README.md

Literature Dialogue Response Task (LDR) 📚 •

⚠️ Disclaimer: this repository setups the task for the predefined train and valid splits. In order to replicate studies on different splits you have to manually update the related parts. We believe that ParlAI supports task initialization in Cross-Validation mode, however it goes beyond the capabilities of this project version.

This repository represent a supplementary material for the nicolay-r/book-persona-retreiver experiments organization 🧪 mentioneed in paper Personality Profiling for Literary Character Dialogue Agents with Human Level Attributes (pre-print) that has been accepted for Long Paper track at LOD-2024.

Adding task into ParlAI 🦜

You have to accomplish three steps below in order to start experiment with ParlAI 🦜 dialgue agents in the related task.

These steps are as follows:

👉 1. Add this entry into ParlAI task_list.py for registering this task:

{
    "id": "GutenbertBookChars",
    "display_name": "GutenbertBookChars",
    "task": "gutenbergbookchars",
    "tags": ["ChiteChat"],
    "description": (
        "Dataset of speaker utterances from ProjectGutenberg with their spectrums"
    )
}

👉 2. Follow the setup.sh to create folder GutenbertBookChars in the ParlAI project.

👉 3. Display dataset data in parlai/scripts/ to make sure that the task is available:

python display_data.py --task gutenbergbookchars

You can also froceed with the related notebook on GoogleColab

References

You can cite this work as follows:

@proceedings{rusnachenko2024personality,
  title     = {Personality Profiling for Literary Character Dialogue Agents with Human Level Attributes}
  authors   = {Rusnachenko, Nicolay and Liang, Huizhi}
  booktitle = {Proceedings of the 10th International Conference on Machine Learning, Optimization, and Data Science (LOD)},
  year      = {2024},
  month     = sep,
  days      = {22--25},
  address   = {Castiglione della Pescaia (Grosseto), Tuscany, Italy},
  publisher = {Springer}
}