contextual Embeddings
│   ├── petr.py             # Pre-trained Encoder Transformer (fine-tuned)
│   └── rl.py               # Reinforcement Learning (self-critical) components
│
├── main.py                 # Main entry point to run training or generation
├── train.py                # Unified training script
├── generate.py             # Script to generate summaries with a trained model
└── README.md               # This file

---

Features

• Multiple Models: Implements several models from the paper: AS2SP, TRCE, PETR, and an RL training scheme.
• Configurable Pipeline: Easily switch between models, data schemes, and hyperparameters via a central config file.
• All Data Schemes: Supports all graph construction (sequence, combination) and transformation (OAMR, OAMRWS, SAMR, SAMRWS) methods described in the paper.
• Modular and Extensible: The clean structure makes it easy to add new models, datasets, or evaluation metrics.

---

Setup and Installation

1. Clone the Repository

git clone https://github.com/MasihMoafi/AMR-Parsing-Summarization.git
cd AMR-Parsing-Summarization

2. Environment and Dependencies

A Conda environment is recommended.

# Create and activate the environment
conda create -n amr_env python=3.10
conda activate amr_env

# Install dependencies using uv (or pip)
pip install uv
uv pip install torch pandas amrlib==0.8.0 penman==1.3.1 transformers tqdm

3. System-level Dependencies

The graphviz library requires a system-level installation.

• Ubuntu/Debian: sudo apt-get update && sudo apt-get install graphviz -y
• macOS (Homebrew): brew install graphviz

4. Download AMR Parser Model

Download the pre-trained AMR parsing model (model_parse_xfm_bart_large-v0_1_0.tar.gz) and place it in the project root. Then, extract it.

# In the project root directory
tar -xzvf model_parse_xfm_bart_large-v0_1_0.tar.gz

5. Configure Paths

Open config/config.py and ensure all paths (e.g., DATA_PATH, MODEL_PATH) are correctly set for your system.

---

How to Run

Use the main entry point main.py to either train a model or generate a summary.

Training

To train a model, specify the train mode. You can override any setting from the config file using command-line arguments.

# Train the model specified in the config file (e.g., AS2SP)
python main.py train

# Train a different model, for example, PETR
python main.py train --model PETR

The script will use the settings in config/config.py to load the correct data, build the model, and start the training process.

Generation

To generate a summary with a trained model, specify the generate mode. You must provide the input text and the path to your saved model.

# Example of generating a summary
python main.py generate \
    --model AS2SP \
    --model_path "path/to/your/saved_model.pt" \
    --text "The boy wants to go to the store and buy some food."

---

License

This project is licensed under the MIT License.